WordCamp Asia 2026 - Conference Day 1 (Growth)

WordCamp Asia 2026 - Conference Day 1 (Growth) 🔗 **Source:** [WordCamp Asia 2026 - Conference Day 1 (Growth)](https://www.youtube.com/watch?v=jDsozp7y8bk) > [!note]+ Document Notes > tags: #wordpress #ai #llm #open-source #web-development #agentic-web #content-management > author: James LePage > source: WordCamp Asia 2026 > reference: https://www.youtube.com/watch?v=jDsozp7y8bk > date created: 2024-12-19 10:30 > > # 🏷️ Title: WordPress and AI: Building an Open Agentic Web >  ## 🧰 Action Inbox - [H1] Update to WordPress 7.0 beta to test AI building blocks — WordPress 7.0 #wordpress - [H2] Install WordPress AI plugin and experiment with abilities API — AI integration #ai - [H3] Point Claude at your site using MCP to test agent interactions — MCP adapter testing #llm ## 🧭 Table of Contents - [[#📝 Summary]] - [[#📖 Main Content]] - [[#🗒️ Notes]] - [[#📚 References]] - [[#🗣️ Paraphrase]] - [[#📰 Newsletter]] - [[#✅ Action Items]] - [[#🧪 Hypotheses]] - [[#📊 Methodology]] - [[#❓ Questions]] - [[#🔮 Future Research]] - [[#💡 Implications]] - [[#🧠 Learning Styles Assessment]] - [[#📊 Accessibility Assessment]] - [[#🌍 DEIB Assessment]] - [[#✅ Logical Validity]] - [[#🛡️ DARVO Analysis]] - [[#🎯 Conclusion]] ### 🔗 References - [WordPress AI Plugin](https://wordpress.org/plugins/ai/) - [Make WordPress AI Team](https://make.wordpress.org/ai) ## 📝 Summary - WordPress 7.0 introduces four AI building blocks to prepare WordPress for an "agentic web" where AI agents interact directly with websites - The building blocks include: Abilities API, WP AI Client, MCP Adapter, and AI Plugin for WordPress - WordPress's 43% web share positions it as crucial infrastructure for determining whether the AI-enabled internet remains open or becomes closed - The shift represents a move from content consumption by humans to AI agents performing tasks on behalf of users - Content creators and site owners need to prepare for AI-mediated interactions while maintaining control over their digital representation ## 📖 Main Content James LePage, Core AI Co-lead for WordPress and Head of AI at Automattic, presents WordPress's strategy for the AI era. The presentation argues that we're experiencing a fundamental shift in how the web is consumed - from humans directly reading content to AI agents acting as intermediaries. The core thesis is that WordPress, powering 43% of the web, has a unique opportunity and responsibility to ensure the AI-enabled internet remains open rather than becoming a closed, proprietary ecosystem. WordPress 7.0 introduces four building blocks designed to make WordPress "agentic-ready": 1. **Abilities API** - Allows websites to publish machine-readable lists of their capabilities 2. **WP AI Client** - Universal interface for interacting with AI providers 3. **MCP Adapter** - Implements Model Context Protocol for third-party AI client integration 4. **AI Plugin** - Reference implementation tying the other blocks together The presentation envisions a future where websites have "site delegates" - AI agents that represent the site owner's interests when interacting with visiting AI agents, creating an "agent-to-agent" web while maintaining decentralized, open protocols. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/jDsozp7y8bk?t=0) — [H1] Opening and Context Setting - Key idea: WordPress's role in shaping the AI-enabled internet is critical given its 43% market share - Evidence: "43% of the web running on WordPress" #### ⏱️ [05:23](https://youtu.be/jDsozp7y8bk?t=323) — [H2] The Web as a Living Graph - Key idea: The internet is a connected graph of entities representing people, businesses, and ideas - Evidence: "The web isn't just a collection of pages, but instead it's a living graph of entities" #### ⏱️ [12:45](https://youtu.be/jDsozp7y8bk?t=765) — [H3] Shifting Consumption Models - Key idea: AI is changing how content is discovered and consumed on the web - Evidence: "AI consuming content. AI didn't exist in the way that it does even 3, four years ago" #### ⏱️ [18:30](https://youtu.be/jDsozp7y8bk?t=1110) — [H2] Content Remains King in AI Era - Key idea: AI systems need high-quality content to function effectively - Evidence: "Content is king again. And the reason is AI needs content to run well" #### ⏱️ [25:15](https://youtu.be/jDsozp7y8bk?t=1515) — [H1] Open vs Closed AI Future - Key idea: The architecture of AI systems will determine whether the web remains decentralized - Evidence: "Is this future open or is the future closed?" #### ⏱️ [32:00](https://youtu.be/jDsozp7y8bk?t=1920) — [H1] WordPress's Strategic Position - Key idea: WordPress has unique assets of distribution, content, and open web foundation - Evidence: "WordPress is by definition a content management system" #### ⏱️ [38:45](https://youtu.be/jDsozp7y8bk?t=2325) — [H1] AI Building Blocks Introduction - Key idea: WordPress 7.0 delivers four foundational pieces for AI integration - Evidence: "AI building blocks for WordPress... gets delivered in its entirety with WordPress 7.0" #### ⏱️ [42:30](https://youtu.be/jDsozp7y8bk?t=2550) — [H2] Abilities API Details - Key idea: Websites can publish machine-readable capabilities for AI interaction - Evidence: "Allows you to tell any AI what your website is capable of" #### ⏱️ [46:15](https://youtu.be/jDsozp7y8bk?t=2775) — [H2] WP AI Client Functionality - Key idea: Universal interface prevents vendor lock-in for AI providers - Evidence: "Allows WordPress to interact with any AI provider... It doesn't lock you into a provider" #### ⏱️ [50:00](https://youtu.be/jDsozp7y8bk?t=3000) — [H2] MCP Adapter Implementation - Key idea: Model Context Protocol enables third-party AI clients to interact with WordPress - Evidence: "Allows third party clients to talk to software" #### ⏱️ [55:30](https://youtu.be/jDsozp7y8bk?t=3330) — [H3] Ecosystem Adoption - Key idea: WordPress community is rapidly adopting the AI building blocks - Evidence: Multiple plugins and providers implementing the new APIs #### ⏱️ [62:00](https://youtu.be/jDsozp7y8bk?t=3720) — [H1] The Agentic Web Vision - Key idea: Future web will include AI agents as active participants alongside humans - Evidence: "The agentic web is simply the internet that exists... with a new player on the board" #### ⏱️ [68:45](https://youtu.be/jDsozp7y8bk?t=4125) — [H2] Site Delegates Concept - Key idea: Websites will have AI representatives to interact with visiting agents - Evidence: "A delegate being something that represents you, something that understands your content" #### ⏱️ [75:15](https://youtu.be/jDsozp7y8bk?t=4515) — [H1] Democratizing Representation - Key idea: WordPress can democratize digital representation in the AI era like it did for publishing - Evidence: "WordPress made it possible for anybody to publish on the web" ## 📚 References - [WordPress AI Plugin](https://wordpress.org/plugins/ai/) - [Make WordPress AI Team](https://make.wordpress.org/ai) - [Model Context Protocol](https://modelcontextprotocol.io/) - [WP Playground](https://playground.wordpress.net/) ## 🗣️ Paraphrase - WordPress is positioning itself as the foundation for an open AI-enabled web by building infrastructure that allows AI agents to interact with websites while maintaining user control - The four AI building blocks in WordPress 7.0 create a toolkit for developers to build AI features without vendor lock-in or closed proprietary systems - The future web will include both human users and AI agents, requiring websites to adapt their representation and interaction models while preserving the decentralized nature of the internet ## 📰 Newsletter WordPress 7.0 is bringing AI capabilities directly to the core platform! James LePage from Automattic shared how WordPress is preparing for an "agentic web" where AI assistants interact with websites on behalf of users. The big news: WordPress 7.0 includes four AI building blocks - the Abilities API, WP AI Client, MCP Adapter, and AI Plugin. These tools let your WordPress sites work with any AI provider while keeping you in control of your data and representation. Why this matters: With WordPress powering 43% of the web, these changes help ensure the AI-enabled internet stays open and decentralized rather than locked into proprietary systems. Ready to explore? Try the WordPress 7.0 beta in WP Playground and install the AI plugin to see how your site can work with AI agents like Claude or ChatGPT. Resources: [Learn WordPress](https://learn.wordpress.org/), [Developer Docs](https://developer.wordpress.org/), [WP-CLI](https://wp-cli.org/) ## ✅ Action Items #### WordPress - Core AI: Finalize WordPress 7.0 AI building blocks, document MCP implementation - Core: Integrate collaborative editing with AI agents, test permissions model - Meta: Update WordPress.org to showcase AI capabilities, create developer guides - Documentation: Create comprehensive guides for AI building blocks, MCP adapter usage - Community: Organize AI-focused contributor days, create learning resources #### DevRel - Developer advocacy: Create demos of AI building blocks, present at conferences - Education: Develop workshops on building AI-enabled plugins, MCP integration - Developer experience: Improve AI plugin documentation, create code examples - Developer marketing: Promote WordPress as open AI platform, contrast with closed systems #### Community - Maintainers: Explore AI integration opportunities in existing plugins - Contributors: Test AI building blocks, provide feedback on APIs - Ecosystem: Build provider plugins for AI services, create reference implementations ## 🧪 Hypotheses - WordPress's AI building blocks will prevent AI vendor lock-in and maintain open web principles - Site delegates will improve AI agent interactions while preserving content creator control - The agentic web will increase demand for high-quality, well-structured content - MCP adoption will standardize AI-website interactions across the industry ### 📊 Methodology - A/B test AI-enabled WordPress sites vs traditional sites for user engagement - Monitor adoption rates of AI building blocks across WordPress ecosystem - Analyze performance impact of MCP adapter implementation ### ❓ Questions - How will search engines adapt to agent-to-agent web interactions? - What governance models will emerge for AI agent behavior on websites? - How will content monetization evolve when AI consumes content without direct visits? ### 🔮 Future Research - Economic impact of AI-mediated web consumption on publishers - Technical requirements for large-scale agent-to-agent interactions - Privacy implications of AI agents accessing personal websites ### 💡 Implications - Practice: Web developers need to learn AI integration patterns and MCP protocols - Policy: Content licensing and AI access rights require new frameworks - Security: AI agent authentication and authorization become critical concerns - Education: WordPress training must include AI concepts and agentic web principles - Funding: Investment in open AI infrastructure vs proprietary platforms becomes strategic ### 🧠 Learning Styles Assessment Reading level estimate: 12th grade (technical presentation with complex concepts) #### 👀 Visual learners - Present: Slides with diagrams of web graph structure, building block architecture - Improve: Interactive demos of AI agents working with WordPress sites #### 🔊 Aural learners - Present: Live presentation with clear explanations of technical concepts - Improve: Audio examples of AI agent interactions, voice demonstrations #### 🗣️ Verbal learners - Present: Clear definitions of technical terms, analogies for complex concepts - Improve: Glossary of AI and WordPress terminology, story-driven examples #### 🧩 Kinesthetic learners - Present: Call-to-action to install and test WordPress 7.0 beta - Improve: Step-by-step tutorials for implementing AI building blocks #### 🧠 Logical learners - Present: Structured presentation of building blocks, clear cause-and-effect reasoning - Improve: Decision trees for AI implementation choices, technical specifications #### 🧑‍🤝‍🧑 Social learners - Present: WordCamp presentation format, community discussion encouragement - Improve: Workshop sessions, collaborative AI experiments #### 🧘‍♀️ Solitary learners - Present: Technical documentation references, individual exploration encouraged - Improve: Self-paced tutorials, personal experimentation guides #### 🌿 Naturalistic learners - Present: Web ecosystem analogies, organic growth metaphors - Improve: Environmental impact discussion of AI systems ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear slide structure, verbal description of concepts - Improve: Alt text for visual diagrams, structured navigation aids #### Keyboard navigation - Present: Standard video controls available - Improve: Keyboard shortcuts for demo interactions #### Contrast and readability - Present: High contrast slides, clear typography - Improve: Ensure all text meets WCAG AA standards #### Text resizing and reflow - Present: Scalable slide content - Improve: Responsive design for various screen sizes #### Media and captions - Present: Live presentation with clear audio - Improve: Closed captions, transcript availability #### Forms and interactions - Present: Simple QR code scanning instructions - Improve: Alternative access methods for QR codes ### 🌍 DEIB Assessment #### Cultural breadth - Present: Global WordPress community focus, international WordCamp - Improve: Examples from diverse regions, multilingual AI considerations #### Inclusive language - Present: Technical language appropriate for developer audience - Improve: Avoid assumptions about technical background, define all terms #### Neurodiversity support - Present: Structured presentation with clear sections - Improve: Visual summaries, consistent formatting #### Representation - Present: Focus on technical content over personal representation - Improve: Diverse examples of WordPress users and use cases #### Psychological safety - Present: Encouraging tone about experimentation and learning - Improve: Explicit support for questions and exploration ### ✅ Logical Validity #### Strong reasoning - Deductive strength: WordPress's market share logically positions it to influence AI web standards - Inductive strength: Current AI adoption patterns support predictions about future web evolution - Causal reasoning: Clear connection between AI building blocks and open web preservation - Analogical strength: Publishing democratization analogy effectively explains digital representation goals #### Fallacies - Appeal to authority: Relies heavily on WordPress's market position without addressing technical merit - False dichotomy: Presents only open vs closed future options, ignoring hybrid models #### Tone of voice - Overall tone: Optimistic and technical, with clear expertise and enthusiasm for open web principles ### 🛡️ DARVO Analysis - Deny: No significant denial patterns observed - Attack: No attacks on opposing viewpoints - Reverse victim/offender: Not applicable to this technical presentation - Emotional cues and rhetoric: Positive framing of WordPress's role, some promotional language for Automattic products ## 🎯 Conclusion WordPress 7.0's AI building blocks represent a strategic move to ensure the AI-enabled web remains open and decentralized. The four core components - Abilities API, WP AI Client, MCP Adapter, and AI Plugin - provide the infrastructure needed for websites to interact with AI agents while maintaining user control and avoiding vendor lock-in. As AI agents become more prevalent in web interactions, WordPress's 43% market share positions it to significantly influence whether the future internet preserves its decentralized, open nature or becomes dominated by closed proprietary systems. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad Purpose: Stage CEE units that become Literature notes. Split level: H3 ### [H1] AI building blocks enable open agentic web > [!quote] [WordPress and AI: Building an Open Agentic Web](https --- ## Full Transcript [00:16] Hello. [00:19] Heat. [00:50] Hey. Hey. Hey. [01:18] Hey, hey, hey. [03:34] Funny. [03:58] All [04:20] right. [04:24] Heat. Heat. [05:01] Got WordPress? Now what you need? [05:04] Jetpack. Jetack makes your site faster. [05:07] Create more content with AI. [05:11] Jetack expands your audience. Publish on [05:14] your site. Share it everywhere. Watch [05:16] your stats grow. Jetack protects your [05:18] site. Your business stays online 24/7. [05:22] Everything is backed up in real time. [05:24] Eight individual plugins. One powerful [05:27] family of products. The best WordPress [05:29] sites have Jetack. [05:32] Managed WordPress hosting made [05:34] effortless, fast, and reliable. [05:37] Pressible geo redundant failover ensures [05:40] zero downtime even during traffic [05:42] spikes. We keep your site secure with [05:45] automatic updates and SSL certificates. [05:48] Woo Commerce stores run smoothly no [05:50] matter the load. Focus on growing your [05:53] business. We'll handle the hosting. [06:01] WordPress.com has everything you need to [06:04] build, launch, and grow a WordPress site [06:07] all in one place. like integrated AI [06:10] tools that help you move faster across [06:12] your workflow without losing control. [06:14] And WordPress Studio, which lets you [06:16] develop locally, then push straight to [06:18] production when you're ready. Launch [06:20] sites for clients knowing they have [06:22] managed hosting, predictable pricing, [06:24] and human support behind them. Build [06:27] sites your clients won't outgrow. [06:29] WordPress.com [08:02] Namaste. The program will begin in 10 [08:04] minutes. Please silence your mobile [08:07] phones. [08:38] Hey, [08:52] hey, hey. [09:31] every [10:16] down. [10:23] All [10:43] right. [11:58] Heat. Hey, Heat. [12:26] Feel [12:49] feel. [13:02] The program will begin in 5 minutes. [13:04] Please take your seat and silence your [13:06] mobile phones. [13:20] Hey, hey, hey. [13:45] Heat. Heat. [14:08] Heat. Heat. N. [14:18] Heat. Heat. [14:40] Heat. Hey, Heat. [17:06] The program is about to begin. Please [17:08] take your seat and silence your mobile [17:10] phones. [18:45] Namaste and good morning Mumbai. [18:51] Welcome to World Cam Asia 2026 and [18:54] welcome to Mumbai. We are so glad that [18:57] you are here. Take a moment, look around [19:00] and soak this in. This event didn't come [19:03] together by accident. We had organizers, [19:06] supporters from cities, countries across [19:09] Asia and beyond. They put their time, [19:12] their energy and their heart to make to [19:16] get this celebration of WordPress in [19:18] Mumbai. So if you are one of them, then [19:21] please stand up. We want to thank you. [19:24] Can you get some feet up? [19:36] Before we go further, we want to take a [19:39] pause. We want to remember someone who [19:41] should have been here with us today. In [19:44] November, we lost Zil Tucker. She is she [19:48] was a beloved organizer from Ahmedabad. [19:51] She always showed up with enthusiasm and [19:54] a smile that lit up the room. Zeal [19:57] encouraged people to get involved. Her [20:00] spirit is very much part of what we are [20:03] building here today. We would like to [20:06] remember Zeal with a minute of silence. [20:09] Please join us as we remember her and [20:11] everything she gave to this community. [21:16] Thank you everyone. Let us carry her [21:18] energy with us throughout the next two [21:22] days. Let let let let us walk you [21:25] through what we have lined up. [21:27] We have two the next two days are packed [21:31] with amazing sessions, workshops, great [21:34] conversation with old friends, new [21:37] friends. Check the schedule board and [21:39] plan your two days. Make sure you do not [21:43] miss the sessions that you want to [21:44] attend. [21:48] A quick note on the venue. You'll find [21:51] the information desks right outside the [21:53] US hall. Our volunteers Our [21:58] volunteers are wearing official WordCamp [22:01] t-shirts, so they are easy to spot. If [22:04] you need any help at all, walk up to [22:06] them and just ask. A quick note, so uh [22:10] during your social posting, be [22:12] considerate of participants wearing this [22:15] red lanyard. They should not be part of [22:19] your social post uh in a photograph. [22:21] Just be mindful of their privacy. [22:24] And now [22:27] none of this that you're seeing would [22:29] have been possible without our sponsors. [22:31] So they believed in this event and [22:33] backed it. Please join us in thanking [22:36] them. We would like to take this [22:38] opportunity to thank our word camp Asia [22:42] 2026 sponsors. [22:44] So, our super admin sponsor, Elementor, [22:48] Hostinger, Jetpack, PayPal, Pressible, [22:52] Salesforce, WordPress.com. [22:56] Our admin sponsors, Google, and Woo [22:59] Commerce. Our editor sponsors, Astra, [23:03] Automatic for Agencies, Blue Host, [23:07] Yoast. [23:09] Our author sponsors [23:11] EasyWP [23:13] by Spaceship, Kinsta, Mini Orange, [23:17] RankMath, RTAM, WP Rocket, [23:22] as well as our other contributors, [23:25] subscribers, and viewer level companies [23:27] participating. A huge thank you to all [23:30] our sponsors. Without their generous [23:33] support, Word Camp Asia simply would not [23:36] have been possible. Make sure to stop by [23:38] the sponsor hall in Jasmine, too, to [23:41] just say a hi and a hello. Give them a [23:43] big round of applause. [23:51] We have planned the program around three [23:53] things that guide us. Uh we're going to [23:57] we're going to get help with expanding [23:59] WordPress, our knowledge, explore the [24:01] scale of WordPress, and also when we go [24:03] back from here, hopefully grow our local [24:05] communities. [24:08] So, here is what I want you to do today [24:11] and tomorrow. [24:13] Uh, in the true spirit of open source, [24:15] be open to having some great [24:18] conversations with someone you don't [24:20] know. Be open to learn something new and [24:24] to teach something to someone that you [24:26] know, to teach something that to someone [24:29] something that you know. We all grow [24:32] when we lift each other up. [24:35] That is what WordPress is all about. Uh [24:38] so now let's roll with word camp Asia [24:40] 2026. I won't and have a good day. [25:21] I think the [25:26] QR code is like [25:40] wind. [26:41] Yeah. [26:45] Yeah. [26:49] Yeah. [26:57] Yeah. [27:01] Yeah. [27:24] Hello Mumbai. [27:26] Hello. Hello. [27:28] Let's give a round of applause to the [27:30] leadership team once more. [27:34] Now, uh my name is Patrick Lumumba and [27:37] I'm from Nairobi, Kenya. Who has ever [27:40] been to Nairobi, Kenya? [27:43] That's nice. Um, I'm so glad to be here [27:47] and uh the first thing that I want you [27:49] to know is that uh we are ahead of [27:51] schedule. Our first speaker of the day [27:53] will be James Gems and it will be uh [27:56] we'll be starting at 9:30. In the [27:59] meantime, I have some things that you [28:02] should know. [28:04] One uh we have QR code. So these are for [28:08] the closed captioning. You once you scan [28:11] them, you'll have them on your phone so [28:13] that you can be able to follow along [28:15] what is happening. [28:19] Now um uh I I I believe you had already [28:22] uh gotten the information that you [28:24] should be silencing your phones. So [28:25] please silence your phones so that we [28:28] don't interrupt um the sessions. [28:32] Again, [28:34] we have the schedules right outside the [28:37] the lotus room. They are on the uh the [28:41] readerboards. So, if you want to know [28:43] exactly where to be at what time, please [28:46] check the the readerboards right outside [28:48] the lotus room one. [28:53] Another announcement that I would like [28:54] to put across is that uh there will be [28:58] family photo. There will be a family [29:00] photo session just right outside the [29:02] lotus room starting from 11:20. [29:05] Then [29:07] lunch will be served from 11:45 to 1:00 [29:11] p.m. So be ready to go for the lunch [29:16] break and lunch will be served at [29:18] Jasmine room uh Jasmine Hall 1. [29:23] Then uh the sessions will again resume [29:26] in all the halls. Uh I mean uh there [29:28] will be some sessions in the workshop [29:31] rooms. The workshop rooms you can find [29:33] them right on level two. [29:38] Okay. [29:42] Who has a question about Nairobi before [29:44] we get started? [29:51] Nice. So one thing that I I I I tried I [29:54] tried convincing the leadership team to [29:57] have World Camp Asia [29:59] uh to to have Nairobi host World Camp [30:02] Asia and they said no. [30:06] I didn't know that you know for the [30:07] World Camp Asia. Their reason actually [30:09] was that Nairobi is not in Asia. So uh [30:13] we can't host World Camp Asia. Where is [30:16] the camera? Nairobi. We tried. [30:19] What campaign is not coming to Nairobi? [30:23] Okay. [30:26] For now I would like to introduce the [30:28] first speaker. Uh the session is going [30:31] to be purely virtual. So uh we will have [30:35] him on the screens [30:39] and our session is going to be about [30:42] WordPress and AI. I know all of us are [30:45] aware of the phrase that AI is here with [30:49] us but for the WordPress community it is [30:52] already here and uh in this session we [30:55] are actually going to get it from the [30:58] people who are actually building it and [31:01] that is James Leage. James [31:05] started off as a freelancer where we all [31:08] I I believe we all did most of us [31:11] started off as freelancers and he went [31:13] on and founded co-founded uh WPAI which [31:18] got acquired by automatic the guys [31:21] behind WordPress.com Jetpack Tumblr and [31:25] all [31:26] and in this session I'm so much excited [31:29] because I would like to know how AI and [31:32] WordPress is actually going going to be [31:34] and if you've been following along [31:37] what AI is coming to WordPress 7.0 [31:42] and James is actually leading the [31:45] project. So uh everyone capesia allow me [31:51] welcome James Lipage. [32:26] Ladies and gentlemen, [32:29] ladies and gentlemen, James will be on [32:31] in u exactly 5 minutes from now. So, [32:34] take a moment, network with uh whoever [32:37] is seated next to you. [32:40] Immediately after this session, I would [32:41] like to know exactly what you've learned [32:43] about the person seated next to you. And [32:45] let it not be someone that you you [32:47] already know. It will be boring. Let it [32:51] be someone you totally haven't seen or [32:54] haven't met before. All right. [33:14] Heat. Heat. [33:46] Hey, [33:54] hey, hey. [34:58] Hey, hey, hey. [35:26] Heat. Heat. N. [36:50] Heat. Heat. [37:17] Heat. Heat. N. [37:52] Hello Mumbai. [37:55] A quick announcement. [37:57] Right opposite uh the building we have [38:00] the US embassy and we are not allowed to [38:04] videograph or take a photograph of that [38:07] place. So be very careful. Don't take [38:10] photographs of the place uh during the [38:13] the whole time you'll be here at this [38:15] event. Another announcement is that uh [38:18] we'll be serving lunch from 11:30 to [38:22] 2:30 p.m. So take note of that. [38:26] Um now is the time. Let us now welcome [38:30] James Leage for the session WordPress [38:34] and AI. [38:45] Hello, Word Camp. How are y'all? Thank [38:48] you so much for having me. I really [38:50] appreciate it and I'm excited to jump [38:52] into this presentation on AI and [38:54] WordPress. I'm also super bummed I can't [38:57] be there in person. I was planning on [38:59] it. I have some family stuff. So, I'm [39:00] giving you this presentation live and [39:02] direct from Charlotte, North Carolina in [39:04] the United States. I'm drinking coffee. [39:06] It's the middle of the night. It's early [39:08] for y'all. Everybody's drinking coffee. [39:10] My goal today is to talk about AI and [39:13] WordPress and speak towards the things [39:16] that have happened in the project and [39:18] what we're building and what we're [39:19] doing, what's coming in the near future, [39:22] but also zoom out a little bit and think [39:24] about the position and place that [39:26] WordPress plays in this internet in this [39:29] AI era. So, it's an interesting [39:32] presentation. It's one where it's kind [39:34] of taken a life of its own as I've put [39:36] it together. So, I'm really excited to [39:38] present to y'all. Hopefully, there will [39:40] be time for questions at the end. I [39:42] think one of my favorite things about [39:43] Word Camps and something that I'm [39:45] definitely already feeling the FOMO [39:47] looking at the video feeds is that in [39:50] person question answer time, the ideas [39:53] shared, the code looked at. Um so while [39:57] I won't have that opportunity in person [39:59] myself uh there are many folks from the [40:02] core AI focuses and functions here and I [40:05] really invite you to use the inperson [40:08] time to the best of your advantage. So [40:11] the obligatory slide who I am my name is [40:13] James Leage. Um I actually had a way [40:17] better introduction from uh our opening [40:20] presenters than I even think I can give [40:22] myself. So thank you for that. That's [40:24] awesome. But my name is James Leage. I'm [40:26] the core AI co-lead for the WordPress [40:29] project. I'm also the head of AI at [40:31] Automatic. Um, to quickly restate my [40:34] background, I got my start in WordPress [40:36] building websites as a freelancer, [40:38] eventually grew into running an agency, [40:41] sold the agency, ran a startup in [40:43] fintech for a little bit, and then [40:46] founded a company called WPAI. [40:48] And WPAI, as the name suggests, did AI [40:52] for WordPress. Um, some of y'all may [40:55] remember the CodeWP and the AgentWP [40:58] products. I really love those near and [41:00] dear to my heart. Uh, we were acquired [41:03] into Automatic and this is where I find [41:05] myself. You can find me online at j.cv. [41:09] That's my personal blog. I have a lot of [41:11] writing on this blog that cover a lot of [41:13] the topics that we'll speak about today. [41:16] You can also find me on X or Twitter and [41:19] LinkedIn at JamesW leage. So that's who [41:22] I am. And before we really get into the [41:24] WordPress stuff, before we get into the [41:26] AI stuff, I I want to take a step back [41:28] and take a look at what WordPress [41:31] actually works within, what it occupies, [41:33] which is the web. And to me, the web [41:37] isn't just a collection of pages, but [41:40] instead it's a living graph of entities. [41:42] And it's really beautiful. This [41:44] collection of entities, these individual [41:47] websites represent people and businesses [41:49] and places and concepts and ideas. And [41:52] each of these is connected through [41:55] links. They have citations. They have [41:57] content that's similar to other [41:58] websites. They have trust signals. So, [42:01] we have this beautiful internet of [42:03] interconnected pieces of the web. These [42:07] individual sites typically contain [42:09] content that's been shaped and crafted [42:11] by people and humans behind it. and it [42:14] forms a genuinely beautiful collection [42:17] of human knowledge, of human ideas, [42:19] ingenuity. This is a visual [42:22] representation of that actual graph of [42:24] the web. And if you look it up, I I [42:26] actually forget what the study is, but [42:28] you you can get to these images and you [42:30] can see the evolution of the web over [42:32] time and the growth and the scaling of [42:35] these really really incredibly connected [42:39] nodes. Each little dot in this [42:41] representation being a website. Websites [42:43] connected through the links through the [42:45] semantic similarity. Each color in this [42:49] graph representing a specific category [42:51] of site. This is what the internet is. [42:54] And again to me it's a beautiful [42:56] representation of people and humanity [42:59] all pushing their hearts and their souls [43:01] and their ideas into a digital space. [43:05] And this structure of the web, this [43:07] interconnected collection of entities [43:09] has been very stable for 30 plus years. [43:13] And what's changing right now because of [43:15] this AI stuff isn't that structure, but [43:18] instead it's everything else. It's [43:20] everything around that structure. And [43:22] one of the big changes that I'm seeing [43:23] today is the shifting consumption model. [43:27] Now we have AI consuming content. AI [43:30] didn't exist in the way that it does [43:32] even 3, four years ago. users are [43:34] reading less. Publishers are changing [43:37] the way they're capturing value. And [43:39] when I say consumption model, it maybe [43:41] sounds like a a big scary concept, but [43:44] it's really just the discovery and the [43:46] reading and the understanding and the [43:48] traversing of the web and the content [43:51] within. And that consumption model is [43:53] changing because of AI. So yesterday, [43:56] and I say this in absolute terms, but [43:58] it's not absolute. This is still [44:00] happening. I still do this all the time [44:01] today. Yesterday, a human would visit a [44:04] site. They'd go and read the content. [44:05] They'd click the links. They'd go [44:07] through the rabbit holes. They'd get [44:09] distracted by an Amazon product and then [44:11] buy it as I did today. Uh, and we go and [44:14] synthesize the answer or we go and take [44:16] the action or we go and achieve the goal [44:19] uh on our own accord. And this has been [44:22] really the way the web has worked for a [44:24] long time. We search on Google, we end [44:26] up on the site. We see the URL on the [44:28] billboard, we end up on the site. We [44:29] click through, we come to our own [44:31] conclusions. And today now we see AI [44:34] beginning to sit between the human and [44:36] the web. And we see this with many of [44:38] these AI tools. We see this also with [44:40] the traditional ways we've interacted [44:42] with the internet. Google introducing AI [44:45] overviews. Something that synthesizes [44:48] answers from dozens or hundreds of [44:51] sources to give you a quick tidbit of [44:53] information that may answer your [44:55] question, may help you achieve your goal [44:57] a bit quicker. a chat GBT pulling web [45:01] searches down to site sources to give [45:03] you answers or research reports or [45:06] recommendations. [45:07] But right now, the user is still [45:09] deciding. They're still acting. They're [45:11] the ones being given this synthesis of [45:14] that web. Tomorrow, things change even [45:18] more. The user begins to delegate goals [45:20] to an agent. And that agent instead of [45:23] the person begins traversing this open [45:25] web, walking through those connected [45:28] lines on the entities, comparing and [45:30] acting, creating a recommendation, [45:33] fulfilling an action. And the user [45:35] chooses to be a part of the outcome, but [45:37] maybe they don't need to be. And I think [45:39] it's also worth defining what an actual [45:41] agent is. An agent to me is something [45:44] that is an AI based system that uses [45:47] LLMs that has agency. Agency meaning [45:50] they can go and do things themselves. [45:53] They get the goal, they have a [45:54] collection of tools, they walk through [45:56] the internet to go and achieve that [45:58] outcome. And this all represents that [46:01] shifting changing consumption model. Now [46:04] instead of taking half a day to do [46:06] research on one specific thing, I can go [46:09] spin up multiple agents. I can use [46:11] OpenClaw. I can use cloud code. I can [46:14] use chatgbt to go and do things for me, [46:17] which means I'm much more enabled, much [46:19] more efficient, and I can do a lot more [46:21] at once. It's a very interesting shift. [46:25] And that shift points to a term that has [46:28] probably been mentioned at word camps [46:30] time and time again. Content is king. [46:33] And content in this AI era that we find [46:35] ourselves hurtling towards still is [46:38] king. Content is king again. And the [46:41] reason is AI needs content to run well. [46:44] The models are already trained. They [46:47] already exist. But what they need is [46:49] information about the world, deep [46:51] expertise on specific topics, creation [46:54] and novel outcomes that these models [46:57] can't do. So right now, humans are still [47:00] reporting the news. We don't have flying [47:01] robots from OpenAI going and [47:03] interviewing us as to how great Word [47:06] Camp Asia really is this year. The [47:08] humans are saying it's great this year. [47:11] The people are writing the posts. We're [47:13] creating the things that are making AI [47:16] useful. [47:17] Which means that to me, AI, generative [47:21] AI, LLM, agents, they're not going to [47:24] replace the open web if we have anything [47:26] to say about it. But instead, it makes [47:28] this open web, this beautiful graph that [47:31] I showed you on slide four, way more [47:33] valuable. the people on the platforms [47:36] that are creating and creating and [47:38] publishing. This is going to matter a [47:40] lot more in this AI mediated internet, [47:44] not less. And to me, the act of writing [47:46] and publishing on a personal level is [47:49] more important now than it's been in [47:50] years. As soon as I saw how the internet [47:53] was evolving, these answers, these [47:55] retrieval systems, these chat bots that [47:58] would draw on sources, I began blogging [48:01] a lot more about me, myself, my [48:03] businesses because I knew that I wanted [48:05] to control my representation, control [48:08] the way I was positioned and placed on [48:10] the internet. This has always been [48:12] something, but it's much more important [48:15] in this AI enabled era. [48:18] But the way the publisher captures value [48:21] is shifting. That content model is [48:23] shifting and it's impacting the economic [48:26] model that publishers big and small, me [48:29] on my blog, huge media conglomerates, [48:32] enterprises, e-commerce stores, commerce [48:35] stores. It's changing the way we capture [48:38] value. Now AI is synthesizing the [48:41] content without always sending visitors [48:43] to the source. And we're seeing [48:44] responses already on the open internet [48:47] like RSL, really simple licensing, a [48:50] standard to monetize and protect [48:52] content. We're seeing Cloudflare block [48:54] AI crawlers by default. That model is [48:57] changing. [48:58] And it poses an interesting question to [49:00] me. Who shapes how the content is [49:03] accessed and how it's synthesized and [49:05] how it's represented in this era that we [49:08] find ourselves rapidly pointing towards? [49:11] And if we want to go higher level, a big [49:14] question becomes what does this future [49:16] look like? Is this future open or is the [49:18] future closed? And I'm sure this is [49:20] another question that we've also asked [49:22] at word camps time and time again. To [49:25] me, open versus closed when it comes to [49:28] this AI enabled internet, this AI era is [49:31] an interesting question. The closed [49:34] approach means that we have proprietary [49:36] layers. They all work together. Content [49:38] becomes material for interfaces. It's a [49:41] walled garden. It's a garden. It's very [49:43] pretty, but you're within it. You don't [49:45] have an understanding as to how it [49:48] works. And the agents that operate [49:50] within these systems, these closed [49:52] systems, don't publish their protocols. [49:54] They don't publish the way they work. We [49:56] don't really understand how to best [49:58] interact with them. And that closed [50:01] walled garden is something I would like [50:02] to avoid in this AI internet. But I [50:06] think it's also something that we've [50:07] always wanted to at least have an answer [50:09] to, which is a big reason why I love [50:12] WordPress in general. And that open [50:15] future, that open approach brings the [50:18] source code to light. It has open [50:20] protocols. It has open data. Your site [50:23] can expose what it is and what it does [50:25] through these standards like MCP. And [50:27] I'll talk a lot more about that [50:28] momentarily. You can choose your AI [50:31] models. You can have your infrastructure [50:32] and the web that beautiful graph [50:35] representation of the web remains this [50:37] decentralized network where you're [50:39] maintaining the control on how you're [50:41] represented. You're owning your content. [50:44] You're being able to move it around. You [50:46] have the benefit that WordPress offers. [50:49] And this isn't something that's [50:50] theoretical. This is something that's [50:52] happening right now. And I absolutely [50:54] see it as an individual working on my [50:57] own personal blog walking through the [50:59] internet. But I definitely see this in [51:01] automatic and in the statistics on our [51:04] users and in the way the internet itself [51:06] is shaping up in the steering groups and [51:09] the committees and all of the different [51:11] things going on online today. This stuff [51:14] is happening now and the outcome of this [51:17] closed verse open how this internet [51:19] actually looks in five years. It really [51:22] to me depends on what gets built and [51:24] who's actually building it. And that's [51:26] when we get to our favorite number. And [51:28] I know this number shifts and changes a [51:30] little bit. This is actually my favorite [51:32] number. So 43% of the web running on [51:34] WordPress. [51:36] That means that WordPress plays a very [51:38] important role in what this future [51:41] internet looks like. The largest [51:44] open-source publishing platform on the [51:47] planet. And I'll actually put a a cross [51:49] through a strike through and just say [51:50] the largest publishing platform on the [51:52] planet. And this is the largest [51:54] publishing platform that has two [51:56] interesting things especially in this AI [51:59] era. Distribution and content. And this [52:02] is really important for WordPress, but [52:04] this is important really for any [52:05] product, any piece of software now that [52:08] these generative AI LLM based things are [52:12] walking around. Distribution is really [52:15] hard to do. Having placement across the [52:18] entirety of the internet is very [52:20] difficult. new things pop up [52:22] consistently, but they don't have the [52:24] positioning that WordPress has, and they [52:27] also don't have the content. WordPress [52:29] is by definition a content management [52:31] system. It's something that manages the [52:33] stuff that a language model must [52:36] interact with to be powerful. So, these [52:39] are two assets that we have as the [52:40] WordPress project. And the third is that [52:43] we have always been the open web's [52:45] foundation. And now this role is more [52:48] and more important. [52:51] And so this was something that I think [52:54] the WordPress project itself began to [52:56] understand as we began to understand [52:59] what LLMs were and how they would work [53:02] and how they would interface with the [53:04] software and what the limitations were [53:06] and what we would expect things to [53:07] happen, how we would expect things to [53:09] shape up in the future. And the project [53:12] was deeply tracking LLMs basically since [53:15] they came out. And we watched and we [53:18] waited for the maturity of an LLM, the [53:20] maturity of the AI industry itself and [53:23] then began building. We started building [53:26] for this interesting change in the [53:29] consumption models, this interesting [53:30] change in the way we interact with [53:32] software. And the first thing we started [53:35] doing on the AI side of things, [53:37] something that gets delivered in its [53:38] entirety with WordPress 7.0, we call AI [53:42] building blocks for WordPress. and [53:44] you'll see a couple QR codes in the [53:46] lower corner of the screen. These are [53:48] things you can scan to get a lot more [53:50] information about it. I wanted to do [53:52] this presentation in a way where we [53:54] spoke about these four foundational [53:56] pieces, but also spoke about the impact [53:59] that they have now that they really do [54:00] exist in WordPress. If you've seen State [54:03] of the Word, if you've seen some of the [54:04] previous Word Camp US presentations, we [54:07] go into a lot more detail about these [54:09] building blocks. But if you scan that QR [54:11] code, it will give you a really good [54:13] overview of these blocks. You can also [54:16] keep up to date with all of the AI stuff [54:18] at make.wordpress.org/ai. [54:21] And so these are the building blocks [54:23] that we need to create to bring AI to [54:26] WordPress. And now today with the [54:29] WordPress 7.0 0 beta with the upcoming [54:32] release and in WordPress 6.9 WordPress [54:35] can do three things that it couldn't do [54:37] before. The first thing, the first [54:40] building block for AI, we call the [54:42] abilities API. And this is in WordPress [54:45] right now as of 6.9. The abilities API [54:48] allows you to tell any AI what your [54:51] website is capable of. Your site can [54:54] publish a machine readable list of [54:56] everything it can do. And that machine [54:58] readable list says if you give me this [55:00] in, I'll go do this out. And by the way, [55:02] here are my permissions and here's [55:05] exactly how I work. So an ability could [55:08] look like updating content, managing [55:10] users, interlinking pages. And when [55:13] presented as a total list of what the [55:16] website can do to AI, that AI can use it [55:19] to interact with software. You can think [55:22] of this as a tool set and your website [55:25] itself being a toolbox. [55:27] Abilities are really fun to me because [55:29] you can also add additional abilities [55:32] with the installation of WordPress [55:34] plugins. Maybe I have a great learning [55:36] management plugin and it has a lot of [55:38] capabilities that I want AI to be able [55:40] to interact with. If that learning [55:42] management plug-in bundles abilities and [55:44] I have an agent that uses these [55:46] abilities or these individual tools, you [55:49] could almost imagine the installation of [55:51] this plugin on my WordPress website [55:53] dumping seven new tools into that [55:56] toolbox that now AI can use. And I have [55:59] to mention also abilities is a very cool [56:01] API. Even if AI didn't exist, it would [56:04] be really fun. But to me, I'm a bit [56:06] biased and a bit focused on AI. So [56:08] abilities API for AI is extremely [56:10] powerful and it's that kind of [56:12] cornerstone thing that you need to bring [56:14] AI to WordPress. The second thing is [56:17] common core as of WordPress 7.0 [56:20] something that exists something that's [56:21] being tested and has been tested for [56:23] some time. But the WP AAI client now [56:26] allows WordPress to interact with any AI [56:29] provider. A provider is a company like [56:32] OpenAI or Google and they give access to [56:36] state-of-the-art AI models. These models [56:39] you can prompt to go do things. These [56:42] models you can give access to tools. [56:44] Tools being abilities in the context of [56:46] WordPress. And this is something that is [56:49] swappable and pluggable. It's a [56:51] universal way to interact with AI. It [56:53] doesn't lock you into a provider. And [56:55] you can choose, hey, I want to use a [56:57] model from my desktop, locally hosted, [56:59] completely private. Or you can say, I [57:01] want to use the best and greatest model [57:02] from Google. [57:04] You can choose anything you want, which [57:07] I think is the biggest point of this WP [57:10] AI client. And what's interesting when [57:13] you take these two things together, the [57:14] abilities API and the WPAI client, you [57:17] end up with a toolkit to build native [57:20] first class endtoend agents and AI [57:23] features into WordPress as a developer. [57:27] And now that things like Cloud Code [57:29] exist and now that these projects are [57:31] extremely well documented, you can kind [57:33] of try your hand even if you don't [57:35] consider yourself a developer. And [57:37] that's really exciting and interesting [57:39] to me. The third building block is [57:41] almost more important than the first [57:43] two, but it's actually dependent on the [57:45] first two. So I put it third. This third [57:48] building block is the MCP adapter. And [57:51] first I'll mention the concept of [57:52] adapter. So we have a list of abilities [57:55] of what can happen on a WordPress site. [57:58] These abilities can be used by AI and [58:00] people alike, but AI to go do something [58:02] on WordPress. But this is within [58:05] WordPress. Outside of WordPress, we have [58:07] a bunch of clients, things like chat, [58:10] GBT, Claude Code. These clients need to [58:13] interface with the tools that software [58:16] has to offer. And the way they interface [58:18] in a primary function is MCP. MCP is [58:22] model context protocol. It's an open [58:24] protocol. It's now under the Linux [58:27] Foundation, which is really exciting. [58:29] And it allows third party clients to [58:32] talk to software. So it allows us to say [58:35] we have agents that can be built within [58:37] WordPress, but we can also make [58:38] WordPress extremely accessible to agents [58:41] outside of it. We can meet users where [58:44] they want to use AI. And this third [58:47] building block gives that holistic [58:50] coverage of AI within WordPress, AI on [58:52] the front end, AI on the back end, AI [58:54] everywhere. [58:56] These blocks make WordPress enabled for [58:59] an agentic era. Without them, we [59:01] couldn't build for AI. With them, we [59:04] can. And all of them ship with WordPress [59:06] 7.0. All of them are open source. All of [59:09] them are readily available. They're also [59:11] not forced. You can opt in, you can opt [59:13] out, but they're available regardless of [59:16] your host, regardless of your provider [59:18] or your budget. And this to me is [59:20] extremely exciting because now with [59:22] WordPress 7.0, we have a CMS that is [59:25] ready for what the future of the [59:27] internet is. And if you if you skin [59:31] scanned the first QR code and are [59:33] reading this building blocks article, [59:35] you'll be like, James, hey, where's the [59:36] fourth? This is the fourth. The fourth [59:39] building block is the AI plugin for [59:41] WordPress. This is something that ties [59:43] all three of the other blocks together [59:45] and says, "Here's what they can actually [59:48] do. here are the capabilities that we [59:50] can offer you as a user of WordPress, as [59:52] somebody who wants to generate images, [59:55] get editor assistance, generate content, [59:58] all built on these cornerstone building [60:00] blocks. So, it's a great reference to [60:02] the code and how everything works, but [60:04] it's also just like awesome to install [60:06] on your WordPress website and get AI [60:09] features. So, this is a plugin that is a [60:12] core canonical plugin, an official [60:14] plugin of the project, and it's [60:16] available for everybody, and it's an [60:18] awesome project. It's consistently like [60:20] every week releasing new versions. I [60:22] think I went on my email maybe like 10 [60:25] minutes before this, and I saw that [60:26] there's a 0.7.0 release out or coming [60:29] out. So, consistently improving every [60:32] day. And for the developers, now that [60:35] these things exist within WordPress 7.0, [60:38] No, you can use these APIs directly. You [60:40] can register abilities in your plugins [60:43] or for your clients. You can query AI [60:45] through the WPAI client. You can expose [60:48] all of this to thirdparty agents through [60:50] MCP. And you begin building your [60:53] WordPress websites, the existing stuff [60:55] that you maintain yourself or you [60:56] maintain for your clients or your people [60:59] towards this agentic era. you begin [61:02] truly enabling your people or yourself [61:05] to be a player in this world and across [61:09] the ecosystem we're seeing the adoption [61:11] of these building blocks really really [61:14] rapidly. So I put this slide together [61:16] and I wanted to find some logos. So I [61:18] started doing research and I already [61:21] know of ACF. I think that was a really [61:23] fun announcement where they announced [61:25] that they support abilities and [61:26] Elementor doing really fun stuff with AI [61:28] and WS form using all three of the [61:30] blocks. But I also started finding [61:33] community contributed provider plugins [61:35] that work with that WPAI client and [61:37] allow me to query things from Mistl [61:42] openish [61:44] AI provider in Europe, open router [61:48] and the list kept growing and growing [61:50] and growing and growing. So, it's not [61:51] just core, it's everywhere in WordPress [61:54] that's now using these blocks because [61:55] they're mature and available. And as I [61:58] put this slide together, I decided to [62:00] write a blog post. So, my blog j.cv/blog [62:04] has a post that says the AI ecosystem or [62:07] the ecosystem around AI and WordPress or [62:09] something like this. And that post got [62:11] so long and so overwhelming and had [62:13] hundreds of links. It just got me so [62:15] excited to see the actual use of these [62:18] building blocks in the real world. [62:21] Something where people are really [62:22] deriving value and building really [62:25] impressive things. This stuff is [62:26] shipping now and it's consistently [62:28] covered in TechCrunch and Ngadget and [62:31] CMS Wire. We have real products both [62:34] within the WordPress project itself, but [62:36] almost more exciting within the [62:38] community. We have real users using [62:40] these things and legitimate capabilities [62:43] being built into WordPress as we speak. [62:46] Even in the contributor day yesterday, I [62:48] saw some really fun stuff shaping up in [62:51] core AI. And this starts us down the [62:53] path, the path to the future. We start [62:57] moving from a CMS to an agentic [63:00] platform. And this is something that's [63:01] more personal to me as opposed to poor [63:04] AI is decreeing that this is becoming a [63:06] gentic platform. But it's something [63:08] where when we manage content and we [63:10] place AI and around it or near it, we're [63:13] able to build towards a future-facing [63:16] agentic platform. So WordPress isn't [63:18] becoming this AI powered CMS. It's not [63:21] becoming something that is sparkle [63:23] buttons everywhere and AI everywhere and [63:26] 14 little clippies trying to joust for [63:28] your attention. But it is becoming a [63:30] platform that agents can operate on [63:33] directly. They can access it through [63:35] MCP. They can access it through the REST [63:37] API, through the WP CLI by directly [63:40] using abilities both server side and [63:42] client side. And this platform is the [63:46] same old WordPress that we know and [63:47] love. It has the same permissions baked [63:49] in the capabilities, the user roles. It [63:52] has the same governance as the project. [63:54] It's open. We can see all of the code. [63:56] There's no back doors. the automation [63:59] and the agents respects the same rules [64:02] holistically as people have for 20 years [64:04] with WordPress and this becomes the only [64:08] agentic platform that has that deep [64:10] distribution that 43% of the internet [64:13] that's already that cornerstone which [64:15] means it opens a door to a really [64:17] exciting future for WordPress and AI [64:21] and I want to take a step back before I [64:23] talk about that future because there's a [64:25] lot more in WordPress and WordPress 7.0 [64:27] know than AI. And obviously I will talk [64:30] about AI, but I do want to shout out one [64:32] of the other flagship cornerstone [64:34] features of this impending release, [64:37] collaborative editing. This is something [64:39] that's really exciting to me. I'm sure [64:40] there's going to be a lot of [64:42] demonstrations and chatter and [64:44] discussion around it during the Word [64:46] Camp Asia inerson sessions. But [64:49] collaborative editing to me is really [64:51] interesting because it was designed kind [64:53] of before we saw all of this AI stuff [64:56] shape up. It was designed for humans to [64:58] work together in a document. And it's [65:00] really fun. It's really awesome to play [65:02] with and explore with people that you [65:04] work with, but also yourself. I had like [65:06] 18 different windows open and I was [65:09] bouncing between all of them. It was [65:11] awesome. You have your blocks that sync. [65:13] You have your paragraphs that sync. It's [65:16] really impressive. But now with AI, some [65:19] of those collaborators aren't going to [65:20] be human. And what's interesting to me [65:22] is that we can use the same [65:24] infrastructure that we've built for [65:26] people for agents. We have the same [65:28] permissions. The agents can now work [65:30] alongside you in the editor. We have [65:32] content guidelines that can steer them. [65:34] We have abilities that can be chained [65:36] together into workflows. And all of this [65:38] to say, we're building these building [65:41] blocks for WordPress. They're shipping [65:42] into WordPress 7.0. There are additional [65:45] features in WordPress 7.0. There's more [65:47] work to do in AI with the AI plugin, but [65:51] there's also more work to do in general. [65:53] As the WordPress platform moves into [65:55] this agentic era, we have options to go [65:58] and explore and build and continue to [66:00] push the future of what this platform [66:03] truly can be and truly can do. So, to [66:05] me, one of those options is [66:07] collaborative editing. Collaborative [66:08] editing with both people and agents. But [66:10] that's just one example of many exciting [66:14] future states of WordPress. [66:16] And that future state starts pointing [66:19] towards the future state of the web [66:21] which I am calling the agentic web. And [66:23] you'll see the agentic web mentioned in [66:26] many marketing documents all over the [66:28] place. I keep seeing it pop up. But to [66:30] me the agentic web is simply the [66:32] internet that exists that beautiful [66:34] graph with a new player on the board. [66:36] Instead of only people now we have [66:38] agents. Those agents being these LLM [66:40] derived systems that have agency that [66:43] can do things for you. We start moving [66:46] from questions to goals and we start [66:48] moving from synthesis to action. That [66:50] changing consumption model that I spoke [66:52] about. But the agents that are changing [66:54] this consumption model are already using [66:57] the same infrastructure that the web has [66:59] that the web ships with that the web has [67:01] had for 30 years. We're not [67:03] rearchitecting the internet in service [67:05] of these agents. These things operate [67:07] browsers. And a great example of these [67:09] things operating browsers is uh Claude [67:12] in Chrome, the chat GBT computer use, [67:16] the perplexity computer. They click, [67:19] they use the area labels, they use the [67:21] internet that already exists. On the [67:24] other hand, other agents are accessing [67:26] websites or these entities on that [67:28] beautiful graph programmatically and [67:31] they're accessing this through MCP, [67:32] which we talked about as a core building [67:34] block. But they're also accessing this [67:36] through traditional means of software [67:38] interaction. And we're seeing a [67:40] resurgence of CLI commands for local [67:43] agents to use and the WP CLI being [67:45] perfectly positioned for agents to [67:48] interact with. We're seeing the REST API [67:50] and application passwords get used more [67:52] and more and more. I'm also seeing a new [67:55] generation of browsers, specifically one [67:58] coming from Google, but I see a few of [67:59] these where they actually consume [68:01] content from the site and they create [68:03] ondemand interfaces, but it's all using [68:05] the same stuff. It's all using the same [68:07] web that's always existed. But that also [68:10] means representation begins to shift. [68:13] Representation, the reason why I even [68:16] started blogging, the reason reason why [68:18] I started increasing my outcome output [68:20] of blogging. When a human visits your [68:23] site, you control the experience. If you [68:25] go to my blog, you'll see an about me [68:28] that I've crafted very specifically to [68:30] explain what I do and who I am. You'll [68:33] see my personal experience, my [68:34] professional experience, the albums that [68:37] I like listening to. You'll get a sense [68:38] for who I am because I've designed my [68:41] site. I have set the layout, the flow, [68:44] the messaging, the way the pages link [68:46] together. So, I control that experience. [68:49] And everybody controls their own [68:50] experience on the web today. [69:10] Folks, we're working to resolve the [69:11] glitch. So stand by everybody. Thank [69:13] you. [69:53] Hey, [70:00] hey, hey. [70:14] Hello Mumbai. [70:16] Uh a very exciting session. There is a [70:19] technical uh difficulty but it's being [70:21] resolved. In the meantime, I would like [70:24] us to reflect on uh what James has [70:27] talked about [70:30] the building blocks of um AI in [70:33] WordPress. [70:36] Who can mention the first one? [70:41] Sorry. [70:43] Yes. Abilities API. The second one from [70:47] the farthest end. [70:53] Okay. [70:54] Here. [70:59] My favorite is the uh the the WordPress [71:04] a AI plugin. The AI plugin. Uh-huh. [71:10] Sorry. [71:14] Yeah. Sorry. Yeah. You see, we have we [71:18] have a favorite, the WordPress AI [71:20] plugin. [71:36] >> Okay. All right. Um, [71:42] what's your favorite [71:44] apart from I know most of us the [71:46] favorite is the uh the WordPress AI [71:49] plug-in. [71:53] Let me get the third building block [71:54] please. The third building block [72:00] MCP adapter and um the workflows API. [72:05] fourth one. [72:11] All right. Now, [72:14] um [72:15] who can explain exactly how you can be [72:19] able to use uh the abilities API? [72:23] All good. Okay. Thank you. So, um I [72:26] think we are good. [72:28] Let's get back to the session. [72:34] I think we're back. I think this [72:36] presentation was so good that my agent [72:40] sabotaged me because I was going to [72:41] share all of its secrets. So, we'll jump [72:44] right back into where we left off. [72:46] Actually, we'll go back one more slide. [72:49] So to me, when an agent visits your [72:52] site, the experience that you've crafted [72:55] yourself, you've put together this [72:57] beautiful flow of information, you've [72:59] created images, that experience begins [73:02] to change. The agent interacts with your [73:05] content differently. [73:07] And that introduces an interesting [73:09] concept to me, which is maybe with all [73:12] these agents interacting with my site, I [73:15] want an agent on my site to represent me [73:18] to them. [73:20] And this is a concept that I've been [73:21] calling the site delegate. This is [73:23] something that I blogged about a lot, [73:25] which is a site delegate. A delegate [73:28] being something that represents you, [73:30] something that understands your content [73:32] and your capabilities and who you are [73:34] and what your personality is. This is [73:36] something that can start interacting [73:38] with these visiting agents on your [73:40] behalf. And instead of these visiting [73:42] agents simply getting the content and [73:44] leaving the site, they begin to [73:46] interact, to ask questions, to go [73:49] deeper. And your delegate begins to [73:51] answer these things to negotiate on your [73:54] behalf. So the web begins to evolve from [73:57] a collection of documents to a network [73:59] of interacting agents. That visiting [74:02] agent represents the user and the site [74:04] agent represents the entity. And they [74:06] communicate, they exchange information, [74:09] they come to outcomes. and you have more [74:11] control over how you're represented in [74:13] this agentic internet. And because it's [74:16] all WordPress and because these four [74:17] building blocks that I tuned back into [74:20] and heard y'all winning on that quiz, [74:23] all of those building blocks result in [74:26] this being a possibility, this being [74:29] open content that you own, protocols [74:33] that we all know and love like MCP [74:36] existing under the Linux Foundation, [74:39] agent to agent as an open standard. [74:41] Nobody's locked into the ecosystem. [74:44] Everybody's able to control their [74:45] representation and we still maintain [74:47] that beautiful decentralized web of [74:50] information. And to me, I call this [74:53] democratizing representation. I see this [74:56] as almost the fourth step of WordPress. [74:59] What needs to happen in the future? 20 [75:02] years ago, WordPress made it possible [75:04] for anybody to publish on the web. You [75:06] didn't need a developer. You needed five [75:07] minutes to install the thing and you [75:10] started publishing. uh you had something [75:12] to say and you said it. And if you look [75:14] at my first blog about amateur radios, I [75:17] was saying stuff about amateur radios [75:19] and it was something that I wanted to [75:21] do. I was enabled by WordPress to do [75:24] that. The publishing was democratist. [75:27] But now the same thing needs to happen [75:29] for the digital representation. And I [75:32] think that this has always been the case [75:33] on the web. What's the best way to [75:35] represent yourself in a pre-AII era? has [75:38] been creating beautiful pages, working [75:41] through the content flows, ensuring that [75:43] you understand how a user walks through [75:46] your site. But now with agents entering [75:49] the chessboard being a new player on the [75:52] map, we need to do the same for digital [75:55] representation. [75:56] And without accessible tooling, the [75:59] benefits of this agentic web, which is [76:01] happening, they'll concentrate to those [76:03] who have the resources to build for it. [76:05] they'll end up in those very prettyl [76:07] looking but messy closed walled gardens. [76:11] But WordPress is the solution to this. [76:14] WordPress can change that. We have the [76:16] building blocks. We're building towards [76:18] this interesting future where agents [76:21] interact on the web, where our websites [76:23] have MCP servers, where we have agent to [76:26] agent, where we're able to interact [76:28] locally through the WP CLI. It's all [76:31] there for WordPress to use. So our local [76:35] plumbers, our journalists, our tiny to [76:37] huge WooCommerce stores, the portfolio [76:40] websites, the J.CVS, blogs, everybody [76:43] gets access to the same stuff, the same [76:46] protocols, the same interoperability, [76:49] the ability to look at the actual source [76:52] code itself. They have the ability to [76:54] shape their representation. [76:57] So what you can do today is very, very [77:00] related to those building blocks. You [77:01] can register abilities in your plugins [77:03] if you're a developer. You can do the [77:04] same if you're an agency. You can expose [77:07] the functionality of your site through [77:08] MCP. You can use MCP plus Claude or [77:12] ChatgBT or Copilot to go and maintain, [77:15] manage, extend, learn more about your [77:17] WordPress sites, clients sites, your [77:20] sites become these programmable [77:22] interfaces. And this is the next era of [77:24] WordPress. This is where everything is [77:26] headed. and you can participate right [77:28] now because it's in WordPress 6.9 and [77:30] it's coming to WordPress 7.0. [77:33] For the creators and the publishers, [77:35] things are shifting, things are [77:37] changing. But I think the best course of [77:39] action is to keep creating, to keep [77:42] presenting who you are and why you do [77:45] it. You being the entity, you being the [77:47] person, but you being the business or [77:49] the e-commerce stop. Structure your [77:53] content intentionally. Make sure that [77:56] you're putting your best foot out there. [77:58] The quality of the representation right [78:00] now and into the future as we move [78:03] towards this agentto agent model that [78:05] may exist really depends on the quality [78:07] of the content. Content is again it [78:09] would kind of come back to the SEO world [78:11] of write good stuff, link that stuff, [78:14] make it well structured, you'll do just [78:16] fine. And for the freelancers and the [78:19] site builders, go and play with 7.0. Go [78:22] and update to, well, I wouldn't [78:24] recommend this on a production site, but [78:26] go and update to WordPress 7.0 when it [78:28] comes out. Go and experiment with the [78:30] beta using WP Playground. You'll see [78:32] that the AI just works. There's API keys [78:35] that get set up in the central [78:37] connectors plugin. If a host sets it up [78:40] for you, it's just there. There's no [78:42] plug-in hunting. And now a website is [78:46] enabled by AI. And for the hosts, go and [78:50] take a look at this WP AI client. You [78:52] have the ability to bundle AI credits [78:54] into your plans because of this unified [78:56] approach. And what that means is if [78:58] every host offered a small amount of [79:01] free AI credits to their users, which is [79:04] extremely possible with this WP AI [79:07] client, everybody who ever uses [79:10] WordPress gets access to AI out of the [79:13] box. And imagine how powerful that could [79:15] be for those creating, for those using [79:17] AI in novel ways through their plugins. [79:20] That would be an incredible world that [79:22] we live and operate in. So right now [79:24] today, you can update to 7.0, the beta. [79:28] You could play around with it in WP [79:30] Playground. You can go to [79:31] makewordpress.orgai [79:33] to understand more about what we're [79:35] actually doing dayto-day in that core [79:37] group. You can go to the core AI Slack [79:40] and marvel at the really fun stuff that [79:42] happened yesterday at the contributors [79:44] day. You can install the AI plugin. You [79:47] can also tear it down and take a look at [79:48] the code because it's a really great [79:50] representation of how these things work [79:52] together. You can point Claude at your [79:54] site and say, "Tell me about myself. [79:56] Help me interlink things. Am I missing [79:58] images? Should I put alt text here? I've [80:01] been using Claude and the MCP stuff. [80:04] It's incredible." And just see what [80:06] happens. Explore. Be in the [80:07] conversation. Be in the arena. [80:10] Understand what this agentic web truly [80:12] could look like because you're actually [80:14] a participant in the early stages of it. [80:17] To me, it's always been empowerment over [80:20] replacement. I want people to be [80:22] empowered to create their best work [80:25] instead of replaced with AI. So, we're [80:28] not adding the sparkle buttons all over [80:29] WordPress. If you want it, there are [80:30] many plugins that will add it. But, [80:33] we're building the tools that make [80:34] everybody more effective at creating. [80:35] And we're building the foundations that [80:37] enable just that. [80:39] The website concept that's not changing. [80:42] The the reason for being digital, the [80:45] reason for being online isn't changing. [80:48] But the people on that board are the [80:50] things on that board are. We now have [80:51] both humans and agents. However, the [80:54] people that build on and for and within [80:56] WordPress, they are more important than [80:59] ever. That open web is worth building [81:02] for. That beautiful graph that I come [81:04] back to is truly beautiful to me and [81:06] WordPress is how we've always built for [81:08] this web. It's our goal and core AI [81:11] absolutely but within WordPress in [81:13] general for this to be the way that we [81:14] do it far into the future as well. So [81:17] let's build that open agendic web. Let's [81:20] do it together. Thank you very much for [81:22] dealing with my interruptions and I [81:25] really appreciate y'all taking the time [81:27] to go and watch this presentation even [81:30] though I was digital. I just want to [81:32] mention that my favorite parts of word [81:34] camps are the hallway tracks. They're [81:35] the things where when I walk off this [81:38] stage, I get stopped by people and have [81:41] these crazy unique conversations and [81:43] these ideas. And as I mentioned in the [81:45] opener, I am bummed out that I can't be [81:47] doing that here. But I do invite you to [81:49] have these conversations. Think about [81:51] what the future of the web looks like. [81:53] Think about how WordPress plays a part [81:54] in it. explore the building blocks, but [81:57] explore the future of the CMS as you [81:59] walk around in person today and tomorrow [82:02] and really enjoy what looks to be a [82:04] great word camp Asia. So, thank you very [82:06] much. You can find me and you can ask me [82:09] these questions at jamesw leage [82:11] specifically on X. Feel free to DM me. [82:14] You can read more about a lot of this [82:16] thinking on j.cv. [82:18] And I don't know if we have time for [82:19] questions, but if we do, I'm happy to [82:21] answer them. If not, enjoy your word [82:24] camp. Awesome. [82:30] >> Thank you, James. [82:33] Do we have a question? If you have a [82:36] question, you can walk uh to the aisle. [82:37] We have microphone there and a [82:40] microphone there. [82:43] Possibly we can only take uh two [82:46] questions. [82:47] Max, [82:54] do you have a question? [82:56] So, my key takeaway from here is um [83:00] after this I'm going to update to [83:03] WordPress 7.0. [83:05] Of course, it's not it's not live. So, [83:07] I'm going to update uh to WordPress 7.0 [83:11] on [83:13] uh not a production site. And then I'm [83:16] going to install the WordPress AI [83:18] plugin. [83:20] And then I'm going to try and point my [83:24] favorite AI client. In this case, I'm [83:27] going to try cloud. I don't know what [83:29] your favorite is and then I'm going to [83:32] see exactly what's going to happen. I [83:34] don't know what you're going to do, but [83:36] at least try and do the same. and to be [83:38] able to also follow along what is [83:40] happening with the WordPress and API uh [83:43] building try and join the core AI slack [83:48] so as to be able to be up to date with [83:51] everything that is happening. If we [83:53] don't have any questions I would like to [83:56] thank James for our wonderful pres our [83:58] wonderful presentation and word campa [84:01] please help me thank gems with a round [84:05] of applause. [84:07] Thank you so much. [84:18] Heat. Heat. [85:59] Okay. [87:01] that [87:03] I don't [88:05] Namaste. The program will begin in 10 [88:08] minutes. Please silence your mobile [88:10] phones. [89:26] Hey. Heat. Heat. [90:26] All right. [90:37] Down. [91:02] Hey, [91:30] Heat. Hey, Heat. [91:38] All [92:00] right. [92:05] Heat. [92:39] WooCommerce is the e-commerce platform [92:41] built for total flexibility. As you [92:44] grow, you can adapt your online store to [92:46] meet your needs with no limits or [92:48] compromises. [92:49] >> WooCommerce for us has been a very [92:50] flexible, customizable platform for us [92:52] to scale our business. So that's why we [92:54] picked in the beginning and it was [92:56] recommended us for to stay and we've [92:58] really grown with it. It's been [92:59] fantastic. That's why more brands trust [93:01] Woo Commerce to power their success than [93:03] any other platform. Because your [93:05] business is yours truly only with Woo [93:08] Commerce. [93:24] Heat. [93:41] Heat. [94:01] The program will begin in 5 minutes. [94:04] Please take your seat and silence your [94:06] mobile phones. [94:12] You [94:17] are [94:22] you. [94:55] and hey [95:21] are feel. [95:26] Heat. Heat. N. [96:01] No one want you. [96:59] Namaste. The program will begin in 10 [97:01] minutes. Please silence your mobile [97:04] phones. [97:56] Hey, hey, hey. [100:09] Hey, [100:13] hey, hey. [100:30] Do I [100:34] know? [103:02] The program will begin in 5 minutes. [103:05] Please take your seat and silence your [103:07] mobile phones. [103:34] Hey, [104:17] Heat. Heat. [105:01] Baby, daddy. [106:19] baby. [106:21] Hey, hey, hey, hey, hey, hey, hey, hey, [106:21] hey, hey, hey. [106:56] Hi, I would like to request everybody if [106:58] you could come in the front seat not at [107:01] the back. We would like to see faces [107:04] peoples. Can I request everybody [108:15] Good morning ladies and gentlemen. I am [108:17] Minatam joining you from all the way [108:20] from Nepal. I'm a digital marketer by [108:23] profession and a proud member of this [108:25] wonderful WordPress community and I'll [108:28] be your MC for the next session. And [108:30] before I start, I'd like to share a few [108:33] things. [108:36] Close captioning is available on your [108:38] personal device for all session. As you [108:40] can see the QR code on the screen, [108:42] please scan and follow the prompts to [108:45] begin receiving captions. And the second [108:48] thing is outside of this building is a [108:50] US embassy premises and it is not [108:52] allowed to take photos and videos. So I [108:55] recommend that you do not click any [108:58] things. [109:01] And for the second session of the day, [109:04] we have something really special, [109:09] a fireside chat that I believe that many [109:11] of us are looking forward to. [109:16] The WordPress touches millions of people [109:19] around the world. But where is it [109:22] headed? AI, community growth, [109:26] contribution pathways and what future [109:29] looks for us in this ecosystem. These [109:32] are the things these are the kind of [109:34] questions we are about to dig in. And to [109:38] help us with this conversation today, we [109:40] have Mary Hubard, an executive director [109:43] of WordPress. And I believe that we do [109:46] not need to introduce her but let me [109:48] share a few things about her. She brings [109:51] 20 years of experience that 20 years of [109:55] industry experience and has previously [109:57] held senior role at Tik Tok, eBay and [110:01] Walmart. She's here to share her vision [110:04] for the WordPress and where it is [110:07] heading. And to guide us this [110:09] conversation, we have a moderator, [110:11] Shilpa Shaha, a founder of Made with [110:14] Heart Innovation, a WordPress committee [110:16] member since 2011 and the voice behind [110:20] WP cubicles. And at the end, we also [110:23] have a Q&A round. So if you have any [110:25] query that you wanted to ask, please [110:28] flee feel free to do that. Now let's [110:32] everybody put your hands together and [110:35] give a warm welcome to Mary and Sila. [110:53] >> Oh, cool. [110:54] >> I get the one. Keep it. [110:57] >> Maybe I could keep keep the pillow. I [110:59] keep saying that I [111:00] >> Okay. [111:00] >> I haven't got to keep in. [111:02] >> I like her cushion better than mine. [111:05] >> You want to trade? [111:06] >> Can we swap? Yeah, that's okay. [111:08] >> Right. So, here's Mary [111:11] >> and we're going to have a quick chat and [111:14] feel free to ask questions. Hold on to [111:16] them till the end and we have these mics [111:19] over here. So, if you have any question [111:21] that you just can't wait with, please [111:23] line up over here and we'll open it up [111:26] to the audience in the end. So, let's [111:28] get started. Yeah. Okay. So, I have this [111:31] little book here. So, you can tell how [111:32] old I am that I still have a book and [111:34] not a phone. So, yeah. So, like Mina [111:38] said, you've worked in the retail space, [111:42] you've worked with Walmart, you've [111:44] worked with Tik Tok, uh you've worked [111:46] across Woo Commerce and WordPress.com [111:49] and now the WordPress project, right? So [111:52] all this experience I mean what lessons [111:55] have you learned you know through this [111:57] experience that helps you understand how [112:01] a complex WordPress ecosystem really [112:04] works you know how to make it work. [112:06] >> Yeah. So my career has been very focused [112:08] on product development [112:10] >> and so and e-commerce. So like I grew up [112:13] in um I grew up really developing [112:16] products to help sell and close [112:18] platforms [112:18] >> right? So like in eBay Motors, we [112:21] developed out the first programs um [112:23] which led to the first PIM systems. Then [112:26] I went to Walmart, created the first [112:27] seller center, developed the first [112:29] third-party marketplace. So my whole [112:30] career was very e-commerce focused and I [112:33] built with great product teams the [112:36] majority of the marketplaces that exist [112:37] today, especially like us. [112:39] >> Um [112:40] >> so what WooCommerce was a closer to what [112:43] you were doing. [112:43] >> Yeah. So, Woo Commerce was very much my [112:46] first move into what open source truly [112:49] was. And it was a a very jarring and [112:52] shocking like experience because [112:55] >> um when I first started, I came in to [112:58] build up the WooCommerce marketplace and [113:00] to understand the dynamics um of 1P 3P [113:03] and how all of that played it was just [113:05] in my in my being. I knew and understood [113:07] that, right? But then in an ecosystem [113:09] that's open source, it's very different [113:11] because it introduces a different type [113:13] of um a different end user, a different [113:15] type of dynamic um and a different I [113:19] guess uh [113:20] >> it's a different way. [113:21] >> It's Yeah. And and it's it's so [113:23] interesting too because like what you're [113:25] promoting or what you're really selling [113:26] isn't necessarily the product itself, [113:28] but it's support, it's updates, and then [113:31] how do you foster um a really clean and [113:34] clear ecosystem where there are so many [113:38] different players in the game and how [113:40] every single person deserves a right to [113:42] be there. Um and so what I really [113:45] learned was that ecosystems matter and [113:47] in order for those ecosystems to [113:49] flourish, you everybody needs a seat at [113:51] the table. Um, and we need to be working [113:53] together. And what that means is [113:55] community. So the people who are [113:57] building the agencies, whether that's a [113:58] company, for-profit, or whether that's [114:01] somebody developing something out, wants [114:03] that premium model. Um, it's the people [114:05] that are building and that's how they [114:06] flourish. So it was a real um quick and [114:10] steep learning curve, [114:11] >> I'm sure, [114:11] >> for open source. But it's what led me to [114:14] to WordPress ultimately um and what [114:17] really got me excited to be a part of an [114:20] open source project because once once [114:22] that door is open, you really don't want [114:24] to [114:25] >> want to see it. Yeah. Want to go back. [114:27] Um [114:27] >> I know the feeling. Yeah. [114:28] >> Yeah. [114:29] >> Yeah. [114:30] >> So that sounds good. So [114:32] >> and so u recently we've seen efforts [114:35] like uh campus connect. [114:37] >> Yes. where you know we are trying to [114:39] bring WordPress to universities, [114:41] colleges and trying to engage with [114:43] students. So how do you think you know I [114:45] mean what role do you see that playing [114:47] in building say the next generation of [114:50] WordPress contributors or professionals [114:52] really? [114:53] >> Oh so like shout out for WordPress [114:55] campus connect to India because it's [114:56] homegrown started here [114:58] >> are you there? started here and it's [115:00] something [115:02] >> um quick ideas at the nan and and Fuja [115:04] and teams that were able to to really [115:07] focus in on students and youth and grow [115:09] that and I think I talked about it when [115:11] I first came back uh I think I talked [115:14] about it in Europe with with Matt uh but [115:17] the idea that something so small um [115:20] continuously running can really snowball [115:22] into many different uh opportunities and [115:26] they all interconnect eventually so I [115:29] actually think that these student [115:30] initiatives are probably the most [115:31] important initiative that WordPress has [115:33] today. If you look at um if you look at [115:36] WordPress globally, you know, it's a [115:39] great product. Um and it's 20 years old. [115:42] Um and so we have an aging contributor [115:46] base. Um we have a hockey stick coming [115:49] with AI around learning and [115:51] understanding. And so we want new eyes. [115:54] We want fresh eyes. We want um to teach [115:56] the next generation of what WordPress is [115:58] and why it matters and why it's so [116:00] important. Um and then we also know that [116:04] like WordPress ecosystems, it thrives on [116:07] community and it thrives on meetups, [116:09] which is why meetups are so important [116:10] and campus connect events are events. [116:13] They're people making connections around [116:15] the world. And that really harbors a [116:18] very positive positive um experience [116:22] that can carry on. And so Campus Connect [116:25] is just one of those avenues, right? But [116:27] we also have um WordPress credits, which [116:29] I I often just refer to as curriculum. [116:32] And what we're seeing is a clean and [116:33] clear pathway from campus connects right [116:36] into the curriculum where we're [116:39] embedding WordPress in to the schools. [116:42] Um and then they actually push into [116:44] student clubs where we're starting to [116:46] see circular environments. [116:48] >> So they get credits for contributing to [116:50] workplace. [116:51] >> Yes. And so so the the There's different [116:54] pathways around education and like one [116:56] of the largest ones that we're doing [116:58] today that I announced I think in [116:59] February is we're doing with the AI [117:01] microcredential [117:03] >> and that's also a student focused piece. [117:06] It's you have campus connect which are [117:08] event-based systems which ultimately get [117:09] you into the credited system which is [117:11] inside schools your contribution hours [117:14] whatever major you have um goes into [117:17] credits to graduate and we're currently [117:19] in 18 schools and growing. there's a a [117:21] great movement that's happening in in [117:23] Europe right now and we've just [117:25] introduced a couple of schools in the [117:26] US, right? [117:27] >> Um and once you're in in there, you [117:29] start to um you start producing like I [117:32] think there's over uh one student did [117:34] 4,000 translations, [117:36] >> right? Huge. It's huge. And the and the [117:38] number of hours that are uh are [117:40] projected just on this current 18 [117:42] schools are thousands upon thousands of [117:45] hours um that help through this [117:47] connection and this AI micro credential. [117:50] Now you start to tie standards. So [117:52] everyone's paying attention to AI. How [117:54] can we how can we help tie that back [117:56] into education and start um helping the [117:59] entire WordPress community base level up [118:02] with AI and then bring the students into [118:05] these paths so they start at the at an [118:07] equal level um and then we can use [118:09] WordPress and its dominance to start [118:12] showcasing exactly what that level means [118:14] globally. So if that certification or [118:16] that credential it's like actually the [118:18] entire WordPress community globally is [118:20] telling us if you have this credential [118:22] this is that standard this means. [118:25] >> So there's a lot of opportunity that's [118:27] coming out of this one simple idea [118:30] homegrown here. Um and I I really [118:32] encourage everybody to kind of see how [118:34] that can be applicable applicable to uh [118:37] to anything that you're doing because [118:39] it's it's going to get larger and larger [118:42] >> and everything interconnects. Right. we [118:44] were talking about. [118:44] >> Absolutely. Absolutely. [118:46] >> Contributor dashboard, um, visibility [118:49] there, credits, these micro credentials, [118:51] like all of those things are to promote [118:53] the next generation, but also level up [118:55] our current contributor base and level [118:57] up the people that are here and showing [118:59] up every day. [119:01] >> Cool. [119:02] >> Yeah, I'm very excited about it as you [119:04] go off. [119:05] >> It's hard not to be. I I totally see [119:06] that. So [119:08] where do you think um you know WordPress [119:11] is doing really well and where do you [119:14] think we might be holding ourselves [119:16] back? I mean [119:18] >> tell us the good, bad, and ugly. [119:20] >> I think what we're doing really well I [119:21] mean we're dominant uh you know and I [119:24] think that in order to stay with over [119:26] 40% of the web um that's no easy feat. [119:29] So I think that we do a lot of really [119:31] good excuse me we do a lot of really [119:34] good um like movement and ideas and [119:37] shake I think where we struggle on the [119:39] product side is it's it's difficult to [119:41] get started right [119:42] >> um it's difficult for so a user to land [119:44] like we're still downloading you know [119:46] you're downloading the WordPress [119:47] software you're moving it I think um we [119:49] can do better there I think a lot of the [119:51] stuff with AI and my mywordpress.net and [119:53] that can really help somebody get [119:55] started in the community. I think we uh [119:58] are a little slow. We're slow to make [120:00] decisions. I think we need to be more [120:02] transparent. We need to lean on project [120:04] leadership to make to be very decisive [120:06] and move. And then for new contributors [120:08] or people who are coming in, we're not [120:10] very um we're not very easy to get [120:13] started. I think that we need we need [120:16] better pathways for contributors to come [120:18] in and get started right away. Yeah. [120:20] >> And um [120:22] >> but I think there contributor days, you [120:24] know, like the one yesterday. I mean, I [120:26] think that's a good place where you I [120:28] mean, even if you're trying to get it [120:29] online, it might be difficult, but you [120:31] know, these contributor days is where [120:32] you [120:33] >> can actually meet and well and it's just [120:35] it's get started. [120:36] >> It it again, it comes back to that [120:38] community. People want to come together. [120:40] They want to meet with other [120:40] contributors and do. But there's a great [120:43] project that's happening right now [120:44] called contributor pathways in the [120:46] project itself. And the idea is that uh [120:49] you take we take the handbooks like the [120:51] team handbooks and we start to show very [120:54] clean and clear paths for somebody who [120:56] comes in like these are the pathways [120:57] that you should start and they're very [120:59] easy to kind of get up and then [121:01] immediately show the impact of your [121:02] work, right? [121:03] >> Because contributors want to be [121:05] supported. They want to see the impact. [121:06] They want that visibility. And so the [121:08] idea that you can come you understand [121:10] where you maybe you choose a team, maybe [121:12] you don't. Um there was a a new concept [121:15] around instead of choosing a team like [121:16] you can actually choose a small project [121:18] within the team [121:19] >> right [121:20] >> so [121:20] >> so that project has a goal and [121:22] milestones and [121:23] >> yeah so it's very clean [121:24] >> there's a clearer sense of achievement [121:26] so [121:26] >> and so you can see it as you do it and [121:28] so not only have you contributed but you [121:30] you've executed on something you could [121:31] be proud of that and say oh yeah look at [121:33] what I have done and then the learning [121:35] that goes into this while you're [121:37] contributing right um and it ties in so [121:40] so cleanly to that contributor dashboard [121:42] board, [121:43] >> right? [121:43] >> Another homegrown project where we're it [121:45] came out of a community summit a couple [121:47] of years ago, [121:48] >> but you know there there's the idea that [121:50] >> it's very hard to track non-code [121:53] contributions and the visibility of that [121:56] within the project is not very robust [121:58] and so how can we get more visibility to [122:00] that whether that is support or even [122:03] presentation of monikers or badges but [122:05] on the other side which I think is [122:07] powerful for any team lead or um you [122:11] know me I I want to see where people or [122:13] why people are leaving [122:14] >> and so there's the these idea around [122:17] milestones when somebody comes in and [122:19] they do their first contribution or they [122:21] show up to a meetup what is the [122:23] stickiness [122:24] >> and where do people exit and then how [122:26] can we interject to either provide a [122:29] mentor or a clearer path and get those [122:31] feedback loops in just to support the [122:33] contributors [122:34] >> like get a buddy whom you can [122:35] >> right or like and so like on the two [122:37] sides of that there's two sides of that [122:39] right there's the the contributor [122:40] experience experience. Um, which we [122:42] should be we should be treating these [122:45] >> all as products. Like I think WordPress [122:47] tends to treat a lot of things like [122:49] infrastructure rather than product. But [122:51] like when you come in and you're [122:53] starting to do your first contribution, [122:54] what would that look like? What would be [122:55] the user experience? [122:57] >> Right? [122:57] >> Then the contributor dashboard helps you [122:59] understand what you're doing, your [123:01] growth trajectory, how you can rise to [123:03] leadership within the project and try to [123:05] make that cleaner and clear, [123:06] >> right? And on the opposite side for team [123:08] leads, it's like what are the gaps that [123:09] you your team might have that are making [123:12] people not want to contribute to your [123:13] team or that isn't very supportive and [123:16] how can we help copy and paste things [123:18] across [123:19] >> bridge the gap. [123:20] >> Yeah. Bridge those gaps and and work it [123:21] out. [123:22] >> Right. [123:24] >> Okay. So, uh, there's a lot going on, [123:27] you know, Gutenberg, um, AI and [123:29] onboarding and releases and so what do [123:32] you think we should really be focusing [123:33] on right now, [123:35] >> you know, with all of this with [123:37] everything that's going on? Um, [123:39] >> you know, I I think we need to do small [123:42] things or we should we should have fewer [123:43] focused items, right? And I tend to [123:46] split um everything between like product [123:50] like core product and then community. So [123:52] like we have a lot of work that's being [123:54] done in the community that I think is [123:56] great and then we have a lot of product [123:57] work that's being being done that I [123:59] think is really great. Um I do think [124:02] that AI and what we're doing with AI [124:04] from a community space. Um we should be [124:07] leveraging our human capital. I think [124:08] it's very important to leverage our [124:10] human capital and then integrate AI into [124:12] anything that could be operational and [124:13] very clean and clear there. I think it's [124:16] important to get on the forefront of [124:18] that now. And then I also think that on [124:20] a product level we do need to be [124:23] thinking about the next iterations the [124:25] future versions of how WordPress is [124:26] going to scale will be with being the [124:29] the open web AI open web right we should [124:33] be thinking about um uh where WordPress [124:37] is going to take a seat at the table [124:38] globally when it comes to the future of [124:40] web development and the future of CMS [124:42] and how how we're looking at it. Um, we [124:45] want [124:47] society is shifting for how they're [124:49] interacting with the web and with [124:51] WordPress. I think we're going to end up [124:53] we will very much lead in this space if [124:56] we can do this right and we can do it [124:58] quickly. People are going to want a [125:00] voice. They are going to want their own [125:02] piece of the web as platform controls [125:05] starts to minimize and AI needs [125:07] something to scrape and AI needs [125:09] something to pull. And this is where [125:10] WordPress can really really grow. Um, we [125:15] push that. We make sure that we are the [125:17] open default. And also as as individuals [125:20] who work within WordPress every single [125:22] day, what are we doing to explain that [125:24] or promote that outside of our [125:26] ecosystem, [125:27] >> right? [125:28] >> Because AI the they're not choosing us. [125:31] We want to make sure that they are like [125:33] this is why the foundation and agnostic [125:35] like we we need to stay here in this in [125:38] this space to ensure that people are [125:39] aware that we're building the [125:41] foundations for the future of the open [125:43] web with with AI. [125:44] >> Yeah. [125:45] >> And that's a very good um [125:48] it's very important. I think it's very [125:50] important for us and and the speed at [125:53] which we achieve that is very important [125:54] as well. [125:55] >> Absolutely. I I second that. So, so if [125:58] you had the chance to rethink or uh [126:02] redesign any one aspect of WordPress, [126:05] you know, uh in order to better serve uh [126:07] the WordPress community, developers, uh [126:10] professionals, uh businesses, uh what [126:12] would that be? And I encourage you not [126:14] to be practical. So, if you had a if you [126:18] could just a magic wish, [126:20] >> um, [126:22] you know, I would pro probably something [126:25] with the plug-in directory. [126:27] >> Ah, I [126:27] >> I think I I spoke a little bit about [126:29] this just a second ago, but I do think [126:30] that we should stop treating the plug-in [126:32] directory as infrastructure and we [126:34] should really productize it. I think [126:35] there's a lot of opportunity here for [126:37] people and businesses within the [126:38] ecosystem to do something, right? [126:40] >> There's been some changes like there's [126:42] been some shifts recently. Um, but I [126:44] think that we can really we can really [126:46] reimagine what that looks like. [126:49] >> Um, some of the other things like that [126:51] if I could just shift uh I I [126:55] every I I've made it a point to where [126:57] when I start finding pages that feel a [126:59] little dated, I just pull up Claude and [127:01] I try to redesign and shift some of the [127:03] pages that we're working on. And I'm not [127:05] great. Uh, but like I think that we can [127:08] modernize a lot of stuff very quickly as [127:10] well. So like um recently back to jobs [127:14] and the microcredential and stuff I had [127:16] run into the the jobs word [127:18] jobs.wordpress WordPress, right? and um [127:20] very quickly within 10 minutes was able [127:22] to redesign that page and I submitted it [127:25] to the team and of course they came like [127:27] oh we don't like yours but we did [127:29] something better and with within 30 [127:30] minutes there was a brand new like [127:33] >> theme based thing and I think that um I [127:35] would make a lot of changes like that [127:37] too because [127:39] >> we are what we present to the world and [127:41] if you have a very dated page it doesn't [127:43] reflect well on our software or like [127:45] what we're doing and how and I encourage [127:47] everybody to do that and if I can do it. [127:50] I think many people can do it cuz you [127:52] know that's good. [127:55] >> So uh looking ahead to the next uh say [127:58] 12 to 18 months um sorry what does [128:01] that's okay. We all have a cough. So uh [128:05] looking ahead to the next say 12 to 18 [128:06] months uh what does success for [128:09] WordPress look like to you? [128:12] >> I mean it's just 12 to 18 months. Yeah, [128:14] I think well [128:16] again success is going to be uh [128:17] WordPress is the default platform for [128:20] open an open AI web driven or driven AI [128:23] web. I think that's my north star. Um [128:27] and also I think that we're getting back [128:29] to a good point. I think that we've [128:30] shown based on the 2026 goals that we [128:32] even set out to achieve. I think that [128:34] we're we're tracking very well, right? [128:36] >> I think we're back to three releases um [128:38] a year. I think we need to standardize [128:40] that more. I think the 7.0 0 and how how [128:42] we're navigating 7.0 and that release [128:44] has has been um has brought to light [128:47] really good good conversations to have [128:50] and um transparent visibility into [128:52] things that we need to be doing. So I [128:54] think continuing down that path. I think [128:56] getting and reactivating Meetup is very [128:58] important and um and I think we're well [129:00] on our way with events. So I do think [129:03] that those are the things that are going [129:04] to shape and change WordPress not from a [129:06] community just not just from community [129:08] but also at the product level and we [129:10] need to nail we need to nail the AI [129:13] pieces we need to nail these releases [129:15] down um and uh within the next 12 12 to [129:18] 18 months I think that would make us [129:20] very or position us very well. [129:24] So um looking out further out you know [129:26] so what do we need to do now in order to [129:29] make sure WordPress continues to lead [129:31] the open web space um you know after the [129:34] next 10 years 20 years so what do you [129:36] think we should be doing now to make [129:37] sure that happens [129:39] >> so I do think that education is one of [129:42] our growth levers um and the reason [129:45] being is because when we think about we [129:46] need to level up our we need to level up [129:48] and educate people on current technology [129:51] um and prepare them for whatever ever [129:54] whatever society chooses the path for AI [129:57] will be [129:58] >> and I think that that's a really [129:59] important conversation to have because [130:02] uh we nobody knows the answer of where [130:04] this will eventually land people are [130:05] making a lot of guesses and there's [130:07] going to be a lot of products that rise [130:08] and fall in between those like in the in [130:11] between this gap [130:12] >> but like when I was at Tik Tok um and I [130:15] was I was responsible for Tik Tok shop [130:17] so I don't know again going back to [130:19] e-commerce right [130:20] >> so my my job was to roll out Tik Tok [130:22] shop in in Europe and the United States. [130:25] And I often talked about Tik Tok's [130:28] dominance and why it's so important to [130:31] actually educate society and then take [130:33] an opinionated stance to push society in [130:36] that direction. And what I've noticed [130:38] like WordPress has dominance, right? But [130:41] we don't we often don't take our seat at [130:43] the table and we should we should take [130:45] our seat at the table. We should be [130:46] opinionated. We should start pushing [130:48] people down a very specific path. And so [130:50] I do think that for for the [130:54] conversations around AI and open source [130:56] and why that matters is not happening [130:58] enough, right? [130:59] >> Um and so [131:01] >> I do out I do go out and speak a lot [131:03] around why WordPress is so important [131:05] from a policy perspective, from a [131:07] literacy perspective of why it matters [131:10] >> to be an to be open source and why [131:12] WordPress is taking the stance for AI in [131:14] this way. And this is why it ties back [131:17] into education because if we can open [131:20] the door through the AI microcredential [131:22] uh through literacy, so ethical AI and [131:26] use WordPress uh as the playground [131:28] within that space. Not only are we [131:31] teaching people how to work within [131:32] WordPress or build WordPress sites and [131:34] understand that, but we're teaching [131:36] people the ethics of AI, why it is [131:38] important, why open source is important. [131:41] >> Um and we can continue down that path. [131:43] And then we've introduced an entire new [131:45] generation for applicable WordPress [131:48] rather than a software that builds in a [131:51] CMS. Now it's actually no I learned AI [131:54] by building WordPress. I ended up in a [131:57] microcredential. I turned and used this [131:59] for credits within school. I now have [132:01] started a student club. I now present [132:03] WordPress campus connect. And I am now [132:06] I've started an agency and within this [132:08] agency I'm working through these things. [132:10] So by tapping into this type of network, [132:14] you end up creating another round of [132:16] community while not leaving the current [132:18] community behind. Instead, we're [132:20] welcoming people in and we're trying to [132:22] create a flywheel around new technology [132:26] within WordPress within new new a new [132:28] age, new eyes and education um to grow [132:32] both product and ecosystem because again [132:36] there there will be a lot in the [132:37] ecosystem that comes. [132:39] >> Sure. Sure. I hope that made sense as I [132:41] >> Yeah. Yeah. Yeah. Yeah. Follow through. [132:43] >> That does does. That's great. So that's [132:46] all the questions I had really here. So [132:49] we're going to open it up. We have some [132:50] time for questions. So if there's [132:52] anybody who has any questions, if you [132:54] could please come to the mics here in [132:56] the aisle, please. [132:58] If there are any questions, [133:01] you can just feel free to walk up to the [133:03] mic, please. [133:04] >> Sure. [133:05] >> You want to just introduce yourself [133:06] first? Just [133:07] >> Definitely. Okay, I'll go ahead and [133:08] introduce myself. Uh my name is Gan [133:11] Prakas. I'm into this tech world from [133:14] last 21 years as well. Working as a [133:16] developer by heart. Started my career as [133:18] a content writer. Went ahead picked [133:20] WordPress in 2011 I guess and then I [133:24] started rolling out with the same as [133:26] well. I do have a a very generic and [133:29] very standard question to all of you all [133:31] of us from on behalf of all of us [133:33] sitting here as well. So when it comes [133:36] to representing ourself as a developer, [133:38] we have a lot of USP which we can [133:41] present uh when it comes to asking [133:44] somebody to choose WordPress as one of [133:47] their CMS for their business itself as [133:49] well. But at some point of time there [133:52] are a lot of CMS currently available in [133:54] the market where people are very much [133:57] comfortable with as well for their [134:00] business for any of their uh e-commerce [134:03] business or having a normal service [134:04] based website itself as well. So I [134:07] wanted to understand let's not think [134:09] about us being a part of WordPress. [134:13] What would be your take to convince me [134:16] or anyone sitting as as one of one of my [134:20] client itself as well to ask them to opt [134:24] and choose WordPress as one of their CMS [134:27] when it comes to having a very common [134:30] very generic requirement as well. [134:33] >> So uh are you looking for help on how to [134:36] make a case for WordPress to your [134:37] clients? Is that [134:38] >> No, no, no. It's it's not about the [134:39] case. I I just wanted to understand this [134:41] from Mary itself as well. So being uh [134:44] somebody who is a part of WordPress, how [134:47] do you like [134:49] explain the USPS [134:51] of choosing WordPress as a CMS [134:55] >> for over the others [134:56] >> for the others as well? [134:58] >> So my default and the way I think about [135:01] it is always like I want to keep my [135:04] data. I want to know who's using my data [135:07] and I want mine to always be mine. And [135:11] what I've seen shift over the last [135:14] couple especially couple of months but [135:16] over the last year is the idea that [135:20] it's what is going to happen to closed [135:23] platforms in the future. So not only is [135:26] WordPress extremely safe and secure, but [135:28] you have the ability to move your site [135:30] and you're never locked into something [135:32] and then the expandability of the [135:34] product itself is endless. [135:37] But I think that the conversation around [135:40] closed platforms today with AI is [135:43] extremely important for your sales pitch [135:46] simply because people don't understand [135:47] it and they don't understand what's [135:50] happening like if you think about the [135:51] laws and the policies that are happening [135:52] throughout Europe or what's happening or [135:54] attempting to be happening in in the [135:56] United States. It's that [135:58] what what are you doing with my info and [136:00] what will happen to this site and where [136:02] will this site go and how much more [136:04] important will it be? Especially if you [136:06] think about commerce. So with commerce [136:09] or even blogs, if you're making your [136:11] money off of your site and it used to be [136:13] or historically about traffic and who's [136:15] coming and how you're going to get ads [136:17] and how that's going to serve up or for [136:19] commerce itself, it's you need you need [136:21] to have a website even on a closed [136:23] platform, you still should have a [136:24] website to validate who you are. You [136:26] don't know the future of what your [136:28] monetization strategy is going to be and [136:30] it's not in your hands. And I think [136:33] that's a very business owners are now [136:35] contemplating that or at least the more [136:37] advanced ones are because it's not in [136:39] your hands and you don't know where [136:42] anyone is going to end with AI. What you [136:44] do know with WordPress and these open- [136:46] source models is that it will always be [136:48] in your hands for now. And as search and [136:51] how people discover shifts, you can [136:55] actually be more agile and make these [136:57] shifts on behalf of your client and [136:59] customer to ensure that you're always on [137:00] the forefront and opening that up [137:03] because what anybody here over 40, [137:05] they're defense in here. Bad audience [137:07] for that one. But the majority of people [137:09] over 40 just go to chat now. They're [137:11] using chat as Google, right? And so like [137:13] if you think about that, the people who [137:15] are gen um Gen Z, they're not even on [137:20] the web, right? They're in these closed [137:21] platforms and they're they're finding [137:23] and discovering through Tik Tok or [137:24] through Snap. And so you cannot [137:28] accurately make those calls today. But [137:30] what you can is you can develop for the [137:32] future and you can attempt to [137:34] futureproof and then make sure and [137:36] remind people that wherever you go with [137:38] WordPress, your data is your data and [137:40] you have the freedom and the ability to [137:42] move. Um, and that's pretty profound [137:46] because Shopify is going to start [137:47] selling your information. If you feed [137:49] into AI, what does that mean before they [137:52] create their own individual platform? [137:53] That's a giant marketplace, which is [137:55] probably where they're headed. But you [137:57] know if you start thinking about what [137:59] that means [138:01] >> absolutely so data security is one of [138:03] the major factor that you have already [138:05] identified but what about the recurring [138:07] maintenance when it comes to opting for [138:09] a WordPress website so there's lot of [138:11] things which comes into the play as well [138:14] uh we talked about the open system and [138:15] the closed system that's that's [138:16] absolutely you were right on it with [138:18] that as well and as a developer as a [138:20] person I definitely going to second you [138:22] on that but when it comes to managing [138:24] those WordPress websites ites [138:27] >> like I I I have a e-commerce store where [138:29] I have like only 20 products available [138:32] on my website but there's a recurring [138:34] maintenance that I have to have [138:36] >> I have to have a hosting I have to have [138:37] a proper hosting as well regular backups [138:40] SSL along with that I have to take care [138:42] of the themes plugins what are things [138:44] that I have used across while building [138:46] the website itself as well [138:49] >> yes [138:50] >> yeah so why we should be like choosing [138:53] WordPress in this case not any of other [138:56] CMS itself as well. [138:58] >> Well, again, I think that it's the the [139:01] it's the the price you pay. Would you [139:03] rather have a closed ecosystem that can [139:05] control your theme and then that theme [139:06] is taken away? Would you like to build [139:09] within WordPress and continue to use [139:10] patterns or or however that is the the [139:12] ongoing maintenance cost? It could be a [139:14] balance behind it. How would you sell it [139:17] or how are you selling it today? [139:19] So absolutely uh there's few things [139:22] which which we take care of it as well [139:26] since if you ask me selling a WordPress [139:28] or suggesting WordPress has been one of [139:30] the core as well because we are a [139:32] WordPress developer. We prefer WordPress [139:35] as a very first preference when it comes [139:37] to suggesting somebody to go for a [139:39] website. [139:40] >> Okay. Because we have been working with [139:42] this community since long as well. So [139:45] being into that comfort zone, we [139:47] definitely put WordPress at the [139:49] forefront when it comes to somebody [139:50] coming up and going out with a [139:51] suggestion as well. Got a very good open [139:53] source like community as well, a lot of [139:55] free plugins as well which you're not [139:57] getting across any of the other CMS [139:59] itself as well. But when it comes to [140:01] choosing these small scale things as [140:03] well where there's a very small scale [140:06] application out there or there's very [140:09] small website with few pages as well. We [140:11] definitely recommend WordPress itself as [140:13] well because and for that we have to [140:16] have very uh streamlined flow which we [140:19] call it as security things as well to be [140:21] there at place uh very less efficient [140:24] and effective hosting itself as well [140:25] which should be secure enough. there's a [140:27] lot of service provider available itself [140:29] as well. [140:30] >> Uh these are the thing that we that we [140:32] are doing that I'm doing generally as [140:34] well but I was look just looking for [140:36] something pretty exceptional which I'm [140:38] not doing it which anyone of us are not [140:40] doing it here as well. So that was [140:42] something I was looking for you. [140:44] >> Yeah. Great. [140:45] >> Thanks. [140:45] >> Thank you. [140:47] >> Yeah. Hi we take a question please. [140:49] >> Yeah. Hi uh Karthik here. I'm from C3 [140:52] digits. So I have a question that uh how [140:56] scalable can we think of vcommerce? I [140:59] mean we have a client who love WordPress [141:02] uh uh because of feature and they have a [141:04] lot of content on their website but they [141:07] have around 1 million SKs which are [141:09] generating dynamically. So they want to [141:12] use vcommerce and we just want to [141:14] provide them a solution that what should [141:16] be it can be a jamstack of headless CRM [141:19] with WordPress or so how scalable we can [141:22] do we can go with WooCommerce can we [141:24] confidently say that yes we can build on [141:26] WooCommerce and it will not you know [141:28] break [141:29] >> I mean I would think so do we have [141:30] anybody specific to WooCommerce I see [141:32] Ryan I mean it take I think I think [141:35] WooCommerce itself is very scalable I [141:37] mean could I confirm that the the like I [141:40] don't know if you're looking for me to [141:41] answer if it's like stable or continued [141:43] or will it always be supported then my [141:46] answers to those would be things would [141:47] be yes but specific to to I guess that [141:51] development question I'm I'm I'm unsure [141:54] but I'm looking at our dev rail [141:57] >> yeah [141:59] sorry [142:08] >> okay sure yeah I would love to talk [142:10] about that. [142:11] >> Yeah. Yeah. [142:13] >> Okay. Thank you so much. Yeah. [142:15] >> Thank you. [142:17] >> Uh yeah. [142:18] >> Hello Mary. Uh myself uh Milan. So my [142:21] question is uh Cloudfare recently [142:24] launched am right and it has provided [142:27] some credits and something I also [142:31] provided a statement that it is open [142:33] source and something like that and [142:34] what's your I want to know about what's [142:37] your thought thoughts on the emash. So [142:40] that's my question. [142:41] >> So we were actually presented mdash a [142:44] couple of weeks before it was launched. [142:46] Um I I think that it's a painted [142:48] WordPress. I don't think that that um I [142:51] mean the the UI and things have shifted. [142:53] Uh I don't think it's as powerful of a [142:55] product as they're marketing it to be. I [142:57] mean in general um in general it's it [143:02] you have Astro and you have a lot of [143:04] what the AI basics are. Um I'm not too [143:07] concerned. I think that we've had a [143:09] pretty strong response and WordPress has [143:11] had a strong response and there's been a [143:12] great developer blogs that talk about [143:15] some of the features that are missing or [143:16] even the direction that WordPress is [143:18] headed and what we're doing to ensure [143:19] that there's a structural and [143:21] foundational AI uh rather than something [143:25] that is you know an agent that's pushing [143:28] off against others just like VCEL and [143:31] other ones. So, um, but I do think it's [143:33] a part of this group and how we're using [143:35] it like with our my dot my WordPress [143:39] >> WordPress.net. [143:41] >> Um, how we're working and using AI [143:44] within WordPress. I think it's up to our [143:46] group and our community to start really [143:49] shaping what our versions of those [143:51] things can be. And we should and we [143:53] should be very vocal about that as well. [143:56] >> Thank you. Thanks. [143:59] >> Yeah. you just uh you want to come up to [144:01] the mic while we'll take his question [144:03] please. [144:05] >> Hello [144:06] uh hi my name is Jaip and I have been [144:09] playing with WordPress since last 10 [144:11] years when I started with WordPress when [144:13] I was 16 years old. [144:14] >> So basically uh I I I discovered that I [144:19] love building things and WordPress was [144:21] the first tool that I discovered that I [144:23] can create different kind of things like [144:25] one day I just watch the social network [144:26] and I wanted to create the next [144:28] Facebook. So I tried installing body [144:30] press and doing thing and I always love [144:32] that thing. So basically I have studied [144:35] mechanical engineering and uh I worked [144:37] in that field also but uh throughout the [144:40] all the time I was uh playing with [144:42] WordPress and web development and [144:43] discovered that I love doing this thing. [144:45] I love building things and because of [144:48] WordPress I managed to get into the [144:50] professional WordPress developer. So one [144:53] of the thing that uh someone who has [144:56] love building things uh and then get [144:58] into the thing that I don't like and I I [145:00] I was successfully managed to get into [145:02] the thing that I like. So I'm really [145:05] grateful to every WordPress contributor [145:07] that contributed to WordPress because of [145:10] all these people today I'm doing the [145:12] thing that I love doing that I can do [145:14] without getting a single penny but still [145:17] I'm getting uh bugs. So uh okay that was [145:21] the my introduction. [145:22] >> That's awesome. [145:23] >> Yeah. [145:23] >> Yeah. That's great. [145:26] >> Shout out to all the world contributors. [145:28] Yeah. [145:29] >> Okay. So my question is whether one [145:31] thing that I have observed that so [145:33] lately I have been working with the [145:34] Gutenberg and I have observed that [145:36] Gutenber is really powerful and the [145:38] entire fullite editing concept that [145:40] WordPress is provide uh promoting is [145:42] really great like recently I built a [145:44] theme with patterns and custom blocks [145:48] and the entire editor is so much [145:50] powerful but one of the thing that I [145:53] observed that one is that not so many [145:56] people are aware like basically those [145:58] who are you. So basically what they are [146:00] doing is like uh they are just [146:03] installing WordPress and going for any [146:04] page builders. And second thing that if [146:07] I'm searching the Gutenberg in the [146:09] WordPress so this there is a rating [146:11] system that every uh like even play [146:14] store or or this have that if someone is [146:17] rated low and if you are promoting the [146:20] newer version the rating is getting [146:22] average rating. So basically what [146:24] happened that I'm not sure but when the [146:26] classic editor to the Gutenberg editor [146:28] switch happened uh not so many people [146:30] liked it so maybe many people rated it [146:32] led but over the time the gutterbang has [146:34] so much become powerful so like one of [146:38] the thing that I feel that uh if there [146:41] was a way to promote that uh people to [146:44] try the fullsight editing concept [146:47] Gutenberg patterns and all and how we as [146:50] a developer can help promote promote [146:53] because uh yesterday I observed that so [146:55] many people are contributing with a core [146:57] editor and all that thing. So how can we [147:01] promote this more? [147:03] So I think the first thing is to to join [147:06] those teams right and bring those [147:08] conversations actually into make and and [147:10] start to develop that out um and own it [147:13] and and see what you can do. I think [147:14] there's a lot of people here like you [147:15] said there's people here that these this [147:17] is where these ideas can flourish and [147:18] you can push to see uh especially if [147:21] you're this excited uh I think that [147:23] passion will carry through and you can [147:25] drive some of these initiatives back and [147:27] there's there's key people I see Devril [147:29] right here I think there are a couple of [147:30] core committers in the back um but [147:32] there's everyone is who is here can push [147:35] um and you can start making those a [147:37] reality I I would always start within [147:39] the make community um and I think I keep [147:43] I keep coming back to this because I do [147:44] think it's important and I said it this [147:46] morning. I said we're in a shift within [147:48] WordPress in general and we've been in [147:50] this shift for the last year, but we we [147:53] really you have the everybody here has [147:55] the autonomy and the agency to go into [147:58] WordPress and start driving these things [148:00] forward. Um, and we should just start [148:02] doing things and see what sticks. And so [148:05] with somebody with your amount of [148:07] excitement and passion, I think that [148:08] you'll be able to drive that. I also [148:10] think you're because you started at 16 [148:12] and because of this excitement, you [148:14] should think about doing these maybe in [148:16] a project within WordPress campus [148:18] connect and start hosting some of these [148:20] things to bring up this these type of um [148:23] conversations because you could feel how [148:25] excited and passionate you are. But make [148:27] those connections. Jonath Jonathan Dasi [148:29] is here. He's in the back. Um great [148:31] great connection for you to start [148:33] talking about ideas like that. Peter [148:34] Wilson was here as well somewhere. But [148:37] yes and get involved in some of the [148:39] releases. [148:40] >> Okay. [148:41] >> Yeah. Thank you. [148:42] >> Thank you. [148:43] >> So everyone just sitting here I urge you [148:46] uh everyone of you is having WordPress [148:47] og profile. So just try Gutenberg uh and [148:51] try how cool is it is and make sure if [148:53] you like it just rate it. So let us [148:56] improve the uh [148:58] >> the what do you say [149:01] >> the rating [149:02] >> the rating the branding and the [149:04] perspective. and and be and be our own [149:06] advocates, right? Go out. We We all have [149:08] the ability to go out and speak and [149:09] promote it and present it. We should. [149:11] >> We should. [149:12] >> Thank you, Maddie. [149:13] >> Thank you. [149:18] >> Go ahead, please. [149:19] >> We'll just have one questions [149:21] >> and then we'll end the session. [149:23] >> Yeah. Hello everyone. I'm Rashmi. I'm [149:25] freelance accessibility consultant and I [149:28] want to ask uh as a community, as a [149:31] platform, what we can do to promote [149:35] accessibility and to make sure that [149:39] every new launch of accessibility [149:42] plugins or themes or patterns are born [149:46] accessible out of the box. [149:48] >> Yeah, great question. And this comes up [149:50] a lot, right? This I think this is like [149:52] the second or third time. [149:55] Um I think with accessibility it starts [149:57] with us individual. I think the the um [150:01] how we can implement or ensure that [150:03] thing is maybe infiltration or even [150:05] training around any other any of the [150:07] other teams. Um I think when we [150:09] discussed this last year it was there [150:11] was the idea to create or an overarching [150:14] program team that can go in and actually [150:16] use this as a bulletointed list of [150:18] things that must be the standard before [150:20] any any release. And I still think [150:22] that's a great place to start. Has [150:24] anything come of that? And we launched [150:26] the program team I think end of last [150:28] year. Have you been able to connect or [150:29] has accessibility team been able to [150:31] connect with the program team at all? [150:33] >> Not yet. [150:34] >> I think that would be a great place to [150:35] start. And also like when we think about [150:37] the templates, I think the template [150:39] templating of the handbooks of what is [150:41] the key pieces that your expertise is [150:43] telling different different teams that [150:45] they should be following would be a [150:47] great place. um Tammy has been doing and [150:50] running the program program team [150:52] relatively loosely but I think maybe [150:54] connect with her and see how we can kind [150:56] of push this out in a more robust way. [150:59] >> Yeah. So yeah that's why I want to make [151:01] sure that is there any process that uh [151:04] every means developer or designer can go [151:07] through so that they don't skip that [151:09] process of following the accessibility [151:11] rules at the level basic level. Well, I [151:15] think there's two things. Anytime you [151:16] introduce a new process, it's easier to [151:18] start with new folks rather than old [151:19] like people who have been here much [151:21] longer. Um, so my first thinking just [151:24] from a product level would be to get [151:26] started with some of these contributor [151:28] pathways so it's integrated into anybody [151:30] who comes in new and then it can start [151:32] to infiltrate into the rest of the of [151:34] the community. And then the second thing [151:36] would be to connect with the program [151:37] team um simply because the importance of [151:40] accessibility and and making sure that [151:43] it's something that is considered [151:44] especially before something is released [151:46] are probably really good places to [151:48] start. [151:48] >> Yeah. [151:50] >> Thank you. Thank you. [151:51] >> So we have time for one more question. [151:53] Do we we can take [151:55] >> I think we need to end this session. [151:57] >> Oh yeah. Okay. [151:58] >> Sorry. [151:58] >> I could I connect with you too after. No [152:01] problem. [152:02] >> So all right. Thank you so much Mary and [152:04] Sila. Let's give one more round of [152:07] applause for them. [152:07] >> Thank you. [152:08] >> Thanks. Thank you. [152:10] >> Thank you. [152:11] >> And on behalf of WKAM team, we would [152:14] like to appreciate you with this thank [152:16] you gift. [152:47] Thank you once again [152:50] ladies and gentlemen. I have few [152:52] announcements to make. We are having our [152:55] family photo to the for lotus foyer. So [152:58] I would like to request everyone to be [153:00] there. If you need any help to find the [153:02] place there are volunteers you can ask [153:04] for the places. So I would request to be [153:07] in that group photo because it is one of [153:09] the most memorable moment of every word [153:11] camp Asia and our lunch will be opening [153:15] from 11:30 to 2:30. [153:18] So please have your food, preserve your [153:21] energy and connect with the fellow [153:22] attendees. And the last one is we have [153:26] our workshops running on the second [153:28] floor room number 205 and 206. And I [153:32] would like to request everybody to uh [153:34] please leave this room as we have to [153:36] prepare for our next session. Please [153:39] look after your belonging. [153:44] >> Jasmine foyer. So we are having our [153:47] family photo at Jasmine Foyer. So I [153:50] would like to request everybody to be [153:52] there. [153:56] Thank you everyone. [155:06] Heat. Heat. [155:39] Hey, [155:47] hey, hey. [156:06] Heat. Heat. [157:22] You [157:26] want you [157:29] back. [158:29] Heat. Heat. [216:03] one. Hey, check. One. 1 2 3 4 5 6. H [216:09] check. EP cable. [232:25] Hey, hey, hey. [233:20] Hey. Hey. Hey. [233:48] Heat. Heat. [234:25] feel. [234:32] Feel feel you. [234:37] Heat. Heat. [236:51] Heat. Heat. [238:31] Heat. Hey, heat. Hey, heat. [239:12] Namaste. The program will begin in 10 [239:14] minutes. Please silence your mobile [239:17] phones. [242:55] The program will begin in 5 minutes. [242:58] Please take your seat and silence your [243:00] mobile phones. [243:05] Heat. Heat. [243:28] Heat. Heat. [244:15] Heat. [244:28] Heat. [244:37] Heat. Heat. [245:17] Hey. [245:25] Hey. Hey. [247:09] Heat. Heat. [247:51] The program is about to begin. Please [247:53] take your seat and silence your mobile [247:55] phones. [248:21] Hello everyone. Um I'm Tess from [248:23] Automatic and I'm really um excited to [248:26] be your MC in Lotus 2 today for this [248:29] session. So thank you very much for [248:30] being here. Um just a little bit of a [248:32] housekeeping note is that closed [248:34] captioning is available on your personal [248:36] device for all of the sessions. So if [248:38] you scan the QR code on the screen and [248:41] you follow the prompts, then you'll be [248:42] able to get captions. [248:44] Um I'm really excited to introduce Ryan [248:47] today. Ryan's a developer advocate at [248:50] Automatic. Um he's a seasoned WordPress [248:52] developer and a regular contributor to [248:54] WordPress and to the Gutenberg project. [248:56] And today he's going to talk about the [248:58] interactivity API. So please welcome [249:00] Ryan Welchure. [249:13] Hello everybody. Thanks Tess for that [249:15] awesome intro. Um my name is Ryan [249:18] Welchure and uh I am from um Automatic. [249:22] I'm a developer advocate there. Uh this [249:24] is my first time in India in Mumbai and [249:26] it's been amazing. I have had to learn [249:28] how to cross the road again. Uh it is [249:31] terrifying. I feel like I'm playing [249:32] Frogger if you know that game. Anyways, [249:34] um but yeah, so today we're going to be [249:35] talking about the interactivity API. Um [249:38] and so [249:40] yeah, so the first thing that we should [249:42] talk about is what is the interactivity [249:43] API? That's kind of the whole point of [249:44] me being here today. Um, so the first [249:47] point I'd like to make is that it's a [249:48] standardized way to add front-end [249:51] interactivity to your blocks. Um, that's [249:53] a really generic statement, but um, [249:55] standardiz is very important concept, [249:56] and we'll get into that a bit uh, later. [249:59] Um, its intent is to replace ad hoc [250:01] approaches. So, sprinkles of jQuery, uh, [250:04] maybe little React islands, other [250:06] JavaScript related applications that you [250:08] sort of embed on your page. Uh, this is [250:10] meant to replace those with sort of a [250:11] one unified pattern that ships with [250:13] WordPress core. [250:15] It can scale from the simplest button [250:18] all the way up to live faceted search to [250:21] navigation or uh sorry instant page [250:24] navigation and even shopping carts. Um [250:26] so you can basically do anything you [250:28] want with it which is really really [250:29] cool. Um blocks can share state and [250:32] communicate which means that you can [250:34] have two completely separate blocks on [250:36] different parts of your page that know [250:37] about each other and and can work with [250:39] each other. Uh so so the example I have [250:41] on the screen here is an add to cart [250:43] button that then talks to a shopping [250:44] cart block. [250:47] Um it's already being used in core. So [250:49] we we we we already have this uh [250:51] functionality in place. It powers blocks [250:53] like the search block, the query loop [250:55] block and the other ones that I've named [250:56] on here. The image light box one was [250:58] actually the the first one which is [250:59] really cool. And now I think the gallery [251:01] block is going to be doing that as well. [251:04] Um so under the hood we have kind of [251:07] three technology. I mean there's a lot [251:09] going on but I'm breaking it down into [251:11] sort of three different sections. We [251:12] have Pact um which is a very tiny [251:15] JavaScript library. Now a lot of [251:17] research and a lot of discussion went [251:18] into what should core use um to power [251:21] this and Pact was sort of the winner. [251:23] It's a tiny library. I think I think it [251:25] shipped 7 kilobytes. Um so it's so [251:28] small. Uh so it's used uh along with [251:30] pact signals to do the hydration the [251:33] client logic and the client side uh [251:35] navigation. So the second part is HTML [251:38] directives. Um if you're familiar with [251:40] Alpine.js and some other frameworks, um [251:42] you can uh these this would not be a new [251:45] concept for you, but the benefit of a [251:46] directive is that both the client and [251:49] the server understand what what that is [251:51] and they can work together. And finally, [251:53] the last piece is the serverside logic. [251:55] And this is handled by WordPress uh PHP, [251:57] specifically the HTML tag processor. [252:00] We're going to get into a little bit [252:01] more uh detail on on how this works, but [252:03] those are the sort of the three main [252:04] elements that kind of make up uh the [252:06] interactivity API. [252:08] So earlier I was talking about the [252:09] standardized approach. The benefits of a [252:12] standardized approach are that blocks [252:14] can communicate cleanly because all [252:17] these blocks work with the same system. [252:19] So you don't have to weird build these [252:20] weird bridges between like a React app [252:22] over here, some jQuery over here, all [252:24] that sort of stuff. It just works. Um [252:26] because they all work together. You can [252:28] mix and and match them. You can nest [252:31] them. You can do some very complicated [252:32] things because they all work together. [252:35] One shared system means left less [252:37] JavaScript is actually shipped to the [252:39] browser. So show of hands. How many [252:41] people here have built a theme or a [252:42] plugin that added a JavaScript library [252:44] for the front end? Okay. Now how so so [252:48] keep those up. Show of the other hand. [252:50] How many people have added a plugin or [252:52] something else that added another [252:53] JavaScript library? Okay. Okay, now show [252:55] the legs. No, I'm just kidding. So the [252:58] thing is every time you add a new [253:00] library that needs to be shipped to the [253:01] browser and you don't know is it coming [253:03] from a CDN, is it the minified version [253:05] or if it's is it 6 gigabytes? Who knows, [253:07] right? So the idea is that with this one [253:10] framework, we're shipping one tiny piece [253:12] of JavaScript to the browser and you can [253:14] do all these same uh functionality. So [253:16] that is a huge win. Um, and so when [253:20] every block uses a standard, features [253:22] like client side navigation become [253:24] actually possible uh because they all [253:26] work together. So we can do really [253:27] really cool things. We can detect all [253:28] the blocks on the page and know whether [253:31] or not they're using the interactivity [253:32] API and then enable full page navigation [253:35] in instant uh navigation. Um, so there's [253:39] there's some core concepts that I'm I'm [253:40] going to talk. Um, the the first is the [253:44] the reactive and declarative mindset. [253:46] This is sort of a a way of thinking [253:48] around how we build these these [253:49] applications and the the benefits that [253:52] these things provide. Um there's a [253:54] difference between the data. So we have [253:56] global state, local context, derived [253:58] state and some config. So all different [254:00] types of data, but you know when you [254:02] should use them and how how they work. [254:04] Um well we're going to talk about [254:06] serverside rendering and WordPress [254:07] specifically. And then finally some [254:09] client side navigation. So let's start [254:12] with the reactive and declarative [254:13] mindset. So declarative means you [254:18] describe what what the UI looks like [254:20] based on state not how to update the [254:23] DOM. So when we talk about declarative [254:26] code versus imperative code we we talk [254:28] about how in an imperative code example [254:32] we we work on the how. We're like how [254:35] does this work? So in this very [254:37] contrived example, we have a function [254:39] and its purpose is to receive an array [254:42] and and return a sum of all the even [254:45] numbers in in that array. So in the [254:47] imperative approach, we're doing things [254:48] like we are micromanaging every aspect [254:51] of this. So we've defined a variable. [254:53] We're doing a for loop. We're iterating [254:54] each instant, each item. We're checking [254:56] is that item odd or even. Then we're [254:58] adding to the I don't need to go through [255:00] every line of code, but you get the [255:01] point. So we're doing everything like [255:03] very very specifically. Whereas in a [255:05] declarative example, so this is the same [255:07] function, does the same thing. We don't [255:09] care how it's doing it. We just care [255:11] what it's doing. And so the point here [255:13] is we're we're using the filter [255:15] function. And the only thing we're [255:16] telling the filter function is how [255:17] should we filter? And then that returns [255:20] an array. And then we're using the [255:21] reduce function to take that array of [255:23] numbers and return a single value. So we [255:26] don't care how it's working under the [255:29] hood. Like I I don't care if filter is [255:30] is using a for loop or whatever it's [255:32] doing. It's just working. And so that's [255:34] that's that's the difference. It's it's [255:35] it's how versus what. [255:39] Um does that sort of make sense to [255:40] everybody? Okay, cool. Um so the next [255:42] part of this is reactivity. So [255:44] reactivity means and not ReactJS [255:46] reactivity just u when the state changes [255:49] the UI automatically updates. So you [255:51] don't have to do any man any manual DOM [255:53] manipulation. That's super important [255:55] because as soon as you get to the the [255:56] the place where you're um having to [255:59] manually add classes and remove classes, [256:01] you you can get into edge cases and find [256:03] weird bugs, it's very very it it can get [256:06] very complicated in very very big apps. [256:08] Um but with the concept of reactivity, [256:11] all we're doing is defining a [256:12] relationship between a directive which [256:15] is and a piece of state and that that [256:18] relationship is never broken. You can [256:20] add as many relationships or as few and [256:23] they they just take care of themselves. [256:24] There's no chance of you uh getting into [256:27] a weird state because your if statement [256:28] didn't have an else that it needed [256:30] somewhere along the lines and so you [256:31] haven't added a class, right? And then [256:34] uh finally, this is kind of a in the [256:36] weeds one, but I wanted to mention it. [256:38] Um so the interactivity API allows for [256:42] mu for mutability. So a lot of sort of [256:45] state management. So like like Redux for [256:47] example um it's this concept of you can [256:50] never modify state directly. You always [256:52] have to create new state and then return [256:54] that state object into your store. Um [256:57] and that is is is all well and good but [257:00] it can lead to a lot of boilerplate. [257:01] It's the code's a lot more more [257:03] complicated than maybe it needed to be [257:05] but with the interactivity API we can [257:07] have a mutability and still have the [257:11] reactivity work. So um it's it's kind of [257:14] a powerful thing. [257:16] So this is in this section. This is the [257:18] takeaway that I want you to have is that [257:20] when you think of um I want you to think [257:22] in terms of the state updates the UI and [257:24] not the other way around. So once you've [257:25] made those relationships, you don't have [257:27] to worry about what the UI is doing [257:28] because the reactivity will handle it [257:30] for you. Okay. So let's talk about the [257:33] four types of data that you can kind of [257:35] get in the um interactivity API. We have [257:39] uh global state. So this is a shared [257:41] source of truth for any interactive [257:44] block and it's kind of how separate UIs [257:47] can can stay aligned. I'll I have an [257:49] example that I'll talk about in a second [257:50] but we'll just you know so this is [257:52] global it's global to every block sorry [257:54] it's global to the block and available [257:56] to any other block. We have local [257:58] context [258:01] this is state that lives on a specific [258:03] element and its children. So this is [258:05] really good when you want to have an [258:06] instance of a block handle its own data [258:09] that's separate from all other [258:11] instances. [258:12] We have derived state and this is kind [258:14] of a kind of a it's a cool concept in in [258:18] in the way that like if you have a state [258:20] item and you need to create some other [258:22] kind of state you can derive from the [258:24] existing state. I've just used state [258:26] like five times in the sentence so I [258:27] don't know if that was clear or not but [258:29] the idea of generating state based on [258:31] other items. Okay. And then finally, we [258:34] have some config. So config is static. [258:36] It it's readon data and it doesn't [258:38] trigger UI updates, but it's really good [258:40] if you needed to have like u like a URL [258:43] to a rest endpoint, for example, that [258:46] that that you wanted to hit. So let's [258:49] look at how we would define some global [258:50] states. We use the the WP interactivity [258:53] state function. Takes two parameters. [258:55] The first is our store name. The second [258:57] is an array of key value pairs. Um, and [259:00] this this acts as like your your default [259:03] values for your state. Um, and this is [259:05] defined in PHP. So I should sort of I [259:07] should have mentioned that. Um, so we [259:09] can define some local context by using [259:11] the ridiculously long named function of [259:14] WP interactivity data WP context. And it [259:18] works sort of the same way where you [259:19] pass it an array of data that that you [259:21] want to pass as context. So if we [259:24] remember local context works on the item [259:26] that it's defined on and all of its [259:27] child items will inherit that context. [259:31] Um you can also define it. I don't know [259:33] if there's a point around here but at [259:34] the bottom there wherever my slides are. [259:36] Um there is a way of manually defining [259:39] it where you pass a Java like a JSON [259:42] object into there. I wouldn't do that. [259:45] It's totally possible but when you get [259:46] really complicated data that can fall [259:48] apart pretty easily and that's why that [259:49] helper exists that function. [259:52] Um so here's a here's an example of of [259:55] defining some a derived state. So on the [259:58] left side we have um a store with a [260:02] piece of state called text. It's a [260:04] lowercase hello world. And then what we [260:06] want to do is we want to show it in [260:08] uppercase. So instead of storing another [260:12] state item state.uppercase text, we're [260:15] going to derive this new piece of data [260:17] from our existing store. And so we do [260:19] that with um and the the store part will [260:22] become clearer in a second, but [260:24] basically what we're doing is we're [260:25] creating what's called a getter [260:26] function, the JavaScript getter in our [260:27] state object there called UC text. And [260:30] we're returning state.ext to uppercase, [260:32] which is just a JavaScript function that [260:33] makes everything uppercase. So we don't [260:35] need to store that. And the benefit of [260:37] not storing this value is that if we [260:40] were to add say an input field where you [260:42] could change that text, the other the [260:45] derived state will automatically get [260:47] those changes. So you don't have to [260:49] update things in two places. This is [260:50] where the the those relationships and [260:52] the reactivity really really helps. [260:55] And finally we have uh config items. So [260:58] we use a function called wp [260:59] interactivity config. It's in the [261:01] render.php. [261:03] Again we pass the store name and then [261:05] some items that we want to have [261:07] available to our application from the [261:09] server. Please don't put your API keys [261:10] in here. Not a safe thing to do. Um and [261:14] this is how we would access it. So [261:15] there's a there's a a function in the [261:17] package called get config and you would [261:19] just access those the way you would [261:21] access um anything you can sorry you can [261:23] dstructure the items that have been [261:25] defined and use them there. Okay so [261:29] which when when should you use these? So [261:31] global state is going to be when you [261:32] want multiple blocks to share or sync [261:34] data. So a real world example is a [261:36] shopping cart and a buy now button. So [261:38] if you have a button over here that's a [261:39] block that represents being able to add [261:41] this to your cart and you have a a [261:43] shopping cart here. They need to know [261:45] what's in the cart. So maybe your button [261:47] block says buy, you know, add to cart, [261:50] but if it's already in there, it says [261:51] remove from cart. So it needs to be able [261:52] to talk. So that's a good example of [261:54] like global skate uh global state. For [261:57] local context, the best example that I [262:00] can think of is a gallery slider. You [262:03] have two instances of of of the same [262:05] block. If you were to use global state [262:07] and you advanced uh one of the sliders [262:10] to the next slide, they would both move [262:12] forward. Whereas with local context they [262:15] manage that data independently of each [262:17] other. So that's an important context. [262:20] Um and derive state should be used [262:21] whenever you can calculate a your new [262:23] value from something in your state in [262:26] your value a value that already exists. [262:29] And finally uh config is whenever you [262:31] need something from the server. Again no [262:34] a no code API keys in there please. [262:38] Okay let's talk about server side [262:40] rendering and WordPress. This is this is [262:43] honestly one of the reasons that the [262:45] interactivity is API is so powerful [262:47] because um unlike react view many [262:51] JavaScript frameworks that use Node.js [262:53] for their server side rendering um the [262:55] interactivity API uses PHP and the HTML [262:58] API so it's rendered on the server for [263:00] us. [263:01] Uh because of that the blocks output [263:04] fully rendered HTML um to your browser [263:07] so the browser gets the right markup [263:09] immediately which is great for [263:10] performance and great for SEO um on on [263:13] the client side, the API just hydrates [263:16] that data that you defined and it gives [263:19] you reactive um behavior without [263:22] rerendering the page. How many people [263:24] have had a React application that they [263:26] render on on the page and they have a [263:27] very specific loading state that kind of [263:29] pulses until the API hits and returns [263:32] that data? We don't really have that [263:34] because all our data is already there [263:36] immediately provided by the server. [263:40] And then uh finally, and this is [263:42] honestly the most powerful part because [263:43] we use WordPress to render. We get all [263:46] the all the all the great interactivity [263:48] stuff, but we also get all you know the [263:49] 20 years of WordPress APIs, all the [263:52] hooks, everything that that that we [263:53] need. So we can filter the output of [263:55] this block. We can modify anything with [263:57] this block on the PHP side using all the [264:00] WordPress a APIs and it just works [264:02] really well. [264:04] So let's look at how we would enable a [264:06] block to use the interactivity API. Uh [264:08] the first step is we're going to go into [264:10] our block JSON file. We're going to add [264:12] interactivity true to our supports [264:14] object. Now with the client side ren uh [264:16] sorry with the client side navigation [264:18] there there's another um it this becomes [264:21] an object and you enable both but this [264:23] is a short a shortand to enable all the [264:26] things. Um after after you you've done [264:28] that we're going to add viewcript module [264:30] to our block JSON. So if if you're [264:32] familiar with with with VueScript to to [264:34] load a file on the front end, this is [264:36] slightly different um because the the [264:38] interactivity API uses ES modules. So so [264:41] we have to load this in in a different [264:43] way. Um and so once we we've defined [264:46] that relationship, we have to modify our [264:48] build process a little bit. So if you're [264:49] using the scripts package, which you [264:50] should because it's amazing, um you're [264:53] going to want to add the experimental [264:54] modules flag to your start and to your [264:57] build scripts. What this does is this [264:58] tells Webpack to output this as a ES [265:01] module. [265:03] Um, and then we would want to uh define [265:06] some global state which we've already [265:08] sort of seen. Um, you can define some [265:10] local context if you need to. Um, and [265:13] you know, so we we've talked about these [265:15] all already. And then then we get into [265:17] the client side store. So the client [265:19] side store is in Vue.j uh the the um [265:22] best practice is for it to be in Vue.js. [265:25] And it is basically we we we we import [265:28] this the store function from the [265:29] package. We call store. We pass it a [265:32] name of our store and an object that can [265:35] that has three three properties state, [265:37] actions and callbacks. So let's talk [265:40] about what what the differences are [265:41] between state, actions and and [265:43] callbacks. So state is going to be used [265:45] to store the global state for your block [265:47] as well as the getter functions that [265:50] return your your uh derived state. So if [265:52] you remember the example of the upper [265:54] state of the uppercase text that was [265:56] stored in inside of state. [265:59] Next we're going to um have actions and [266:03] actions are intended to be functions or [266:05] callbacks that are triggered when a user [266:07] interacts with something. Um so like an [266:09] on click an on mouse over an on scroll [266:12] anything a user does you you would put [266:14] the call back for that in in actions. So [266:17] then we have callbacks. Now callbacks is [266:20] very similar to actions except the [266:22] intent here is that it's for things that [266:24] are programmatically happening or [266:25] automatically happening. So there are [266:27] directives that run when the block [266:29] loads. So the user doesn't initiate [266:31] that. It's just sort of a programmatic [266:32] thing. Now there's a lot of gray area [266:35] here and technically you don't even need [266:37] to use these names but this is the best [266:38] practice. I I would recommend sticking [266:40] with state actions and callbacks, but if [266:42] you really really wanted to, you could [266:44] call it programmatic functions instead [266:46] of callbacks, and it would still work as [266:47] long as you accessed it correctly when [266:49] you were uh connecting directives. [266:52] Um the the store is self-referencing. So [266:55] when you define a store, you can return [266:57] the state, you can sorry, you you can [267:00] dstructure the state or the actions or [267:02] the callbacks and use them in other [267:03] parts of your store. So you'll see this [267:05] a lot with state. So in my example here, [267:07] I've got a um an action that that is [267:10] going to update the state. And so what I [267:12] need to do is dstructure state out and [267:14] then access state.ext and modify it [267:17] there. So that's that mutability that we [267:19] talked about where you can just directly [267:21] access things. [267:27] Don't fall. Okay. Um couple more [267:29] functions to talk about. When we define [267:31] local context, we need to be able to get [267:33] access to that context. So there's a [267:34] function called get context which does [267:37] exactly that. Um in the example that [267:39] that you're seeing here, we have a [267:41] button that when the user clicks it, it [267:43] it runs a actions.log function and in [267:46] there we're going to get the context. [267:48] Now if you notice the context is defined [267:50] on the parent. So any child element gets [267:54] access to that context as well. So and [267:56] it's going to we're we're console [267:58] logging is open and in this case it's [268:00] false because that's what the what what [268:02] the data is saying. And then the get [268:03] element function actually gets us access [268:05] to the element that is invoking this [268:07] callback. So in this case um if we were [268:09] to if we were to um console log [268:12] element.attributes [268:13] we would get all the attributes for for [268:15] the button that called it. [268:19] Um so now we've we're kind of set up [268:22] ready to roll. So now we can start uh uh [268:24] defining our interactive relationships [268:27] uh using directives. So the most [268:29] important directive is WP interactive. [268:32] This enables interactivity mode. It [268:35] should go on the root element of of your [268:37] block and without this nothing works. [268:40] So, so it's pretty important. So, you [268:42] you can you can define it in two ways. [268:45] The top example is sort of the more [268:46] recent example. And the value that goes [268:49] in there is the name of the store that [268:51] you've defined in your WP interactivity [268:53] state function and in your your your [268:55] Vue.js. This store name is what connects [268:57] all the dots. So, it has to be the same [268:59] in every file. You can also do it the [269:02] other way which is a bit below. It's [269:04] it's kind of older. Um it works fine but [269:07] you'll see probably more of the top [269:09] example than the bottom example but but [269:10] they're both perfectly valid and work. [269:13] Um so let's talk about directives. So [269:15] this is an example of of a full [269:16] directive. Um the directives got four [269:18] parts. The first part [269:21] is the data dash. Now this is just an [269:24] HTML data attribute if you're familiar [269:26] with that. We need this to make this [269:28] valid HTML code. After that, we have the [269:32] name of the directive. They're always [269:33] going to be prefixed with WP and then [269:36] something. So, WP dash and in this case, [269:38] I'm using bind. The next part is what I [269:42] I like to think of as the parameter. So, [269:45] some of these directives need a bit more [269:47] information. So, for example, the point [269:49] of WP bind is to create a relationship [269:51] with a particular HTML attribute. So, [269:55] we're binding that attribute to a piece [269:57] of state. So, we need to tell it which [269:58] attribute. So in this case, we're going [270:00] to be using the hidden attribute. This [270:02] could be uh Arya label. It could be any [270:05] valid HTML attribute. Um and so that's [270:09] that's why I kind of call it the [270:11] parameter. And and note that it is uh [270:13] the name of the directive and the [270:14] parameter are separated with two dashes. [270:17] And then finally, we have this last one [270:19] which is a use like a a unique [270:21] identifier. And the reason that this [270:23] exists is because if you were to have [270:24] two two directives with with the same [270:27] name, they are technically invalid HTML. [270:30] So if you had a a data-wp- [270:33] um uh init and you had two of those [270:36] because you wanted to do two different [270:37] things, you would need to add that this [270:39] last section to make it valid HTML. [270:43] Okay. And so uh directives directive [270:47] values are sorry I I read that slide and [270:49] realized I I have a typo and it just [270:51] sent my mind in a completely different [270:53] direction. Um the values are connected [270:55] to the store. So in this example we are [270:57] we are making a whole bunch of [270:58] connections. So you can see we've [270:59] defined our our interactive our WP [271:02] interactive with the store name and it [271:04] matches the the store name in in the [271:06] Vue.js. Without that things will not [271:08] work. We have a WP innit that's [271:11] accessing that callback in it. um which [271:14] is in our callbacks um object and then [271:17] then we have a number of other things. [271:19] We have our bind and then we have the [271:21] buttons that are calling. So so once we [271:23] make these connections they start to [271:24] access uh different parts in the store [271:26] and really we've got our our functioning [271:28] application. [271:31] So there's a lot of directives. I'm I'm [271:33] not going to go through them all um [271:35] because that's what the docs are for. [271:37] But um I do I have kind of classified [271:39] them a little bit. Um [271:43] whatever. So we have ones that I'm [271:46] referring to as data providers. So this [271:47] is the interactive in the context. We [271:49] have ones that are attribute based. So [271:50] we know about WP bind. There's class, [271:53] there's style, there's text and things [271:55] like that. So um we have event based. So [271:59] there's there's WP on and WP on async. [272:02] There's always an async version of the [272:04] on version. So we can have on window and [272:06] on document and there's async versions [272:08] of those as well. Um the reason that we [272:10] have async um is because it's basically [272:14] like adding an event listener with the [272:16] passive true set in like so in [272:20] JavaScript if you do add event listener [272:22] there there is an option to pass it as a [272:24] passive one and that that works a lot [272:27] better. I think it doesn't mess with the [272:29] main thread as much in JavaScript. So, [272:31] I've seen uh some examples where you [272:33] have it and you have a like your scroll [272:36] will be deferred until something else [272:38] loads. And so, that's where the async [272:40] comes in to help manage that. Um, we [272:43] have some programmatic stuff watch and [272:46] net and and run. These are all sort of [272:48] these would all all be using callbacks. [272:50] And then we have looping in iteration. [272:52] This is really cool. I don't have time [272:54] to get into the examples, but this [272:55] allows you to build it like nested loops [272:57] and things like that. Sorry, like list [273:00] like [273:01] ordered and unordered lists. Does that [273:02] mean um and stuff like that? Yeah. So, [273:06] finally we are into the client side [273:09] navigation and this stuff's super fun. [273:11] Everything I say I'm like this is [273:12] amazing. Everything about this is [273:13] amazing. So, it's silly. Um so, when we [273:16] talk about navigation, we have sort of [273:18] the traditional way that we click [273:20] through a website. We've been doing this [273:21] now for however long the internet's been [273:23] around. So this is not new to anybody, [273:24] but we click a link, we do a full page [273:27] reload, we we get the white flash, you [273:30] know, all the JavaScript initializes, [273:32] all the stylesheets reparse. You know, [273:35] there's a perceived uh delays. And [273:37] that's sort of like like the way that [273:38] we've always done this with client side [273:41] navigation, especially with the [273:42] interactivity API. We still click click [273:45] a link, but now only the content areas [273:47] swap. So we're able to define sections [273:51] of our pages that can be like we can [273:56] magically look ahead and see what is on [273:58] the new page and just bring that over. [274:00] It's really really powerful. Um our [274:02] interactive state is preserved so we [274:04] don't we don't have start from from [274:06] scratch. Um shared assets uh stay loaded [274:10] and then we have a near instant [274:12] transition which is really cool. So, if [274:13] you've used the query loop block and [274:15] you've enabled the um uh the the [274:19] functionality to not do a page reload [274:20] when when you navigate to the next set [274:23] that that uses this and it's it's it's [274:25] so powerful. So, so how does it work? [274:27] Well, it it works using um this package [274:30] called the the interactivity router. And [274:32] basically what we're doing is behind the [274:34] scenes we're fa we're fetching the HTML [274:36] for for the new page in the background. [274:39] Um and we'll talk about the [274:40] technicalities there in a sec. [274:42] Um, and then when we fetch that page, [274:46] we're looking for a matching ID between [274:48] places on the page. It's technically [274:50] called a router region. And then once we [274:53] have those, we we just bring them over [274:55] and and and swap those sections in. [274:58] And and so everything else is preserved. [275:00] Even so, if there's a section that [275:02] doesn't have a router ID, it doesn't get [275:04] changed. Everything stays the same. And [275:06] the client side states the same as well, [275:08] which is really cool. Um, so the way [275:10] that this is done is is with a directive [275:13] called a WP router region. And so what [275:16] this is is just it's defining an ID that [275:19] the the router uh package can go and [275:22] find and check to see if it's on the [275:23] current page and just bring that in, [275:25] right? Um, so this is kind of a visual [275:28] example of sort of how it works. On the [275:30] left, we've got our initial view. And in [275:31] this case, we we've got four sections. [275:34] And I realize now that like I don't [275:36] think we would ever change the header [275:37] and the footer, but that's the example I [275:39] have. Um, so any of the regions that [275:41] match in the new page get automatically [275:44] updated in the current page that you're [275:46] on. But the important part is the not [275:47] updated stuff stays exactly the same. So [275:50] if you didn't want to update certain [275:51] parts of your page, you could literally [275:53] just just only update the title and the [275:55] featured image or something, right? So [275:57] you you can define that. [276:00] So the way we trigger uh this is we use [276:03] two functions. There is a prefetch [276:06] function and there's a navigate [276:07] function. So the the code in on the left [276:11] hand side in render.php is all it's all [276:13] pretty straightforward. But the but the [276:15] the theory here is that we use prefetch [276:18] to get the page before we navigate. So [276:22] we're adding an an on async mouse enter [276:25] to preload the page. And then when we [276:28] click, we're gonna use the we're gonna [276:30] actually navigate. And so what that does [276:31] is that that that makes it super duper [276:33] fast. So on the JavaScript side, if [276:35] you're not familiar with the syntax, [276:37] what we're doing here is we we're using [276:39] yield and import to dynamically import [276:42] these functions from this package. And [276:44] what that does for us is that it means [276:47] that if someone loads your page and they [276:49] never navigate away, this stuff is not [276:51] loaded. It's only loaded if it's needed. [276:53] And so, you know, it's a bit of a of a [276:55] performance thing. And so that's why [276:56] we're and there's so we're using [276:57] generator functions and yield import. [276:59] The reason that that is done that way as [277:02] opposed to something like async await [277:03] has to do with the way that we keep um [277:05] our our context scope. Uh it's really in [277:08] the weeds, but just just do it this way [277:09] and it'll work. So the idea is that when [277:11] we mouse over a link, we prefetch the [277:13] page and then when we click, all that's [277:16] already been done and it just makes [277:17] makes the change and super easy. Um, so [277:20] this is sort of how the the state [277:23] happens on the client side during [277:24] navigation. We we we use a soft merge [277:27] strategy. So nothing's lost. It's only [277:30] added to really. So like in in in [277:32] example, we have count we have counter [277:34] and username. Those are persisted and [277:36] then any new items that come from from [277:39] the new page are just sort of added to [277:41] the the current state. [277:45] Um, so sometimes when we do that, we're [277:48] going to need to get data from the new [277:50] page because what if the because like [277:53] the post ID may change and that might [277:55] affect some things that you're doing or [277:56] or it might be a different set of [277:57] taxonomies. There could be reasons why [277:59] you would need to get that global state [278:02] from the the new page that you're [278:04] loading. So there's a couple of [278:05] functions that are available to us [278:06] called get server state and get server [278:08] context. So what these do these return [278:10] readon values can't can't change them [278:12] but this allows you to get like the data [278:15] from from the new page and then [278:16] potentially update your state. So in [278:18] this example we we you know just imagine [278:22] with me if you will that we've tied this [278:24] to a watch. So, anytime the navigation [278:26] changes, we are going to grab a bunch of [278:29] stuff from the new page and we're going [278:31] to update the current store just to make [278:33] sure that things work. Like, for [278:34] example, product count or if if certain [278:37] things are expanded. Um, oh, I think I [278:41] have a lot of time left here. Um, so [278:43] what's next in resources? So, there's [278:45] lots of things I couldn't talk about. [278:46] Um, but I have some homework for you and [278:51] I'd like you to take a look at this [278:53] stuff. So, one of the really cool things [278:54] about the interactivity API is that [278:56] TypeScript support is like amazing. It [278:58] is built for for TypeScript. I use [279:01] TypeScript when I build with it. I just [279:03] didn't add it to this example because I [279:04] didn't want it to, you know, I didn't [279:05] want that to be a blocker for [279:08] understanding, but it is very very good. [279:10] Um, the WP directive that we talked [279:13] about that that looping stuff is so cool [279:15] and like highly recommend you go check [279:17] that out. the full page uh navigation [279:20] mode. There is anam there is an [279:21] experiment in the Gutenberg plugin that [279:22] allows you to do full page uh [279:24] navigation. So it's very cool. It's all [279:26] built with the interactivity A API. Um [279:28] get server state server context. This is [279:31] really really good. We have a very [279:32] complicated site. Um you should really [279:34] understand the nuances of this and how [279:36] it works. And then oh and and the recent [279:38] changes in 69 and 7 and 7.0 when that [279:41] comes out which was supposed to be [279:42] yesterday. So the slides [279:44] um anyways so there's like nested router [279:46] regions. There's an attach to thing with [279:48] router regions, which is really cool. [279:49] Which means when you navigate to that [279:51] page, if that region doesn't exist on [279:53] your current page, it's added to the [279:54] page. So you can do things like popups [279:56] or cookie notices, all kinds of cool [279:58] stuff. Um, and then we have the official [280:01] docs. So go and read these docs. They [280:04] are like excruciatingly detailed and I [280:06] mean that with all respect and love [280:08] because they are so well put together [280:10] and there's so many things and you can [280:11] just pass that to code and it'll just [280:13] ingest it and you'll be good to go. So, [280:15] I would recommend reading all of this [280:17] stuff. Um, [280:19] cool. And so, okay. So, and and that's [280:21] it for me. So, um, you can find me on [280:24] Twitch and YouTube. I keep looking at [280:25] the wrong screen. Uh, Ryan Welchure [280:27] Codes. And I would really love it if you [280:29] could, um, give me some feedback on the [280:31] talk if, you know, I love all feedback, [280:33] especially, well, I love good feedback, [280:35] but bad feedback is is well appreciated [280:38] because it makes me a better speaker. [280:39] Um, yeah. Thank you. [280:45] Thank you so much. Thanks so much, Ryan. [280:47] We do have some time for questions. So, [280:48] there's two microphones um up there, one [280:52] in each aisle. If you wanted to go stand [280:53] behind a microphone if you have a [280:55] question for Ryan, [280:59] perfect. [281:02] >> Yep. Just up there to the microphone. [281:04] >> All right. Uh thanks, Ryan. [281:06] >> No problem. [281:06] >> Uh great feature. Um I primarily work in [281:09] Ruby on Rails and um thanks for doing [281:12] this because we have something similar [281:14] called HTML HTML over the wire hot wire. [281:18] Okay. [281:18] >> Which uses stimulus JS very similar. Um [281:23] um so they have also uh added the [281:26] ability to create mobile apps [281:30] uh u using that framework [281:34] >> using the hot over the wire [281:35] >> hot wire HTML over the wire because it [281:38] kind of um makes a traditional website [281:41] into a more app-like experience. [281:44] >> Okay. Uh do you think um I can use the [281:49] interactivity API for a native mobile [281:52] experience? [281:53] >> Um well it is it is HTML, JavaScript and [281:55] CSS at the end of the day. So if you're [281:57] using a framework that can translate [282:00] that into an app. I know like I might be [282:02] dating myself but there was like Cordova [282:04] and stuff like that that used that to [282:06] you know to be turned into a binary for [282:08] phones and whatnot. Um like yes really. [282:10] Um, now the interactivity a API will not [282:14] enable that for you. But it if your [282:17] whatever framework you're using to [282:19] generate this app allows you to have [282:21] JavaScript, HTML and CSS um and and a [282:24] WordPress background, [282:25] >> right? [282:26] >> Or back end, I should say. Um, then yes, [282:28] I mean there's no reason why why it [282:30] wouldn't. You're going to run into to [282:32] some things like the the the powerful [282:33] part of this is that it's not just a [282:35] front-end technology. Um, you do rely on [282:38] the PHP. [282:39] >> Yes. and and and the server. So, I mean, [282:43] I don't see why you couldn't is is is [282:45] the short answer. So, yeah. [282:46] >> Thank you. [282:47] >> Thank you. Appreciate it. [282:48] >> Thank you. I meant to say as well, [282:50] please say your name and where you're [282:52] from before you ask your question [282:53] because we'd love to know who's asking [282:55] as well. There's one over there. [282:57] >> Hello, Ran. [282:58] >> Hello. [282:59] >> Thank you. It was nice listening to you. [283:01] >> Oh, thank you. [283:02] >> Uh my name is Himadri [283:04] and uh I'm a WordPress developer. uh I [283:08] wanted to ask a question that uh can you [283:11] think of a use case where the [283:13] interactive interactivity API cannot be [283:16] or you don't you will not suggest to use [283:18] interactivity API? [283:21] >> Good question. Um short answer is no. Um [283:24] I've seen some really cool things built [283:26] in it. Um [283:29] it [283:31] you're really building an interactive [283:33] experience. So, in a place where an [283:36] interactive experience doesn't make [283:37] sense, this probably doesn't make sense. [283:39] Um, but I don't I mean I can't think of [283:43] any reason why you wouldn't want to use [283:46] >> for example um as uh I've seen I've not [283:51] done this practically ever so I may be [283:53] wrong in this but uh for example some [283:58] pages u or or a website is loading uh [284:04] multiple values with each request. [284:08] >> Okay. because it keeps the old values [284:10] intact on that page and will keep adding [284:14] new values. So it will maybe load the u [284:18] are you talking about like the soft [284:19] merge strategy that we talked about [284:21] where where when you load a new page it [284:22] just adds more to state. Is that what [284:23] you're Yeah. So that's that's where that [284:25] that get server context and get server [284:27] state uh comes in handy because you're [284:29] right like if you just keep adding stuff [284:31] if every page adds new new context or [284:34] sorry new state you will probably at [284:36] some point have to run some cleanup [284:38] >> but that's that's not really a [284:40] limitation of the system and you know if [284:43] if we didn't have the means to get the [284:46] new stuff from the new page then that [284:48] would be a problem but but because we [284:49] have it we can just do our due diligence [284:51] as developers and clean up after [284:52] ourselves as as needed. But yeah, I mean [284:55] I can't see why you there would be any [284:58] reason you wouldn't want to use like [285:00] anything that you can build in [285:00] JavaScript you you can build in this [285:02] because it's JavaScript. It's just [285:03] JavaScript. It's just um it's [285:05] standardizing things. Um it's locked [285:08] into WordPress on the back end. So [285:10] serverside rendering becomes completely [285:12] native. So there's a lot of benefits to [285:13] it. But I mean if if there is an example [285:16] I' I'd love to see it. I'd love like [285:17] challenge accepted. I'll see if I can do [285:19] something with it, you know. So yeah. [285:21] >> Thank you. [285:22] >> No thank you. [285:24] We've got time for another question if [285:26] anybody has one. [285:28] >> Here we go. [285:33] >> Hello. [285:34] >> Hello. [285:35] >> So, thank you for the like workshop. I [285:37] had one doubt. Uh because interactivity [285:40] interactivity API supports mutability, [285:43] right? [285:44] >> So, what happens uh when there are like [285:46] race conditions like there's a [285:47] background thread going in the like [285:50] browser. Yeah. because browser not [285:52] support web workers. [285:53] >> Okay, [285:53] >> so what happens when the background is [285:56] updating the state while the UI also [285:59] like getting the values? [286:01] >> That's a good question. Um I we'd have [286:03] to sort of look at the at the the case [286:06] in point. Um you would probably want to [286:09] structure your code in such a way that [286:10] that you can uh watch for values [286:13] changing. So, for example, there is a [286:15] there's a WP watch, which is if you're [286:17] familiar with React, it's very much like [286:19] a use effect where you and it's very [286:21] it's hilarious. I haven't figured out [286:22] how, but it's like voodoo magic because [286:24] it knows exactly what you want to watch, [286:25] but you don't tell it somehow. So, it's [286:27] crazy. I don't know. But what you could [286:28] do there, you could add a watch that is [286:30] is checking the status of like a request [286:33] or like maybe working with a web worker, [286:35] whatever it is, and then when things [286:37] change, we we update the values. the the [286:40] beauty of the reactivity is that it [286:42] handles a lot of that itself. So there [286:44] might be cases where you have to check [286:46] for use for for race conditions because [286:49] if you're hitting a third party and you [286:50] know all that stuff that comes with [286:52] hitting a an API um I think I'd need to [286:57] know the specifics to be able to kind of [286:58] say well I think you should do it this [287:00] way but I I think it will for the most [287:03] part handle itself because as soon as [287:05] the state is updated the UI changes. [287:07] >> Yeah. Right. [287:07] >> Because I had tested out because React. [287:10] >> Can you speak to the microphone a little [287:11] bit? [287:11] >> Yeah. Okay. [287:12] >> So in generally react or any other thing [287:14] that suppose like immutability the main [287:17] reason would was that it makes [287:19] predictable changes. Y [287:20] >> and because the state is not changing [287:24] >> uh it makes like race condition very uh [287:26] like simple because if the values is [287:29] updating the context the context still [287:33] can use the old values [287:34] >> right [287:35] >> so that's why I had this doubt because [287:37] we are changing the values directly so [287:40] uh sometimes it may be that there is an [287:42] undefined case [287:43] >> well let's let's talk after um and you [287:46] can show me an example because I I work [287:48] closely with the team that builds this [287:49] and they are some of the smartest people [287:51] I I know, way smarter than me. Um, so I [287:54] would love to show them if if you have a [287:55] race condition uh use case, I'd love to [287:57] show them and I'm sure they have a [287:59] solution to the problem. [288:00] >> I don't have right now. I have to offer [288:01] like test fit first. [288:02] >> Okay. So, you're just talking like [288:03] theoretically. [288:05] >> Yeah. I I've never heard of real big [288:08] race condition problems, but I mean, if [288:10] you're doing something super super [288:11] complicated, I'd love to see the edge [288:12] cases of that for sure because that's [288:14] what I do. And so, you know, bring it [288:16] bring it to me. I can bring it to them [288:17] and they'll tell me something smart to [288:18] tell you and then I'll look smart and [288:20] we'll all win. So, [288:23] >> thank you so much, Ryan. Um, we have a [288:26] something to present to you for as a [288:28] thank you for being a speaker. [288:29] >> Oh, very cool. [288:30] >> Thank you very much. [288:31] >> Get a photo. [288:32] >> Oh, yeah. Sure. [288:33] >> Yeah. Thank you. Thank you very much. [288:45] Just before I jump upstage, can I grab a [288:47] selfie with the audience? I never do [288:48] this. I always want to. I always forget [288:50] to. [288:50] >> So, sorry. [288:51] >> I can take it if you want. [288:52] >> No, it's okay. It's got to be selfie. [288:54] All right, everybody say hi. [288:56] >> Yay. Thank you. Awesome. [288:59] >> Thank you so much. [289:00] >> My kids will think I'm so cool. [289:01] >> It's time to get ready for the next [289:02] session. So, if you want to just sit [289:04] tight, we've got the next session [289:05] starting here shortly. [289:10] Another [289:14] one. [291:48] Heat. Hey, Heat. [297:12] Heat. Heat. [297:34] Baby. [297:44] Hey. Hey. [298:16] Heat. Heat. [298:38] Heat. Heat. [302:53] Heat. Hey, Heat. [303:34] I [303:38] love [303:46] you. [304:02] Heat. Heat. [305:04] The program is about to again. Please [305:06] take your seat and silence your mobile [305:08] phones. [305:46] Hi everyone. There's some seats up here [305:48] if anyone wants to come up and sit a bit [305:50] closer up. Make the speakers feel nice [305:53] and welcomed. [305:55] Got a whole chunk of seats near the [305:57] front up here. So, please come on in. [306:04] Um, I'm going to get us started. So, hi [306:06] everyone. Welcome to Lotus 2. Thanks for [306:08] joining us for this session. I'm Tess [306:10] from Automatic and I have the honor of [306:12] being your MC in here for the next [306:13] session. Um, just a housekeeping note [306:16] that we've got closed captioning [306:18] available on your personal device for [306:19] all of these sessions. So, there's a QR [306:21] code on the screen. If you scan that, [306:23] then you'll be able to follow the [306:25] onscreen prompts to begin receiving [306:27] captions. [306:30] So, there's some seats up the front and [306:32] there's a section over here um if anyone [306:36] wants to sit down. Awesome. [306:40] Okay, so our first lightning talk today [306:43] um is Miriam Schwab who is joining us [306:45] virtually. Miriam is the head of [306:47] WordPress at Element, a longtime [306:50] contributor to WordPress and a regular [306:52] speaker at Word Camps. Today she's [306:54] speaking about MCPS. Welcome Miriam [306:57] Schwab. [307:12] Can you hear me? I can't tell. Oh, yes, [307:16] I hear my voice. Okay. [307:18] Hi, everyone. Um, thank you for joining [307:22] me remotely. I'm very sad that I can't [307:26] be there with you in person. And I was [307:27] really looking forward to attending work [307:30] camp Asia and particularly visiting [307:32] India and Mumbai for the first time. But [307:35] you know sometimes we make plans and [307:37] they don't work out. Um but I hope you [307:40] can meet some of my colleagues there and [307:42] uh maybe visit our booth and I hope to [307:44] see you at another word camp. [307:47] So I'm going to be speaking about MCPS [307:51] and specifically [307:52] about how they are in many ways the the [307:56] next natural evolution of how WordPress [307:59] has been extensible until now. How [308:02] WordPress has always had an [308:04] infrastructure that encourages [308:06] innovation [308:07] and creating new products, plugins, etc. [308:12] for WordPress and um how all of that [308:16] experience with that type of approach to [308:18] development and innovation has actually [308:21] is actually positioning us in the [308:22] WordPress ecosystem really well to take [308:25] things forward. So um I'm not sure if [308:28] you can see my slides. Can you? [308:32] Yes. Thank you for nodding. I appreciate [308:34] it. Okay, [308:37] great. Um [308:39] okay so [308:42] what are MCPS? It's model model context [308:45] protocols. That's why um the title of [308:47] the talk is hooks filters and app [308:48] context. Why MCPS are the hooks of the [308:51] AI era. So let's go to the next slide. [308:57] I'm going to assume you can see the next [308:58] slide. Um okay. So I'm going to take you [309:02] back to once upon a time in the early [309:04] days of WordPress when many of us [309:08] including myself would unfortunately [309:12] make changes to WordPress directly in [309:13] core which was very very bad practice [309:16] highly not recommended but in the early [309:18] days we didn't necessarily know better. [309:20] We were learning things, we're testing [309:21] things out, and WordPress already was [309:24] really fun to work with and make changes [309:26] to. And so, in many cases, we would just [309:28] go straight to core and make those [309:30] changes. Um, we're talking I've been in [309:33] the WordPress space over 20 years, so [309:35] this was quite a while ago, but I [309:37] remember that that was how many of us [309:39] approached doing work in WordPress. Um, [309:43] but the hooks and filters did exist. It [309:48] just like editing or felt somehow more [309:50] immediate and it was just it was the [309:52] lazy way of doing things which sometimes [309:54] we do and so we would do that. Um and [309:56] and that was how we were doing things. [309:58] Um next slide. [310:01] So that was all fine until it really [310:03] wasn't. An update would come out a [310:05] WordPress update and it would write over [310:07] the changes that we made in core. And [310:10] eventually that happened enough times to [310:12] enough people that we started to really [310:15] understand the value of the hook system [310:17] and that we should all get behind it. Um [310:19] it wasn't just a nice to have, it was [310:21] the right way to work with WordPress. [310:23] Not because editing core was impossible, [310:25] but because extending it through the [310:27] shared system was so much better for [310:28] everyone, including myself and including [310:30] others. [310:33] And that shift to working with the [310:35] ecosystem, it really took hold obviously [310:38] and developed into something much [310:40] broader. But it wasn't just a technical [310:42] shift. It wasn't just about that we [310:44] could hook into WordPress. It was a [310:46] cultural one. The community collectively [310:49] decided that the future of WordPress [310:51] development and customization should be [310:53] the shared protocol, not the shortcut. [310:56] The decision to build this way turned [311:00] WordPress into one of the most or the [311:02] most successful open source ecosystems [311:04] in history with tens of thousands of [311:06] plugins, thousands of themes, companies [311:09] built around it, hosting companies, page [311:11] builders like Elementor, e-commerce [311:13] platforms like Woo Commerce, all [311:16] interacting with each other without [311:18] coordinating an event, without requiring [311:20] permission, and without even needing to [311:23] actually know each other. Although it is [311:24] always better to work together if you [311:26] know each other, but people could just [311:28] build and extend WordPress or other [311:31] products without actually having to know [311:33] the people behind that product. We just [311:34] needed to agree on the interface. Next [311:37] slide. [311:39] Oh, sorry. Now, next slide. Also, I was [311:43] talking about this slide, but I didn't [311:44] say next slide. Anyways, next slide. [311:47] Okay. So, we're at that moment again [311:51] with AI and shared interface, that [311:54] common protocol that we all have been [311:57] working off of is what I want to talk [311:58] about today because I think we're [312:00] standing at a very similar moment, [312:02] except this time it's not about plugins [312:04] talking to WordPress. It's about AI [312:06] agents talking to your WordPress site [312:08] and from the other side making sure that [312:10] your products can talk to AI. And we as [312:13] an ecosystem, the WordPress ecosystem, [312:15] we're uniquely positioned to grab this [312:17] opportunity and continue to be a leading [312:20] platform due to the decades of [312:23] experience and our very rich history as [312:26] an open and collaborative ecosystem. [312:28] Next slide. [312:33] Okay, so we know that the concept of [312:35] hooks was key to WordPress's growth, but [312:38] let's just take a look at what it [312:40] technically achieved. I know I said the [312:41] cultural aspect was very important but [312:43] obviously the technicalities are [312:44] important as well. So what did it offer [312:46] us? It offered us an open infrastructure [312:49] and a standard that everyone can rely on [312:53] which is actually quite amazing [312:55] considering the scale of WordPress. The [312:57] standard protocol, the hook system, has [313:01] held up and been very reliable all these [313:04] years through all the evolutions of web [313:06] technologies, the different approaches [313:09] to building that we've all taken on over [313:11] the years. It's really held up very well [313:13] and there is no gatekeeping behind it. [313:16] You just find the hooks and you extend. [313:19] There's clear extension point. It's [313:21] quite clear. There's there's structured [313:23] pathways for external code to interact [313:25] seamlessly. with WordPress and with [313:28] other products which I'll talk about and [313:31] um it was about building the rails, [313:33] setting the infrastructure up and then [313:35] getting out of the way of the community. [313:36] I know that particularly in at work camp [313:40] Asia in my experience when I'm there a [313:42] lot of the attendees are developers [313:44] building products and solutions for [313:46] WordPress and WordPress enables that [313:50] because of the infrastructure that's [313:51] there and you can just do what you want [313:54] and do what you think is needed and [313:56] innovate around WordPress and the [313:58] friction is very low or non-existent [314:01] almost. [314:02] So let's just look at a real world world [314:04] example which I know very well. [314:06] Elementor. Elementor is a plug-in. And [314:09] why is it possible to do what it can [314:12] what it does? Because of the [314:14] extensibility of WordPress because of [314:15] this hook system. When WordPress was [314:18] built and launched, it just when [314:19] Elementor was built and launched, it [314:21] just needed to know where the hooks were [314:22] and hook in and it worked with [314:25] WordPress. There was no need to ask [314:27] anyone for permission to integrate with [314:28] WordPress. There's no need for special [314:31] partnerships or licensing agreements [314:32] which you may see in thirdparty [314:33] proprietary systems. Just use the hooks. [314:36] And this carries on to the world of [314:38] plug-in extensions or add-ons. By [314:40] adopting this mindset of extensibility [314:42] ourselves when third party developers [314:44] want to add a custom functionality to [314:45] Elementor today. They just register [314:48] through our extension API and their [314:50] widget appears in the editor alongside [314:51] the ones that we built natively. No [314:53] involvement from us required. It's very [314:56] possible that many of you in the [314:57] audience now have developed extensions [314:58] for elementary. We know that a lot of [314:59] products have been developed in the Asia [315:01] region for elementary and we appreciate [315:03] it and you can just do it. Of course, [315:05] we'd like to know you as well, but you [315:06] can do it without even talking to us or [315:09] there's no need for permission. Just do [315:10] your thing. Now, let's say someone wants [315:12] to build a solution for WooCommerce [315:14] sites that also use Elementor. So, it's [315:16] a cross product situation. No problem. [315:19] This type of cross product development [315:20] happens all the time. We're not [315:22] involved. And because of that, it [315:24] compounds the innovation flywheel that [315:26] exists in WordPress even further. [315:29] Why is this good for all of us? More use [315:31] cases means more users who are having [315:34] their needs answered, which means more [315:36] developers building more plugins, which [315:38] means even more use cases. This is a [315:40] flywheel that spins because the [315:42] infrastructure has been open. The [315:44] extension points were well defined and [315:46] no single entity controlled who could [315:47] participate. People can innovate freely [315:50] without barriers or friction. Next [315:53] slide. [315:55] That's the real power of a protocol. Not [315:58] just as it connects things, but it also [316:00] connects people and it makes connection [316:02] the default which leads to growth being [316:05] the natural outcome. And that is what we [316:07] have seen over the last 20 years where [316:09] WordPress has kept growing to reach a [316:11] point where it's over 40% of the [316:12] internet. Next slide. [316:17] Okay. Okay, so what does this mean for [316:18] where we are today? [316:20] WordPress is doing exactly this again [316:23] right now for the AI era by evolving to [316:26] meet the demands of today based on the [316:29] wisdom and experience of the past [316:31] decades of successful growth and [316:33] innovation. Next slide. [316:38] What the core AI team is building is the [316:41] hook system for AI. It's what I said in [316:44] the beginning, a type of natural [316:45] evolution of our approach to making sure [316:48] that WordPress infrastructure is [316:50] extensible and ready for everyone to [316:52] work on. And the hooks, the hooks of the [316:55] AI era are what the core AI team is [316:58] building out now in a very smart way. [317:01] These are common protocols, shared [317:03] infrastructure, open by default and [317:06] notably built in exactly the way open [317:08] source projects build things which means [317:10] anyone can contribute, anyone can build [317:12] on it and no single entity controls who [317:14] gets access. Sounds familiar, right? [317:17] This is very similar to the hooks um [317:20] approach that WordPress has had until [317:22] now. What this means practically is that [317:24] the opportunity is right in front of all [317:25] of us. the new hook, the abilities API, [317:29] the MCP adapter, the WPA client, they're [317:32] all here for us to build on in the same [317:34] way that we have until now. The same [317:36] instinct that drove developers to learn [317:38] hooks, share what they figured out, and [317:41] build plugins that solve problems nobody [317:42] else had thought to solve yet. That [317:44] instinct is exactly what's needed now. [317:48] Next slide. [317:52] Okay, so WordPress is ready for AI, [317:55] especially once WordPress 7.0 is [317:58] launched is released. We are an AI [318:02] product. [318:03] But many of us are maybe thinking or are [318:07] thinking or have considered what about [318:08] all the advant coding tools that are out [318:10] there? How do we compare? Many of these [318:13] tools are indeed impressive. But think [318:15] about what they're actually offering. A [318:17] managed environment, proprietary data [318:19] format, a platform that controls what [318:21] you can build and how you can extend it. [318:24] The AI is very capable, but the [318:26] ecosystem is in many ways a walled [318:28] garden and the people building on these [318:31] platforms are siloed from each other. [318:33] Everyone is solving the same problems [318:34] independently. I'm building a plugin to [318:37] do a thing and you're building a plugin [318:38] to do a thing. You know, it's like so [318:41] they're but at least in WordPress, we're [318:43] sharing it with each other. They're [318:44] they're building a whatever. Nobody else [318:47] knows that they built it. Everyone's [318:48] repeating themselves. They're not [318:50] learning from each other. So nobody is [318:53] building on what the person next to them [318:54] figured out. There is obviously some [318:57] level of sharing, but it's not the same. [318:58] They're not, you know, releasing their [319:00] code out into the internet for people to [319:03] learn from and build on. So why should [319:06] each of us reinvent the wheel every [319:07] single time? [319:09] That's not just frustrating, it's [319:10] actually very wasteful in time, in [319:13] energy, and in the very real compute [319:15] resources that go into running these [319:17] things. In those wild garden [319:19] environments, whatever you want to do [319:21] and need to do, in almost every case, [319:23] you're going to burn through tokens. You [319:25] don't almost have an option to do things [319:28] yourself and have that control and not [319:31] use AI tokens if you don't want to. Next [319:33] slide. [319:37] But beyond the technical architecture, [319:39] there's something deeper here that I [319:40] think is one of our strongest advantages [319:41] and it's harder to quantify. It's the [319:43] culture. Many of us who came up in open [319:46] source learned by reading people's code. [319:48] I know I did. I would dig into themes [319:49] and plugins. And that's how I learned to [319:51] build with WordPress. We figured out how [319:53] things worked by digging into plugins [319:54] someone else had built. By seeing how a [319:56] problem we were wrestling with had [319:57] already been solved by someone in a [319:59] different country who published their [320:00] solution for free. That's how this [320:02] community has always grown and is how [320:04] human progress works in general. Every [320:06] generation builds on the breakthroughs [320:08] of the one before it. You can't advance [320:10] by starting from zero every time. That [320:13] culture of sharing, learning from each [320:15] other and building on top of what [320:16] already exists can't be manufactured. [320:18] You can't force it into existence in an [320:20] industry that wasn't built around it. [320:22] It's either there encouraged and [320:24] celebrated or it isn't. In WordPress, it [320:27] is and it has been for over 20 years. [320:29] That's not a feature you can copy. It's [320:31] a compounding advantage that the vibe [320:33] coding world simply doesn't have yet and [320:35] may never have in the same way. Next [320:38] slide. [320:41] I've been working hands-on with this [320:43] from the ecosystem side, working with [320:45] third party teams to support MCP [320:47] integrations between their products and [320:48] ours. And the dynamic I've seen [320:50] firsthand is almost identical to what [320:52] Hooks created in the plug-in ecosystem. [320:54] I've been in conversations with teams [320:56] from companies that are technically [320:57] competitors of ours. Working through MCP [321:00] integrations together because the [321:01] protocol makes it the logical thing to [321:03] do. Our shared users benefit when we [321:06] collaborate and the protocol makes [321:07] collaboration easier than the [321:09] alternative. [321:11] Next slide. [321:14] What all this means practically is that [321:15] the opportunity is right in front of us. [321:18] The new hooks are here. The abilities [321:20] API, the MCP adapter, the MPI client. [321:23] The same instinct that drove developers [321:25] to learn hooks, share what they figured [321:27] out, and build plugins that solve [321:29] problems nobody else had thought to [321:30] solve yet. That instinct is exactly [321:32] what's needed now. So, I encourage you [321:34] start learning this infrastructure if [321:36] you haven't already. If you have, learn [321:37] it more. There's always more to learn [321:40] with AI. It's endless. Experiment with [321:42] it. When you figure something out, share [321:45] it because someone in this community [321:46] will build on it just like we've always [321:48] done. Now, it's the ecosystems turn to [321:51] do what is always done. Build on the new [321:53] infrastructure. Expose your plugins [321:55] capabilities through the abilities API [321:57] so AI agents can discover and use them. [321:59] Find the use cases your specific [322:01] audience needs. the niche, the edge [322:03] case, the very specific problem that [322:05] cannot be addressed in core and solve [322:07] it. AI enables us to do way more than we [322:10] could in the past. You really can solve [322:12] niche edge cases now, your own and [322:15] others. Share what you learn. I highly [322:18] encourage that. It's really in the [322:20] beginning when I started using [322:21] WordPress, I I had a blog where I would [322:24] write all of my learnings and I've [322:25] actually started doing that again now on [322:26] my own personal site with regards to AI [322:29] and WordPress and AI in general. I [322:31] highly recommend that we all share our [322:33] learnings and document them for each [322:35] other. We can all do much better when we [322:37] don't have to constantly reinvent the [322:39] wheel. Write about it. Talk about it at [322:41] events like this one, like I'm doing [322:44] right now. Hello. Put your code [322:46] somewhere people can read it. Let the [322:48] next person build on what you figured [322:50] out. Next slide. [322:54] There's no doubt that the AI era is [322:56] going to change how we build, maintain, [322:57] and grow websites. It has already. like [323:00] these changes are already happening. Um, [323:03] and it's actually very exciting. I am [323:06] having such a good time building with [323:08] WordPress. It reminds me of the early [323:10] days when I first discovered WordPress [323:11] and was building with it. It's that same [323:13] level of excitement, maybe even more. [323:16] It's a really exciting time to be [323:17] building in our ecosystem. And what's [323:19] exciting is that everything that got us [323:21] here, open infrastructure, shared [323:22] knowledge, the longtail served by a [323:24] thousand different developers solving a [323:26] thousand different specific problems is [323:28] exactly what positions us to leave what [323:29] comes next. This is how WordPress got to [323:32] where it is. That's how 60,000 plugins [323:34] happened. Not because one company [323:36] planned it all out, but because [323:37] thousands of people over 20 years, many [323:40] of you included, kept extending the [323:43] platform one use case at a time and [323:45] sharing what they built. The era doesn't [323:47] change that dynamic. It accelerates it [323:49] and it gives us a much bigger surface [323:50] area to work with. The history of [323:53] collaboration, extension, and shared [323:55] knowledge isn't just something to be [323:56] proud of. It's what positions WordPress [323:59] to be a leading solution in the AI era. [324:02] Not despite how we've always worked, but [324:04] because of it. Next slide. [324:08] So, let's get together and build what's [324:11] next. I'm excited to see what we all [324:14] create. Um, I think these are very [324:16] exciting times with a lot of opportunity [324:19] and especially now that AI is going to [324:21] be baked into WordPress in the way that [324:23] it is, which is very suitable to [324:26] WordPress, not forcing anyone, but it's [324:27] there for us to choose to use. There's a [324:30] lot more opportunity there as well. So, [324:32] encourage everyone to learn and share [324:33] what you're learning. And thank you for [324:35] listening. Um, you can find me online, [324:40] social media, all the places, post at [324:42] Slack, make WordPress Slack, and my [324:44] email address is here if you want to [324:45] reach out. I hope you enjoy the rest of [324:47] this amazing event. It looks amazing, [324:49] and I hope that I'll be able to see you [324:50] in person at another event soon. Thank [324:52] you very much. [324:56] Thank you so much, Miriam. Next up, we [324:59] have Sum. Sum makes the web safer and [325:02] faster. At Campus Press, he helps [325:04] universities and enterprises with [325:06] large-scale support, CMS migrations, and [325:09] accessibility audits. Today, he's [325:11] speaking about CSPs. Please welcome [325:14] Suman Basulli. [325:22] >> So, [325:24] hi everyone. Good afternoon. I'm Suman. [325:27] So on a daily basis we uh receive [325:31] request to do accessibility audits [325:50] I think I'm audible. Yeah. uh so data [325:54] migrations and to implement uh content [325:56] security policies across like different [326:00] uh sites. [326:02] So most of the time those are like [326:05] multi-sight networks uh where like [326:08] implementing one CSP policy would impact [326:12] hundreds or sometime potentially [326:15] thousands of sites. [326:17] So what is CSP? So [326:21] let's say your uh browser is a nightclub [326:24] and CS CSP is the bouncer of that [326:27] nightclub. So whenever a request hits [326:32] your browser on your site, CSP is going [326:35] to check if it is allowed or denied. If [326:37] it is not in the allowed list, it will [326:41] block that website or script from [326:44] executing. [326:46] no matter if it is like a genuine [326:48] request or a malicious one. So why CSP [326:52] matters? Uh basically CSP is there to [326:56] prevent excss and other client side [326:58] issues. But in modern web security CSP [327:02] is also a requirement of various [327:05] standards like PCIDSS [327:08] and various other things. [327:11] So [327:13] content security policy has different [327:15] directives. So these are some of those [327:18] directives. [327:19] So every directive takes a source. Now [327:23] sources can be like internal source, [327:26] external source or none. [327:29] Internal sources uh are mostly self. If [327:33] you define self, it will u take the it [327:37] will load everything from your like host [327:39] that is your website. If external [327:42] sources is defined, it can be a domain [327:45] name, it can be domain name with port, [327:47] it can be a domain path. Then um if none [327:52] is set, nothing will be loading. And [327:54] there are also some scheme sources which [327:58] uh are like HTTP then data and blob and [328:03] then also for scripts and styles uh we [328:06] can define unsafe inline. If it is [328:09] defined then the inline uh [328:13] then the inline like uh I would say [328:15] scripts for example on click or any [328:18] event listener would be like enabled. If [328:21] hash and nons is used, the browser will [328:23] check if the hash and the nons of that [328:26] specific request is matching with what [328:28] is defined inside CSP. And there are few [328:32] uh script only CSP which we won't like [328:34] get into here. [328:37] Now for example, this is a content [328:39] security policy. Here the default is [328:43] default can be set as self or none. And [328:46] then we have like script uh to be uh the [328:49] script source is example.com and unsafe [328:52] inline is enabled. Style is not [328:55] accepting any source. Image is not [328:57] accepting any source. Font is not [328:59] accepting any source and connect source [329:02] is again example.com. So what will [329:04] happen when this content security policy [329:07] is when your browser sees this CSP? It [329:10] will first look into the child [329:13] directives that are script, style, [329:16] image, font and source. Default source [329:18] is always a global rule. If nothing is [329:22] defined, what is defined inside global [329:24] source like default source would be [329:26] used. For example, if style source is [329:29] not defined and the default source is [329:31] set to self, [329:33] then uh the browser will allow all the [329:37] styles from my specific uh like site uh [329:42] on that like website. So it will first [329:45] look for like child script uh child [329:48] directives and then it will uh fill in [329:52] the blank for all other source that are [329:54] not defined from the default source. [329:57] So let's suppose this is a website and [330:01] we will be implementing uh CSP locally [330:04] on our browser here. So to do so first [330:07] you would need to open the chrome web [330:10] dev console. From there on the first [330:13] request you will right click and select [330:16] override content and then it will ask [330:19] you to select a folder. Once you have [330:22] selected the folder it will ask you if [330:25] you want to enable allow editing files. [330:29] Once it is done a similar interface will [330:32] be shown. [330:34] Then uh now before moving uh forward if [330:37] you want to uh like stop the override [330:40] you you need to un uncheck this checkbox [330:42] and clear all the contents. So after we [330:46] have enabled uh override content we can [330:48] override the headers in a similar way [330:51] and once it is done then we will see a [330:55] button called add header. [330:57] Now with this button we can add our CSP. [331:01] For now let's say let's add like default [331:04] source as none. And then if we reload [331:07] the page you'll see everything is like [331:09] broken. Right now the image seems broken [331:13] but below there there are content. So it [331:16] is not loading any uh source styles [331:19] fonts nothing. It is just raw HTML. Now [331:23] if we like open the console then we will [331:26] see a bunch of errors. [331:29] We won't go into it right now. Before [331:33] going into this, we would enable uh like [331:36] uh for images, scripts, styles to load [331:39] from our website itself. Once it is [331:42] done, if we again reload the page, we [331:45] will see the issue is still there. So if [331:49] we now open the chrome console we will [331:51] see there is a recurring error which [331:54] which says applying inline style [331:57] violates the following content security [331:59] directive and it is saying to enable the [332:02] unsafe inline keyword. If I enable it [332:05] and then reload the page I can see that [332:07] some of the like styles are there. Still [332:10] I don't see my image or anything. So [332:13] this is the iterative process. We will [332:15] again open the Chrome console and see if [332:18] there are any errors. I still see there [332:20] are like lot of errors. If I expand [332:22] those errors, I would see my browser is [332:25] asking me uh and saying that these [332:28] things violate these security policy and [332:31] I need to fix this. So I will implement [332:34] accordingly. So here I enabled WP.com to [332:38] load all the images from and various [332:41] other things. And then once I reloaded [332:44] the page I see my site is now like [332:46] looking good but still there are uh some [332:49] issues like on the left hand side of the [332:51] emoji I see uh image is broken. [332:55] If I open the chrome console I will [332:57] still see that there are many more [332:59] issues to fix. So I will go ahead and [333:01] fix those. So now if I see on my console [333:05] there are no errors. So this is how like [333:09] CSP is generally implemented. So I just [333:12] showed how it is done locally. Similarly [333:15] the similar thing can be done with [333:18] report only mode. So for this we you can [333:20] use uh you can enable the report only [333:22] mode on your like browser uh on your [333:25] like web server or maybe you can use a [333:27] plug-in but the difference is you need [333:31] to enable a report to URI and you need [333:34] to configure this with your login tool. [333:37] You can use any logging tool and once it [333:39] is configured all the errors that I like [333:42] showed in the browser, it will be passed [333:44] on to your login tool login tool and [333:47] from there you will you will do the [333:49] iterative process that I just showed and [333:52] then once uh the like we have resolved [333:56] all the errors you can remove the report [333:58] only and then it will be enforced in [334:02] production. Once it is enforced in [334:04] production, you won't see errors. The [334:07] browser will start to directly block the [334:10] requests. [334:11] Once we have reached this stage, I would [334:14] recommend to still enable the report [334:17] report to URI because in CSP you have [334:20] fixed all the errors that you currently [334:23] have on your like site right now. But [334:26] just in case if you use a a different [334:29] service later for example YouTube if [334:32] YouTube is not like defined on your CSP [334:35] later you do a YouTube embed it won't [334:38] load it will be like blocked so you you [334:41] still need your logging tool to log [334:44] those issues and report to you so that [334:46] you can iteratively like fix it. So [334:49] that's it. Thank you. You can scan this [334:51] QR code. A full technical guide is [334:53] available here and with all the slides [334:55] and if you have any questions you can [334:57] like leave a comment here. That's it. [335:05] >> Thank you so much Simon. We just have a [335:07] little gift to say thank you for [335:09] speaking. [335:27] Thanks very much. Okay, next up we have [335:30] Maitri. Maitri is the director of client [335:33] delivery at RT camp and a committed [335:35] member of the WordPress community, [335:37] frequently volunteering and speaking at [335:39] word camps. Today she's speaking about [335:42] project delivery. Please welcome Maitri [335:45] Chioan. [335:50] >> Thank you Tess. Hi everyone. I'm Maitri [335:53] and I lead the client delivery at RDAM [335:56] where we work on large complex projects [335:59] that usually involve some moving [336:01] timelines, multiple stakeholders, and [336:04] enough ambiguity to keep things very [336:06] interesting. So I'm responsible for our [336:09] client relationships as well as how our [336:11] teams deliver. After being in this field [336:14] for a few years, one starts to notice um [336:18] similar types of problems usually appear [336:21] across projects and some patterns start [336:24] to emerge. Now just raise your hand if [336:28] you've been a part of a project uh where [336:32] you had a great start and you managed to [336:34] release it but somewhere in between [336:37] things got a little complicated. [336:40] Yeah. All right. Quite a few of us then. [336:43] So today I'm not going to be walking you [336:46] through any frameworks cuz you've all [336:48] shipped some projects and you know how [336:50] they're supposed to go. But what I do [336:53] want to talk about today is that [336:55] somewhere in between because here's what [336:59] I've noticed. Projects don't usually go [337:02] sideways because of one big thing. Um [337:05] there's rarely going to be a moment [337:07] where you're going to be able to point [337:08] out and say, "Yep, that's where [337:10] everything went south." [337:13] It's usually smaller than that or maybe [337:16] a handful of moments that have happened [337:18] sometimes quietly or sometimes not so [337:21] quietly. [337:25] Um, and they're almost always avoidable. [337:29] So, if somebody had to just say [337:31] something about it or maybe asked a [337:34] question that was looming in their head, [337:36] those moments could have been avoided. [337:39] Today, that's what we are going to be [337:41] going through. We're going to be talking [337:42] about four moments that I see coming up [337:46] usually on many projects. What's [337:49] actually happening behind them and how [337:51] we can try to avoid it. Now, I'll tell [337:54] you something up front before we get [337:56] into it. None of the moments that I'm [337:58] going to talk about today will surprise [338:00] you cuz I'm sure you have all [338:02] experienced it in some way or the other. [338:05] But what is different is whether you [338:07] were able to catch it early and do [338:10] something about it. So let's get into [338:12] it. [338:15] The first moment usually happens during [338:17] the most optimistic phase of any project [338:19] life cycle. That is your kickoff. [338:22] Everyone is confident. Your client is [338:24] nodding. You are nodding. Everything [338:27] looks great, right? But that's also the [338:30] very same moment where everyone involved [338:33] is starting to think about a slightly [338:36] different picture for that final [338:38] product. Now we might have all [338:40] experienced this. Uh a project kicks off [338:44] you by reviewing your signed off scope [338:47] really well. both sides walk off very [338:49] confidently and then weeks later you [338:52] have a stakeholder join your call and [338:54] say oh I assume that that integration [338:57] was also going to be carried forward was [339:00] that not discussed and no for your team [339:03] that was definitely not discussed but [339:06] neither side is wrong as such they're [339:09] just working off different versions of [339:12] that scope document one that is written [339:15] and one that is being assumed [339:18] The gap between those two versions was [339:21] there from day one, but maybe it just [339:24] didn't surface until a few weeks into [339:26] your project. That is what I call the [339:29] alignment illusion. The moment where [339:32] everyone is saying yes, but they also [339:35] mean something slightly different. It is [339:38] what happens when um a written scope [339:42] meets some human interpretation. [339:44] And in my experience, it is almost never [339:47] about how that scope contract was [339:49] written. The contract was probably fine. [339:52] It's about what everyone assumed was so [339:55] obvious that they didn't feel the need [339:57] to say anything about it. So, how do we [340:01] address this challenge? [340:03] Well, one of the first things that we do [340:05] on our projects is conduct a dedicated [340:09] assumption session. Not just a scope [340:12] review because that happens any which [340:14] ways but it is a conversation where we [340:16] ask the client what are the assumptions [340:19] that we are making for each of the [340:22] scoped items. Is there anything obvious [340:24] about it that we are not talking about? [340:27] Now think about something as [340:29] straightforward as migrating content [340:31] from an old instance into a new [340:33] instance. It sounds very simple, right? [340:36] But there are going to be some questions [340:38] that come up. Does any cleanup need to [340:40] be done for that decade old content? Uh, [340:43] do they want all the assets to be coming [340:45] over? Do they plan to maybe tweak some [340:49] of the content before the new site goes [340:50] live? Now, nobody has spoken about it [340:54] earlier on because they might have [340:56] assumed that the other side already [340:58] knows about it. What comes out of that [341:01] assumption session is a document that is [341:04] shared between both teams and it talks [341:08] about what both sides are treating as [341:10] true before they go into the project [341:12] cycle. It's small it sounds like a very [341:15] small thing to be honest but when an [341:18] assumption breaks later on it is going [341:20] to be the difference between um having a [341:23] quick conversation and a month of back [341:25] and forth. [341:28] The next moment usually happens when we [341:30] are right into that execution phase and [341:33] it is where things get a little bit [341:35] sneakier. [341:37] You are in middle of the project, things [341:39] are moving and the client sends you a [341:41] message. Um, hey, just a quick one. Can [341:44] you add a filter to this page? It should [341:46] be a small request, right? And yeah, it [341:49] is a small request. You should be able [341:51] to do it in the next couple of hours. [341:53] So, you say yes. Then a week later you [341:57] get another request where they say um [342:01] hey we would love to change how the CTAs [342:03] are looking in that hero section. Can [342:05] you please make that tweak for you? You [342:08] go um all right sure I can do that. And [342:11] then again you get another request. Hey [342:14] we saw how the navigation is now looking [342:16] but we want to revisit it. Here are the [342:19] new updates for that. And then suddenly [342:22] you you have a team that is completely [342:24] stretched out. The timeline has shifted [342:27] for you and nobody can exactly explain [342:30] how all of that happened because every [342:33] single ask that came to you in uh during [342:37] the middle of the project was a small [342:39] and a reasonable one. But it was the [342:41] accumulation of these small tasks that [342:44] nobody ended up tracking. [342:46] Now this is one of the most common [342:48] patterns that one sees in delivery. Team [342:52] ends up absorbing weeks of unplanned [342:54] work through these small requests but [342:57] together they could actually form a [342:59] whole project entirely. [343:02] I think of it as a slow puncture. You [343:04] don't notice it immediately, so you keep [343:07] on driving. But a few days later, things [343:10] just start to get heavier and that's [343:13] when you end up noticing what you have [343:14] been compensating for. [343:17] The answer is not going to be to say no [343:19] to everything that comes your way. But [343:22] every time you're going to say yes to [343:24] something, something else needs to give. [343:27] Maybe more time or less scope or a [343:30] conversation about budget. [343:33] What we have found is um what we have [343:36] found that actually works is making [343:38] those tradeoffs very much visible to [343:40] your client. Something as simple as hey [343:43] we can take this on but here's what it [343:46] means for your timeline or we can add [343:49] this new request but here's something [343:51] that we need to drop. [343:55] We keep a running record of um all these [343:58] trade-offs that we are going to be [344:00] making through the project. Every time a [344:02] new request comes in, it goes in that it [344:05] goes in that tracker. What was asked, [344:08] what was agreed, and what had to move in [344:11] order to make room for it. It feels like [344:14] a very uncomfortable conversation to [344:16] have with the client for every small [344:18] request, like you're trying to be [344:20] difficult, but honestly, you're not. [344:22] You're just trying to keep the scope, [344:25] timeline, and budget honest for both [344:27] sides. And when both sides can start [344:31] seeing the full picture, it's um those [344:34] conversations start to get a lot more [344:36] easier for you. [344:38] Now, while you're still in that [344:40] execution phase, the third scenario [344:43] starts coming up. Things are moving. [344:45] Your project uh your team is doing well. [344:48] You all are able to check off all the [344:49] boxes as per your plan maybe. And [344:51] suddenly a new face appears. Maybe a [344:55] senior stakeholder who was away for a [344:57] long vacation has now come up. Or maybe [345:00] a new head of product has joined. Or [345:02] maybe the original decision maker that [345:04] you were working with has now got pulled [345:07] into something else and handed over the [345:09] project to somebody entirely new. And [345:11] this new person has some questions. They [345:15] are going to be reasonable ones like why [345:17] was this approach taken? or why did you [345:20] move away from this idea or who made [345:22] this decision and suddenly everything [345:25] that felt settled for you doesn't seem [345:28] to be that way anymore. It's not because [345:31] the decisions were wrong but it is [345:34] because that context that led to those [345:36] decision only stayed with the people who [345:39] were there in that room and now the [345:42] people in the room are different. [345:45] I have seen this scenario add many weeks [345:48] into the project quietly. Not because [345:51] the stakeholder is trying to be [345:53] difficult. They are just doing their [345:54] job. But because nobody had a record of [345:58] the decisions that were made throughout [345:59] the project. So every conversation that [346:03] was already finalized gets reopened. [346:07] What would be the best way to record [346:08] these decisions now? Well, it is of [346:11] course by maintaining a shared decision [346:13] log, not just some long detailed minutes [346:16] of meeting that you send out every [346:18] single meeting and well half of the [346:20] people don't end up reading it fully. It [346:23] would be something rather simpler where [346:26] it's a shared unified record of all the [346:29] decisions that are being made throughout [346:30] the project like what was decided, who [346:34] decided it and uh why was this done? [346:39] It doesn't need to be a very long [346:41] document. It could be just a few quick [346:43] lines after every significant call or a [346:45] milestone. But what this does, it turns [346:48] the institutional memory into something [346:51] that can travel beyond the people who [346:54] are there in that call. Now, when that [346:57] stakeholder arrives, you can just hand [347:00] them a map that says here's where we [347:03] have been, here's why we made those [347:05] decisions, and here's where we are now. [347:08] That's going to be a difference between [347:10] a two week derailment and a 30-minute [347:13] catch up for that new stakeholder. [347:16] And the final moment usually happens [347:19] when you're just around the end of that [347:21] project life cycle. Um, this is going to [347:24] be the first time that your clients are [347:27] seeing the product entirely end to end. [347:31] Now, think about the first time that you [347:34] might have used uh a new product, not [347:37] just seen a demo of it or read up about [347:39] it, actually used it end to end, and [347:43] then you think something is a little bit [347:46] off about it. It's not broken, but just [347:50] something is not quite right. [347:52] That's actually the feeling that most of [347:55] the clients feel during the UAT phase. A [347:58] document and a real product are entirely [348:01] different experiences. [348:03] You can read a scope and picture [348:05] something or you can approve a design [348:08] and picture something, but when you're [348:11] actually clicking through the final [348:12] product, that picture starts to shift a [348:15] little bit. Again, that's not a client [348:18] who's trying to be difficult. They're [348:20] just being human. So what actually going [348:23] to what actually helps during this phase [348:25] is not more testing or more review [348:28] cycles. It is a conversation that should [348:31] have happened much earlier into your [348:33] project. Maybe around the same time that [348:35] you're reviewing your scope and talking [348:37] about the assumptions that are being [348:38] made. [348:40] You need to be documenting what is the [348:43] acceptance criteria for each of the [348:45] scoped items. It can't be a feeling [348:48] where they say, "Oh, it should just feel [348:50] intuitive." Or, "Oh, it should just feel [348:53] a little fast." These need to be actual [348:56] measurable statements like this feature [348:59] works when this happens or this [349:02] functionality works when it can give [349:04] this particular output. It needs to be [349:07] objective goals and not just some [349:09] subjective ideas. Now, as we all know, [349:12] we call this an acceptance criteria [349:14] document. [349:16] It is not going to be a long one. But [349:19] what it does do is when your UAT phase [349:21] arrives, the question in the room [349:24] everyone ends up addressing is is this [349:27] what we had agreed to achieve and not [349:30] something like hey do you like what is [349:33] what you're seeing. [349:35] So without without this particular [349:38] document, every single opinion that you [349:40] encounter during the UAT phase is going [349:43] to be a valid reason for a revision. The [349:47] line between the bug and an enhancement [349:49] request starts to blur and your nearly [349:52] done moment that you were feeling starts [349:55] stretching even further ahead. [349:57] Now tell me how many of you here have [350:01] been a part of a project life cycle that [350:04] went as per plan exactly how you had [350:07] thought about it since day one. No [350:10] changes in your plan at all. [350:13] All right. Not many of us. Me neither to [350:16] be honest. So here's the thing between [350:19] that gap of plan A and reality. It is [350:23] not a flaw in the process. It's actually [350:25] a feature of working with people. Your [350:28] priorities can shift. Your context is [350:30] going to change. Maybe what seemed [350:32] obvious at the beginning is going to [350:34] look slightly different 6 weeks in. [350:37] That's not a failure. That's just a part [350:40] of project delivery. Here's something [350:42] that I've noticed when you look back at [350:44] all the moments that we just discussed [350:47] that these are some moments that you can [350:49] really plan for. The goal was never to [350:52] eliminate that gap. The goal was to be [350:55] the person in the room who saw it coming [350:57] and put something in place before it [350:59] became a blocker. Now, I've been doing [351:03] this for a while and the thing that I've [351:05] come to respect most about project [351:06] delivery is that the crit is that the [351:09] critical aspects of management are [351:12] always going to be invisible. [351:15] Your plan is going to be visible. Your [351:17] final product is going to be visible. [351:20] But everything in between like the calls [351:22] you might have taken to keep everything [351:24] on track or the moments where two people [351:26] had completely different ideas or the [351:30] decisions that held everything together. [351:33] None of that is going to show up [351:34] anywhere. And that's really what [351:37] delivery is. The invisible work that [351:40] makes the visible product possible. And [351:44] that gap between your plan A and [351:46] reality, that's not failure. That's [351:49] actually where the real magic happens. [351:51] Thank you so much for joining me. [351:58] >> Thank you so much, MRI. Stay here. We've [352:01] got something for you just to say [352:02] thanks. [352:27] Thank you so much everyone. Um we're [352:29] getting ready for the next session now. [352:31] So sit tight, connect with somebody that [352:33] you don't know and uh look forward to [352:36] the next session. Thanks a lot. [352:56] Hey. [353:04] Hey. Hey. [353:26] with you. [353:29] Hey [354:07] I feel [354:11] I feel [354:23] Heat. Heat. [354:43] Heat. Heat. N. [355:10] Heat. Heat. [356:17] feel. [356:28] Hey [356:32] Heat. Heat. [356:56] Heat. Heat. [357:32] Hey, [357:48] hey, Heat. [358:06] Heat. [358:12] Heat. Heat. [358:52] Hey, hey, hey. [360:22] Language. [362:07] The program is about to begin. Please [362:09] take your seat and silence your mobile [362:10] phones. [363:12] Uh I hope everyone is enjoying the [363:14] session till Now uh so far I'm picking [363:16] up some great insight. Uh before [363:19] starting the session I want to share [363:22] some quick housekeeping notes that close [363:24] captioning is available on your personal [363:26] device for all the sessions. Scan the QR [363:29] code for this room and follow uh the on [363:32] the screen prompt to begin receiving the [363:34] caption. [363:36] Hello and namaste. I am Indira Vishwas. [363:38] I am a QA engineer at RTam. I'm from [363:42] Pune, India and this is my first time [363:45] ming and if you see me nervous that's [363:47] because I am. Uh but I'll try my best to [363:50] give um I will try my best to give uh [363:53] all the possible delivery. Uh so uh [363:57] let's start with a session. Uh so next [364:01] we have a speaker uh [364:05] who started his journey as early as when [364:08] he was 13 years old. uh and since then [364:11] he hasn't stopped. He leads a different [364:14] uh uh vertical like uh store store apps, [364:18] icegrams and partler. Uh in his talk he [364:21] will share what actually happened when [364:23] his team all goes in AI the wins the [364:26] failures and everything in between. [364:29] Please help me to welcome Nav Ma on the [364:31] stage. [364:40] Thank you. [364:43] Okay, I'm going to get the clicker. [364:57] We need the slides to go back, [365:03] Tommy. Yeah. Okay. All right. So, uh, [365:07] Indira said that she was nervous. I am [365:11] too. You know, whether it's the first [365:13] time or the hundth time, every time I [365:16] get on stage, there are always [365:17] butterflies. And I think that's for a [365:20] good reason. [365:22] But, uh, welcome. Until we figure this [365:24] out, just settle in. Maybe think about [365:27] your own AI experiments as well. And, [365:31] uh, once we're ready, we can begin. [365:38] All right. [365:42] Yeah. Settle in. Settle in. [365:49] So the year was 1458 [365:54] and Thomas [365:56] is a scribe [365:59] writing verses manually on pages. [366:04] his fingers aching with pain due to all [366:07] the careful strokes that he has to take. [366:12] It takes him one day to write one page. [366:16] The skills he's mastered have taken his [366:19] own life [366:21] and the love and devotion [366:25] that he writes each letter with. [366:28] And then [366:30] a new machine somewhere in the [366:32] neighborhood [366:34] starts printing books [366:37] at a speed and cost that Thomas cannot [366:41] even match. [366:44] Patrons who paid gold to Thomas earlier [366:47] for his work have now started buying [366:51] machine printed books [366:54] and they're marveling [366:56] at Johannes Gutenberg's invention. [367:02] Printing press [367:05] wiped the scribes. [367:10] Thomas was forgotten. [367:15] And [367:18] we are at an interesting crossroad today [367:24] because every new technology [367:28] goes through an entire disruption [367:30] process [367:32] whether that was printing press or [367:35] whether that's the railways the [367:37] electricity the assembly line [367:41] you like you know the PC PC, the [367:43] internet, the smartphone, [367:47] there is a huge disruption, a creative [367:50] destruction that happens in the process. [367:56] So today we have face a very [367:58] uncomfortable question. [368:02] Ain [368:10] gone are the good times. [368:16] AI, [368:20] right? [368:22] Will AI [368:24] replace all the web agencies? [368:28] Will AI replace all the developers? [368:32] Will it replace all the content writers, [368:35] marketeteers, copywriters, [368:37] graphics people? [368:41] And deep underneath all of this is a [368:45] very personal question. [368:49] Will I be the next Thomas? [368:53] Will I be able to survive? [368:57] But I believe it's not a time to hide. [369:01] It's time not to be the frog in a [369:05] heating pot. [369:09] You know, it's time to take action. [369:12] So, that's what we did. You know, we had [369:15] heard um all the stories of um how AI [369:21] can convert your designs into your apps [369:23] like fully built apps. So last year when [369:28] we wanted to redesign ice express our [369:31] email marketing solution we said let's [369:33] try AI. So we took the Figma design and [369:38] we give it to cursor and all the other [369:41] tools and the first output was quite [369:45] impressive [369:48] but not really usable. [369:51] So we're trying a few things and then we [369:54] took the Figma design and hooked up the [369:56] Figma MCP with uh you know GitHub [369:59] copilot and claude and tada [370:03] the output was actually usable. It [370:06] generated flawless you know react code [370:08] that we could use in WordPress [370:11] and we were excited. [370:13] Why? Because what could have taken us 9 [370:16] months, [370:18] we completed the entire process of [370:21] developing that plug-in like complete [370:23] redesign of the product in just 2 to 3 [370:25] months. [370:29] That was amazing. [370:32] And then the bug reports started coming [370:35] in [370:39] in the excitement of faster code, faster [370:44] prototype, faster output in the race to [370:47] 10x the productivity. [370:50] We did not test the thing enough [370:53] and then it started breaking. every [370:57] release would you know cause more bugs [371:01] and then [371:04] we would be just you know spending time [371:06] fixing all the AI slop. [371:10] So a large portion of the time after [371:12] that actually went in fixing all the [371:14] bugs that AI had mixed. And why? Because [371:18] what looks good on the screen, AI [371:21] probably forgot that particular edge [371:23] case, the workflow, the rules that we [371:27] wanted it to have. We just assume [371:28] because it looked good on the screen and [371:30] then we had done it in the Figma and the [371:32] output worked. We thought it will take [371:34] care of it. [371:36] But no, that's not how it happens. [371:41] You know u in the wonderland Alice in [371:45] Wonderland if you heard the story [371:47] there's a character called white rabbit [371:49] so the white rabbit goes like this he's [371:51] checking time you know we are late we [371:52] are late come fast come fast [371:56] running around like that [371:59] so AI enters into our life something [372:03] like that you know we are late we are [372:06] late fast let's do things fast. [372:10] So that's the white rabbit. [372:14] The unfortunate thing is [372:17] speed [372:19] does not equal direction. [372:24] You can generate tons of code [372:28] but that's just code. [372:30] Where is the vision? Where is the [372:32] architecture? Where is the business [372:34] logic? That is something that you still [372:37] need to feed. [372:40] You can delegate the task of you know [372:42] doing some code but please do not [372:45] delegate the task of architecting the [372:47] solution [372:49] because what happens is that you know [372:52] developers and everybody goes through a [372:55] behavior arc. [372:57] So first we have like this complete uh [373:01] disbelief [373:03] that this will not work you know AI will [373:05] not work. We begin like that uh it may [373:08] be working for you for our task it [373:10] doesn't work and then we start [373:12] experimenting it a little bit and then [373:14] we get comfortable. [373:17] Once we get comfortable the next stage [373:19] is blind trust. [373:25] That happens every time. No matter what [373:28] is the technology. [373:31] If you pick up your phone to check time, [373:35] do you ever question whether this is [373:37] right time or wrong? You don't [373:41] because you have blind trust in that [373:43] technology. [373:46] That's a very dangerous spot to be in [373:49] especially with AI. [373:52] All right. So, we all go through that. [373:55] So we went there experienced that but we [373:59] couldn't stop. [374:01] So in one of our uh hackathons one store [374:05] apps engineer he took on the task the [374:08] hackathon is for two days and he took on [374:10] the task that I'll build a completely [374:11] new plug-in only using AI very good and [374:17] he had it ready in two hours like the [374:20] plug-in was ready the entire UI the [374:21] whole plugin everything works you know [374:23] it just checks out things into WCommerce [374:25] whatever like you name it it's there [374:30] he didn't win in the hackathon [374:33] because the next two days [374:36] he fixed bugs after bugs after bugs [374:39] after bugs. [374:43] So it's it's not superhuman speed. [374:47] Sometimes it's super human technical [374:50] debt that you get with AI. [374:53] So you got to be careful how you [374:56] navigate that. [374:58] So you know while the technical team was [375:01] doing all these experiments the [375:03] marketing team was like you know [375:07] magic potion magic potion [375:10] magic potion. So what the magic potion [375:13] does turns magic. So marketing was then [375:17] like all over the place applying AI. [375:22] So they would take obviously [375:25] all the 200 blocks and then write them [375:28] and rewrite them with AI [375:32] correct [375:35] and not only that they would actually [375:38] take landing pages [375:41] u SEO AEO competitor research uh email [375:47] scripts [375:49] social media memes [375:51] what [375:52] ads [375:55] everything that you can imagine under [375:57] the sun throw it to AI [376:01] and it works honestly it did work you [376:06] know the some of the rankings improved [376:10] uh we started ranking in the AI [376:12] placements you know whatever AI engines [376:13] started citations we got start you know [376:15] some citations there [376:18] the speed was amazing [376:21] We could produce five times the content [376:23] in the same speed. [376:28] But the magic portion also has its [376:29] weakness. [376:35] So you know when I thought about [376:37] marketing [376:40] I think the biggest problem with AI is [376:44] not that it can produce bad content or [376:47] bad output. [376:49] In my opinion, the biggest problem with [376:51] AI using in marketing or anything else [376:54] for that matter is that it is extremely [376:57] good at producing mediocre [377:01] quality content that looks acceptable [377:06] and it can do that at unprecedented [377:08] speed and scale. [377:11] See what happens is if something is [377:13] obviously bad, it's going to get [377:15] rejected. [377:17] But if something seems decent, it's [377:21] acceptable looking, [377:23] it's going to get published even if it [377:26] is mediocre. [377:29] Now, what's the problem with this? [377:31] How the web works, [377:33] if the entire web is filled with all [377:36] this mediocre content, [377:39] then succeeding [377:42] is going to get harder and harder. [377:47] because everybody's doing it. [377:52] So, the magic portion may be good, but [377:55] you really need to think about what's [377:56] the long-term effect of this. [378:00] And here's another thing. AI can be good [378:03] at all this work. But do you think AI [378:07] can produce trust, [378:10] originality, [378:13] taste? [378:15] And do you think that matters with us [378:17] humans? [378:19] Yes. No. [378:20] >> Yes. [378:21] >> Yeah. [378:22] So more than anything else, what matters [378:24] for you and me is trust, originality, [378:27] and taste. [378:31] So when you're thinking about marketing, [378:33] the real question, the mature question [378:35] to ask would be, can I use AI [378:39] to produce something without losing [378:42] trust? [378:45] That's the right question to answer. [378:50] So you know technicals, engineering was [378:53] doing something, marketing was doing [378:54] something. [378:56] Now let me ask you another question. [378:59] What would you do if you found a magical [379:03] hammer? [379:05] What would you do if you found a magical [379:07] hammer? [379:12] You'll find every possible nail. [379:17] Yeah. So, this is a picture of u an art [379:21] poster that I created for my home [379:24] recently. So, we're doing the home [379:26] renovation and it's just nostalgia. It's [379:29] my memories, you know, it's my first [379:30] computer and whatever not, but [379:33] completely AI made. Okay, not completely [379:35] AI made, but a large portion of this is [379:37] AI generated. And not just this, I [379:40] actually, you know, wallpapers in the [379:42] home we customize using AI. What was I [379:45] doing? Got a magic hammer. So, got to [379:47] use it. Yeah. [379:51] But not only that, at work, we did a lot [379:53] of other things also. uh quite useful [379:55] ones. So we took our support channels [379:58] and we wrote u you know a rack based [380:00] system that would answer support [380:02] queries, answer support tickets in [380:04] finance. we started uh asking AI [380:07] compliance questions and the finance [380:10] team would also you know go and fill in [380:13] Excel formulas with AI now or the HR [380:17] will create you know uh fun activities [380:20] team training programs and even [380:23] responses to emails using AI [380:26] very common we've all done this [380:32] so when we have a magical hammer it is [380:34] very obvious obvious that will try to [380:37] hammer everywhere. [380:41] Is it good? [380:44] No. Should be good. It's magical. [380:49] Not everywhere. Yeah. So, how many of [380:52] you have been burnt by AI? Like you [380:56] tried something, you expected some [380:57] results and it didn't work out the way [381:00] you expected. [381:04] Yeah. Yeah. All of us [381:08] they say that today 70% of the AI [381:11] projects are not generating the ROI. [381:17] There's a lot of hype [381:19] but what's the return? [381:24] So this is what we started thinking [381:25] about next that okay we did all this now [381:27] all the hype [381:29] we should be AI native. If we are not AI [381:33] native, why will customers buy us? [381:36] No, we need to have AI native in our [381:38] marketing pitch on the sales pages [381:40] everywhere whether we are or not, [381:43] you know. [381:46] So, you know, we have another product [381:48] called Putler which is an analytics [381:50] tool. So, it aggregates transactions [381:52] from you know multiple channels and [381:53] gives you clean crisp dashboards. [381:57] So I said you know what people are [381:59] drowning in data and numbers so let's [382:02] create something that will give them [382:04] insights and let's use AI for that [382:07] remember [382:10] so we you know we took on like about 3 4 [382:14] months did a lot of research [382:16] um the team that's working on is sitting [382:18] here but still uh you know we did a lot [382:21] of research and we worked a lot on it [382:23] and we said you know let's do an MVP [382:25] can't launch the full product [382:28] to let's do an MVP and let's send an [382:30] email with these insights to our [382:33] customers. Make sense? Small risk. [382:37] So we took our AI engine, we ran the [382:39] data through it, we generated insights [382:40] and we send emails to customers [382:44] and we get a response. [382:47] Stop this AI junk. [382:52] Ouch. [382:54] That hurts. [382:59] That's how it is. [383:02] You know, some of this whole AI thing is [383:04] like the chashire cat here. [383:07] It smiles widely. It's very confident [383:12] and then suddenly disappears. [383:17] So some of the times you know it'll all [383:19] looks rosy and promising but then where [383:22] is the substance? [383:24] People want substance. [383:29] So then what happens is people go [383:30] through this curve. This is the classic [383:33] hype cycle um from Gartner. And whenever [383:38] there is a new technology, this is what [383:39] it goes through. There is some trigger [383:43] that raises the expectations [383:46] like tremendously. We have like [383:47] unrealistic expectations from a new [383:49] technology and we try it out and then [383:52] there is this whole trough of [383:54] disillusionment. [383:57] Our initial expectations get broken. We [384:00] ex you know we hit failures and then we [384:02] like the whole illusion goes away and [384:06] then we are in the trough. [384:09] I think a lot of us are in that trough [384:12] right now. You know, we've experimented [384:14] with AI, we've seen some success, but [384:16] we've seen a lot of failure [384:18] and like now what? But we are also aware [384:22] of the risk. [384:24] Yeah, we don't want to be the Thomas or [384:26] the frog. [384:28] And we are at that trough. [384:31] What's important when you're in the [384:33] trough is to not stop learning [384:38] because this is the perfect time to [384:41] climb that slope of enlightenment [384:46] and once you climb that slope of [384:48] enlightenment then it you know persist [384:50] and then you have productivity and all [384:52] the results. When will this arrive? I [384:55] don't know. Will it arrive? Yes. [385:02] Right? So keep that in mind that this is [385:04] going to happen. This is how things [385:05] work. [385:08] But now you know we have the reality [385:11] also that we need to look at. [385:14] So what is the reality that we saw with [385:16] all these experiments? Now the first [385:18] thing that we realized was the proof of [385:19] concept is easy. [385:24] You know if you are using AI you can [385:27] impress yourself in 10 minutes. [385:29] with a result. You can impress your boss [385:32] in one hour. [385:34] You can impress the world with a social [385:36] media screenshot immediately. [385:42] But [385:44] the reality hits when it has to go to [385:46] production. [385:48] Proof of concept is easy. Production is [385:51] super hard because you need to take care [385:54] of guard rails. You need to take care of [385:56] quality control. You need to take care [385:57] of operational reasons. You need to take [386:00] care of so many things to make something [386:02] successful in production. [386:05] And let me tell you what, [386:08] most of us [386:11] aren't even capable enough. [386:15] We've worked in WordPress [386:18] and the skills that this whole AI thing [386:20] requires is very different. [386:24] The complexity is so hard. so different [386:27] that we need to upskill ourselves a lot [386:31] and that upskilling [386:34] no matter how quickly we want it to [386:36] happen it doesn't happen that fast. [386:40] So this is another thing that we [386:42] experience that yeah there is you know [386:44] complexity versus capability gap. [386:48] Even if you go after that the landscape [386:51] keeps changing there is something today [386:54] 3 months later that whole thing is [386:56] replaced by something else. [386:59] Then wait next 3 months both of these [387:01] things are replaced by something else. [387:03] You're lost you know running around fast [387:06] fast fast fast fast fast fast fast fast [387:06] fast fast fast fast fast fast fast fast [387:06] fast fast fast fast fast fast fast fast [387:06] fast fast fast fast fast fast fast fast [387:06] fast fast fast fast fast fast fast fast [387:07] fast fast fast fast fast fast fast fast [387:08] fast fast fast fast fast fast fast fast [387:08] fast fast fast fast fast fast fast but [387:08] then what is happening delays delays [387:11] delays [387:14] because the whole landscape is also [387:15] changing what next [387:20] important one it costs money to do [387:24] research all that time costs and more [387:27] importantly like we realized [387:30] you don't want customers telling you [387:31] stop this AI junk [387:33] because that's costing you reputation [387:36] and that reputation is more important. [387:41] If I send you an email written by AI, [387:44] how many of you will recognize it that [387:48] see look around. [387:51] So if we are using AI to write an email, [387:53] what do you expect? [387:55] Why do you expect different results? Why [387:58] do you expect people to click on that AI [387:59] email? Everybody here has used AI so [388:03] much that they can immediately tell if [388:06] something is AIdriven. Like you can [388:08] easily tell this image is AIdriven, AI [388:10] generated, isn't it? Can you not? [388:13] It's obvious. [388:18] So [388:19] that's the thing. And the last thing is [388:22] this that when you have a lot of these [388:25] ideas, you have the hammer, you have a [388:28] lot of ideas, you're like, "Okay, let's [388:29] build this plug-in because we can build [388:30] it quickly in two days." You know, let's [388:32] create this new site because we can do [388:34] that in half an hour. So, you'll have a [388:36] lot of ideas and then you'll have to hit [388:37] this which is the effort impact matrix. [388:41] Classic, very standard in management. [388:43] What you have to do is you have to look [388:45] at the efforts versus the impact. If [388:48] something is low effort, high impact, [388:51] do you do it or not do it? You do it. [388:54] Obvious. If something is high impact, [388:58] high effort, would you still do it? [389:01] Maybe yes. That's a call you have to [389:04] take. [389:07] The problem is [389:10] AI does not fit into this dynamics. [389:15] One because our estimates of effort [389:22] are always less. We will always [389:24] underestimate efforts and we always [389:27] overestimate [389:28] impact. [389:32] And there is this whole uncertainty [389:33] around AI. [389:35] So this matrix does not work. So what we [389:38] did, we kind of expanded it a little [389:40] bit. We said let's kind of refine it. [389:42] And this is the new metric [389:45] where what happens is you you've created [389:47] a space for a lot of loss generators. [389:51] Meaning an AI project is not necessarily [389:54] generating revenue or the return on [389:56] investment. It actually can make you [389:58] losses as well. Okay. Some of the things [390:02] can be money pits where you know you're [390:03] throwing money at it. It's just eating [390:06] money not returning enough. It's not [390:08] necessarily loss generator. It's not a [390:10] revenue accredited service or product [390:12] that you're building. [390:15] The other thing to note [390:17] is look at the size of these quadrants. [390:20] Are they equal? [390:23] They are not equal. [390:26] we the quadrants are in AI's favor so to [390:32] say they are not in your favor [390:35] because the high impact loweffort zone [390:39] is so small. [390:45] So, you know, whenever you're looking at [390:47] something to hit with that hammer and [390:50] you have that bright new idea about AI, [390:53] please run it through this matrix. [390:56] Does it make sense? Is it worthwhile? [391:00] Because [391:02] you know, while AI makes making things [391:05] easy [391:07] and fast, [391:09] not everything is worth making. [391:12] You got to pick your battles because you [391:14] can't fight all the battles at the same [391:16] time. You have to pick your focus. Okay? [391:19] And to pick focus, you'll need something [391:21] to make your choices based on. [391:25] Okay? [391:28] Let's go to another story. [391:34] Rafi Ricoran, [391:36] he is the CTO of Mozilla and he was the [391:40] head of the self-driving cars like the [391:43] taxis division at Uber. [391:46] So last fall he was driving his uh Tesla [391:50] X in the full self-driving mode and he [391:53] was navigating the Bay Area residential [391:55] area which is you know where he has [391:57] driven hundreds of times in full [391:59] self-drive mode and he was just kind of [392:01] dropping his son to the boy scout [392:04] meeting [392:05] and uh it's just going in full self [392:07] drive mode everything is all well and [392:10] there is a turn the car takes the turn [392:14] but suddenly the steering jerks and the [392:16] car decelerates [392:18] and before Rafi could take control of [392:21] the steering they just hit the wall. [392:27] Thankfully Rafi and his sons were not [392:30] harmed. They just had a little bit [392:31] concussion and some bit of headache etc. [392:33] No major injuries but the car was [392:36] totaled like completely gone. [392:40] When he took the matter to insurance, [392:44] Tesla rejected this is not their fault [392:48] because the driver did not act in time. [393:02] See what happens is that whenever any [393:05] automation works well, [393:10] we develop a misplaced confidence in the [393:13] automation. [393:17] Uh you know psychologists they have a [393:19] word for this. This is uh called the [393:21] vigilance u deterance or something. [393:25] Okay. where what happens is that you're [393:27] not that much vigilant for something [393:30] that has been working well. [393:33] You know, we become automatic, right? [393:36] That things work and they will work. [393:39] So the systems are designed [393:43] to give you confidence. [393:46] So you become complicent. [393:48] But if something fails in the system, [393:51] who's responsible? [393:55] you are to blame. [393:59] That's a sad thing. [394:02] What does that mean for us? We cannot [394:05] let our vigilance drop. [394:08] Very simple. No matter how many times [394:12] this has succeeded, [394:15] there is a chance that it can fail. [394:18] Murphy's law always works. [394:23] So you know then I have kind of three [394:26] buckets in which I try to or want to put [394:30] our AI experiments. [394:32] Autopilot with a review, co-pilot and [394:37] advisory. So what are these three gears? [394:40] Autopilot is for things that are very [394:43] easy that you can outsource to AI. Okay. [394:46] But then you do some review [394:48] you know very simple things like doing [394:50] summaries doing some basic research uh [394:52] rewriting something here and there or [394:55] you know even writing some test cases [394:56] which you have speced well so a lot of [394:59] these things can be given out on [395:01] autopilot to AI and then you do a review [395:05] a simple review a quick review that [395:07] should be enough that's the autopilot [395:10] with review mode that's one gear [395:13] the next is co-pilot So all your uh you [395:16] know draft generation, code generation, [395:19] design generation, [395:21] any of these things where you use AI and [395:24] you use it as an assistant to you. You [395:28] are still driving it. You're making the [395:31] decisions but the AI is assisting you. [395:35] Right? That's the co-pilot mode. And [395:36] this is where most of the productivity [395:38] happens today. And if you're not using [395:41] it, please go ahead and use it because [395:42] there is still ton of productivity to [395:45] unlock [395:46] with the co-pilot mode. The advisor [395:50] mode, I think, is very interesting. [395:52] It's very interesting and I've tried it [395:54] and it works beautifully well. So what [395:57] is the advisor mode? [396:00] The way AI agents, AI LLMs are designed [396:03] is that they've trained on a lot of data [396:06] and today they can also, you know, [396:07] search internet and get more [396:08] information. [396:10] So what if AI could be your advisor? AI [396:14] could be a coach. [396:17] AI could be the person who questions [396:19] your assumptions. [396:21] It could be somebody who asks you for [396:24] and you know finds your blind spots. [396:27] What if you could use AI like that? So [396:29] what if you ask AI think like this [396:32] particular expert [396:34] and then tell me what should we do in [396:36] this situation [396:39] or challenge your assumptions about [396:42] something or you can ask AI you know an [396:44] interesting thing that I've done is that [396:47] do something in one AI then pitch it in [396:50] another AI find faults in this [396:53] find the loopholes [396:55] and that works very [396:59] So, whatever you're doing in AI, figure [397:01] out which gear you want to drive in, but [397:06] never driverless. [397:09] Still not that time. [397:12] Easy. [397:14] All right. [397:16] So, let's move to the next. [397:20] What would happen [397:22] if you could use AI [397:25] and one person could do everything from [397:27] design to deployment? [397:33] You may have heard that story recently [397:34] of Medvi. So there is this company uh [397:37] and two brothers. They built this whole [397:41] $1.8 billion company just two people [397:45] rest is AI. [397:47] Now I don't want to vouch for them etc [397:50] etc because I think they did a lot of [397:52] dark patterns and lot of practices that [397:55] I wouldn't kind of you know do a lot of [397:57] unethical things in my opinion but it's [398:00] a proof that you can actually do wonders [398:03] with AI [398:05] just two people company [398:08] now so if you have a bunch of smart [398:10] folks like this in your team you know [398:13] friends and then if you use AI [398:17] you can actually create miracles. [398:22] So when you want to, you know, implement [398:23] AI in your organizations, there are some [398:26] challenges that you will face. So I'm [398:28] just going to go quickly because there's [398:29] a lot of nuances to this. Your situation [398:31] may be different. So I'm just going to [398:34] name some open source tools that may be [398:36] helpful to you. The first challenge that [398:38] you'll face is how do I share my API [398:40] keys with my team? Okay, so you can use [398:42] light LLM. they had a security incident [398:44] recently but still by frost these are [398:47] good tools which will allow you to share [398:48] the API keys so you don't have to give [398:50] your actual openi key or the cloud key [398:54] to your developers or your team they get [398:56] a you know uh virtual key and you can [398:58] manage everything at the light lm that's [399:01] like the top proxy kind of a layer works [399:03] very well you can use langfuse or gizard [399:06] for prompt management uh like you have a [399:09] lot of prompts and then you want to see [399:10] which prompt works well which prompt [399:12] costs more uh you want to run [399:14] evaluations on it like okay is this good [399:16] enough output or not uh this card you [399:18] can use even if you wanted to do some [399:19] security checks on the output uh these [399:22] tools are amazing well-built again [399:25] everything that I'm mentioning here is [399:26] all open source uh if you wanted to do [399:29] workflows then you can use nin most [399:32] people have heard of it uh or you can [399:34] use langraph which is another [399:36] orchestration or workflow tool you know [399:38] works well with the whole AI chain [399:43] There will be chances where you'll have [399:45] to use your own knowledge base like a [399:47] support or anything that requires [399:49] specific knowledge that is core to your [399:51] industry or the problem that you're [399:52] solving. You'll have to use a rag. [399:54] You'll have to use a vector database [399:55] there or if you wanted to test some [399:57] things easily on your machine, you can [399:58] use anything LLM. [400:02] See one thing, one problem with the [400:04] whole AI thing is that a lot of times we [400:06] take it as a blackbox. AI generates [400:08] something and we don't even know what's [400:11] happening. behind the scenes. So that's [400:14] a black box. That's a very high risk. [400:18] So if you don't wanted to do that and [400:19] you wanted to try some things your own, [400:21] you can actually run LLMs on your own [400:23] computer. And for that you can use Olama [400:26] or you can use LM Studio. Um I would [400:29] highly recommend trying them out. You [400:31] know, very easy. Some of the models are [400:32] not so huge. They'll run on your CPUs [400:35] with decent amount of RAM and it's just [400:37] fun to explore things like that as well. [400:41] So, you know, these are some tools that [400:44] you can potentially use when you wanted [400:45] to apply AI to your organization. [400:49] So, we talked a lot about AI. I've told [400:51] you a lot of stories. Uh, but what next? [400:58] The times are still uncertain. [401:02] The ride is likely going to be brutal. [401:05] The oceans are already red and they are [401:08] likely going to get bloodier. [401:10] The competition is going to be intense. [401:13] The survival game will require us that [401:16] we reinvent. [401:20] Uh I told you the printing press killed [401:23] the entire scribes industry. [401:27] But what also happened [401:29] is that it gave rise to completely new [401:32] industries. [401:34] So every job that is taken away in a [401:37] technology disruption is a specific job. [401:41] But every technology disruption also [401:43] gives rise to new industries which are [401:45] completely different and new. [401:50] The new industries that the printing [401:52] press gave rise to are what like [401:54] authors, publishers, bookshops, online [401:57] stores. [401:59] All of that came because of the printing [402:01] press. the whole knowledge the whole [402:03] like WordPress is probably a result of [402:06] printing press you can trace it back to [402:07] that a lot of uh you know progress in [402:10] the society you can trace it back to [402:12] Gutenberg [402:15] so while the times may be tough and as [402:20] you go through that messy middle it can [402:23] be very scary and indeed it is all of us [402:26] are scared [402:29] but you know what matters is that you [402:31] focus on the other side which is the [402:33] blue ocean. [402:35] Think about what are the opportunities, [402:38] what new opportunities can you create? [402:42] What are the problems that AI will [402:44] create that you can solve? [402:47] There is this huge market opening up for [402:49] fixing the AI slop. [402:52] What are the other problems like that [402:54] that you can probably solve? [402:56] or you know if you look beyond our [402:59] technology bubble then what are the [403:01] other problems in the real world that [403:04] you can solve with technology [403:09] so think about going beyond think about [403:11] expanding think about new markets blue [403:14] oceans [403:16] and where you can position with that [403:18] again you'll have to stay focused you [403:20] know pick your battles well but I think [403:23] see the whole WordPress game will [403:27] Website making is not going to be the [403:28] same. Careers will change. Businesses [403:32] will have to change because AI is not [403:36] going away. It will change transform a [403:38] lot of things will change but you know [403:41] human beings are resilient [403:45] and we always invent new ways of [403:48] survival and growth and I think that's [403:51] what we need to do today [403:54] and the thing that we'll have to [403:56] probably practice if I wanted to kind of [403:58] summarize and give you one thing [404:01] you know that'll be this stay curious [404:05] stay grounded [404:06] This is what will help you navigate [404:09] through this whole AI wonderland, right? [404:11] Thank you very much guys. [404:20] >> Thank you Nero. We have few minutes for [404:23] having few questions. Please use the [404:25] microphone uh provided in the aisle if [404:27] you have few questions. [404:32] So if you have any questions please come [404:33] to the mic introduce yourself uh you [404:36] know say your name where are you from [404:37] and then the question in as uh short [404:40] manner as you can and I'll try to you [404:42] know be my guest you know answer. [404:45] >> Hi everyone my name is Jal [404:47] >> and I'm from Boda. [404:49] >> So basically uh we talk about AI slope. [404:53] So my question is like uh we have seen [404:57] that the customer that used to outsour [404:59] us to work to us have started developing [405:02] their own thing. [405:03] >> Yeah. So basically let's say they are [405:05] creating a landing page but as a [405:08] developer I know that this is AI slope [405:10] the purple color website with the inter [405:12] fonts and easily readable but they have [405:15] this bias that like human have this bias [405:19] if they have created thing like if any [405:22] kne guy creates a poem he thinks that [405:25] this is kn but a poet know that this is [405:27] not nice. So when this uh bias is going [405:31] to get fixed and if let's say like as a [405:37] designer and developer I have found the [405:39] equation to create the design which is [405:42] not a AI slope. Now I have this formula [405:46] eventually everyone is going to develop [405:48] their own formula. So is it going to be [405:50] a new AI slope or the problem is going [405:53] to get solved. So this is my first [405:56] question and second question is like uh [406:00] we have recently seen that anthropic and [406:03] even anti-gravity and all the things [406:05] organization uh have done this notorious [406:08] thing that they changes the limits of [406:10] the plans. So like organizations like [406:13] anthropic have this premium models but [406:16] eventually this organization are going [406:18] to have monopoly. So they have right [406:20] like we talk about uh that graph that [406:23] you show that uh the co there is a cost [406:26] of investment but uh are we going to [406:29] make profit out of it? So if AI is going [406:31] to be in this thing there is going to be [406:32] a investment and if we are not uh [406:35] developing something making profit then [406:37] it is not worth it. So let's say [406:39] tomorrow anthropic creates something [406:41] that uh increase the price. So one of [406:44] the solution that I believe is like [406:46] local models is going to be the part of [406:49] a future. But I want to take your [406:50] opinion like I personally use Quain 3.5 [406:54] coder in a very low-end PC. [406:56] >> It works and even Google has just [406:58] released one app uh Google AI age in the [407:01] play store. So you can download the [407:03] local model in there and um my phone has [407:06] good GPU so I just ran it and it work [407:09] smoothly in the 2GB 3GB model. So is [407:12] this future or what's your take on this? [407:14] So these are my two questions. [407:16] >> Okay, good questions. Thank you. Let me [407:18] answer the second question first before [407:20] I forget the first question. Um [407:24] everybody is investing billions of [407:26] dollars into AI. [407:28] Will they want some return on it? [407:30] Obviously. [407:33] Will they start charging for it? I mean, [407:35] they're charging for it already, but [407:37] they're not charging enough. So, will [407:38] they have to raise prices? Yes. [407:41] Will we have to pay for it? Yes. If you [407:44] want to use it, you'll have to pay for [407:45] it. Do you currently pay developers? [407:53] So, if some AI tool is doing that job, [407:56] it's just replacing somebody, right? And [408:00] these kind of [408:02] things will keep on happening because [408:04] see the other thing is that people want [408:05] monopoly because they know this is like [408:07] you know the winner whoever is the [408:09] winner will have the largest market [408:11] share. So good thing for us is that [408:12] there are so many people competing. [408:15] Uh the bad thing is also the same that [408:17] you know if somebody wins and then we [408:19] will be kind of captive to that if we [408:22] have just gone used to it. So your [408:24] option of you know doing some things [408:26] locally is very good but that won't work [408:29] for everybody. [408:32] Correct. So a large majority of the [408:36] population will start using some or the [408:38] other AI just like that. [408:41] Yeah. So that's kind of my take on that. [408:44] Okay. Coming back and linking this to [408:46] the first question. [408:50] You may be the best artist, [408:53] but you need an appreciator of that art [408:57] to really value that work. [409:01] General population is not that [409:03] appreciator. [409:05] And most of the population is probably [409:07] not even going to be able to afford it [409:08] even if they wanted to. [409:11] Correct? So what people really think [409:13] about is does it work for me? Does it [409:17] get my job done? If it does, they'll do [409:21] it by all means and they have the right [409:22] to do it. So, we can't be purists [409:25] thinking that okay, you know, this is [409:27] the purple tailwind obviously [409:29] recognizable design that why should they [409:31] do that? Oh, well, they should do it [409:32] because it works for them. If it works [409:35] for them, yes, by all means do it [409:38] because it enables people to fulfill on [409:41] their dreams. [409:43] Now, okay, yes, there is going to be a [409:44] time when everything is going to be [409:46] average. That's what I talked about, [409:47] right? Like if everything is average, [409:49] every website looks purple and interfont [409:54] and what where is the you know way to [409:57] differentiate one brand from the other. [409:59] Yes. Then there'll be a comeback [410:01] >> of the artists maybe. Let's hope for [410:04] that. [410:04] >> Okay. [410:05] >> Yeah. Thank you. [410:06] >> Thank you. [410:07] >> Thank you everyone. We are taking on [410:09] time. If you have any questions, you can [410:11] take that offline. Uh thank you for the [410:15] great session. would like to give you [410:17] something. [410:17] >> Yes. Thank you. [410:45] Thank you everyone. Uh we will we are [410:47] starting in some time. Please uh be here [410:51] connect with someone new and share your [410:53] insights. [411:25] I don't think [411:37] Hey, hey, hey. [412:10] 49. [412:22] Heat. Heat. [413:11] Heat. Heat. [414:17] Heat. Heat. [415:39] Heat. Heat. [418:39] Heat. Heat. [419:04] you. Nah. [419:10] Heat. Heat. N. [419:20] Heat. Heat. N. [419:57] Heat. [420:15] Heat. [420:33] feel. [420:41] Hey [421:44] Hey, hey, hey. [423:00] So our next speaker is WordPress core [423:03] committer principle. [423:05] uh software engineer at Blue Host. He [423:08] has been coding since years helping the [423:11] mentoring the contributors and making [423:14] WordPress a welcome to everyone. Uh in [423:17] his talk he will uh explain or share his [423:20] thought how automation is not just a [423:23] good to have uh it will be an investment [423:27] how it can be an investment to WordPress [423:29] or opensource [423:31] uh as an investment. Uh, please welcome [423:34] Jonathan on the stage. [423:38] >> Thank you. [423:51] I don't have the time. I wish I had the [423:54] time. [423:55] Everyone in this room has said this [423:57] before or some variation of it. Some of [424:00] you probably said it today or maybe this [424:02] week. [424:03] Time is our most precious resource and [424:06] it's something we can't replenish and [424:08] nobody knows how much of it we have [424:10] left. [424:12] So even as you sit here, the time you're [424:14] spending in this room is not [424:15] replaceable. So I'm thankful you come [424:17] here and I hope that you get something [424:19] valuable out of it. But have you ever [424:22] considered how much time you're wasting? [424:26] How much your team is wasting your [424:28] community? Whenever I find myself saying [424:31] some variation of this, I try to take [424:33] inventory on this and think about where [424:35] my time is actually going. [424:41] There's a running joke with me and my [424:43] colleagues where uh when I submit a work [424:45] camp talk, the joke is which book which [424:48] principle from this book am I going to [424:49] focus on this time? And this is a book [424:52] if you're not familiar by Carl Fogle. [424:55] And it was published in 2005. But [424:57] despite it being over 20 years old, I [425:00] still find it to be one of the best uh [425:03] books and resources on what it means to [425:05] start and maintain or participate in an [425:08] open source project today. [425:10] So I I highly recommend that you take [425:12] the time to read this u or if you [425:15] believe in the joke, you can go and [425:16] watch my past talks to learn more. [425:19] Myself and other core committers, we [425:21] frequently reference this book when [425:23] we're trying to decide how to handle [425:24] certain situations. [425:26] This is especially important when [425:28] there's no precedent or the situation [425:30] that we're in has never happened before. [425:32] The decisions that we make will help [425:35] guide the project in the future and will [425:36] be used as justification going forward. [425:40] And so it's important that we understand [425:42] all the decisions we make from not just [425:44] a technical but also a community [425:46] perspective. [425:49] I recently reread the book while [425:51] participating in a book club with some [425:53] other members of the community and it's [425:55] always surprising to me how many new [425:58] details I find when I read through this [426:00] book. I walk away from here from every [426:04] time I pick it up, I walk away with some [426:06] new takeaways and some new ideas based [426:08] on where my focus is or what I'm working [426:10] on at the time. And so there were [426:13] several things that stuck around with me [426:15] this time reading it. And today we're [426:17] going to talk about one of the [426:18] principles in the book. Um I am a builds [426:22] and test tool component maintainer. And [426:24] so I spend a lot of time thinking about [426:27] um the efficacy and the efficiency of [426:29] our our workflows and our our tooling, [426:31] the sustainability and maintainability [426:33] of them. [426:35] And so the automation ratio is something [426:37] that [426:39] usually stands out but a little bit more [426:41] this time. [426:42] And so a quick breakdown of this is that [426:45] um it's a rule of thumb that automating [426:47] a common task is worth at least 10 times [426:50] the effort a developer would spend doing [426:52] that task manually once. And often times [426:56] when it's a very complex task that can [426:59] very easily go up to 20 times or even [427:01] higher. It's very easy to be wrapped up [427:04] in low-level work and we can fail to see [427:07] the big picture where we realize that [427:10] everybody is just wasting a lot of time [427:12] on something that can be automated. [427:16] Even those who do realize it, they're [427:18] not annoyed enough or they're not [427:19] burdened enough to actually take the [427:21] time to automate the task and they just [427:24] keep going through the motions and and [427:26] doing these these tasks that, you know, [427:28] are not really a great use of our [427:30] attention and time. [427:35] He writes about what makes this the most [427:36] compelling thing to him is that it's not [427:40] just automating the task. It's the time [427:43] to perform the task once times the [427:46] number of times that somebody performs [427:48] that task and then multiply that by the [427:52] number of people who actually perform [427:54] that task. And when you perform all that [427:57] math, you end up with essentially the [427:59] total potential time saved by automating [428:02] that task. [428:06] So at the beginning, you may not realize [428:09] uh that benefit because there's a very [428:12] often large initial investment in [428:14] automating something. You have to spend [428:16] the time researching and implementing [428:17] and testing uh that automation to make [428:20] sure it's what you need. But it doesn't [428:22] stop working. It continues to work. It [428:24] keeps running every release cycle, every [428:27] time a pull request is open. However, [428:29] it's configured to run, it will just [428:31] continue to run most of the time. We'll [428:34] talk about that a little later. But [428:36] these returns compound. And so in month [428:38] one, you may not realize that [428:40] investment, but in month 20, you'll have [428:43] had a huge return on that investment. [428:50] across the WordPress project. GitHub [428:52] actions are the primary tool that we use [428:54] to automate tasks and um whatever we [428:58] have that needs to be automated. There's [429:00] currently 459 files with automation [429:03] logic related to GitHub actions in all [429:06] of the WordPress repositories. It [429:08] represents 433 unique workflows or [429:12] automations across 113 repos with 651 [429:16] jobs and almost 3,000 individual steps [429:19] performing a single command or action. [429:28] One example of this and some examples [429:30] that everybody should look into adding [429:32] is unit testing. So unit tests are a a [429:36] good way to automate confirming that the [429:39] desired outcome is uh produced when you [429:43] run certain steps in your software. In [429:46] the WordPress or uh in WordPress develop [429:48] specifically the main WordPress [429:49] repository, there's 3,441 [429:53] assertions that are performed in the [429:55] test suite. [429:57] That's divided up over 29,000 tests. And [430:00] if we assume that each test method takes [430:02] three minutes for someone to perform [430:04] manually, it would take a single [430:05] individual over 60 days working non-stop [430:09] to run all of these tests. [430:12] So the reality would be that no one [430:15] would ever run these tests and they [430:17] would be very unclear when they were [430:19] run. Um they'd be very subject to error [430:21] and variance because different [430:24] environments are uh various environments [430:27] are different. there would be [430:28] inconsistencies and it would be very [430:30] hard to aggregate these results to [430:32] understand when specific problems were [430:34] introduced. [430:38] Another example is u a tool in PHP [430:41] that's called PHPCS which is basically [430:44] to ensure consistent styles in your code [430:47] across the entire codebase. And so this [430:50] is really beneficial because it makes [430:51] your codebase easier to read to [430:54] understand. It's more approachable. more [430:56] people will contribute to your project [430:57] overall. You're more consistent and it [431:00] can also make it easier to write tests. [431:06] Another good example of a common [431:08] automation is uh dependabot on GitHub. [431:12] And so in WordPress develop there are [431:16] 2600 dependencies defined just through [431:20] mpm and node. And it's impossible to [431:23] manually track these even monthly. [431:25] Someone would have to manually go to [431:27] each one or subscribe to each repository [431:29] for when new releases are published. [431:31] Each dependency may not use GitHub. Some [431:33] may use some other repository management [431:35] software. And so it really would [431:38] probably not take not as long as running [431:40] those unit tests, but it still is a [431:42] significant time effort that you need to [431:44] consider. [431:48] There's also many tools that help [431:50] prevent introduction of security [431:52] vulnerabilities. And so some of these [431:54] are actually a part of the coding [431:56] standards workflows as well, but there [431:59] are separate ones. And the idea here is [432:01] that you will be notified of [432:04] vulnerabilities within your software, [432:05] but also you can set it up so that [432:08] you're notified when you will introduce [432:09] them. And so they never actually make it [432:11] into your software because the pull [432:14] request is flagged and you have to fix [432:16] that before merging. [432:20] There's also many hidden benefits to [432:22] investing in automation and these are [432:24] second or third degree impacts that are [432:27] not very obvious especially when you're [432:29] getting started with automation. [432:37] One of my favorite things about [432:38] automation is that it forces you to be [432:41] intentional in how you automate [432:43] something. [432:45] You can't take something that you don't [432:47] understand or know how it functions and [432:50] write something to that automates that. [432:53] You have to take the time to fully [432:54] understand it so that you can ensure the [432:57] right situations are tested, the right [432:59] controls are in place and that you're [433:01] actually having the desired outcome. [433:07] Often times that results as well in [433:09] higher quality code. So many times when [433:12] you write unit tests, you discover that [433:15] your functions, you discover that your [433:17] functions are not actually responsible [433:20] for a single thing, also known as the [433:22] separation of concerns. And so you would [433:25] spend some time refactoring them. It [433:26] would make it easier to test. And that [433:29] automation, investing in that automation [433:31] makes it far easier for you, uh, makes [433:34] your quality of your code much higher [433:35] than it would have been if you had not [433:37] automated those tests. [433:41] When you understand the automation and [433:43] how it works, one of the benefits is [433:45] that you eliminate randomness. And this [433:48] is also known as being deterministic. So [433:50] every time that workflow runs, if the [433:52] conditions are the same, you will get [433:54] the exact same result. If a line of code [433:56] changes, it might be different for the [433:58] first time, but then subsequent runs [434:00] will then be consistent with that [434:02] previous one. [434:07] When you understand how to how to [434:09] structure your automation, a lot of [434:12] times somebody can read that automation [434:14] file and they would understand what the [434:17] purpose is, what it's trying to do, and [434:19] how it's doing it. And that's just [434:21] because by nature, a lot of automation [434:24] is structured through maybe YAML files [434:26] or configuration files and it just makes [434:29] it easier to read. But you have to be [434:32] careful to name things correctly and [434:34] organize them in the right way. And so [434:36] it does still take a bit of care to make [434:37] sure that you realize this benefit. [434:44] Another great benefit of this, [434:45] especially in a uh project like [434:47] WordPress, is that there's lots of uh [434:50] there's lots of reasons why [434:53] code is in the condition that it's in at [434:55] this point today or the given point when [434:58] you add automation for a specific thing. [435:01] And oftentimes as contributors [435:03] throughout the years move on or take on [435:05] new jobs or they um you know retire then [435:11] that that is very hard to preserve in [435:13] some ways. A lot of times we encourage [435:16] our core committers to blog more or to [435:19] um be more detailed in their comments or [435:21] to add pages to our core handbook as [435:23] ways to document that knowledge that we [435:26] have and those shared uh the shared [435:28] learnings from different initiatives. [435:31] But automation is also another great way [435:33] to codify that and preserve that for [435:36] contributors in the future and ourselves [435:38] when we return to things. I can't tell [435:41] you how many times I return to code and [435:42] I look at who made the change and it's [435:44] me and I think what was I thinking or uh [435:47] I don't remember what I was thinking and [435:49] you know reviewing the automation can be [435:51] very helpful to jog my memory or [435:53] understand what we were trying to [435:54] accomplish. [435:59] So what does this mean in the context of [436:01] open source projects? [436:03] Open source projects are very uniquely [436:05] resource constrained. Oftentimes [436:08] maintainers are volunteers. They're [436:10] partially sponsored. Some of them are [436:13] fortunate like myself to be fully [436:14] sponsored, but there's very few. They're [436:17] juggling many, many things. We have in [436:21] such a large project like WordPress, [436:23] there's many things that need to be [436:24] considered, taken care of. and when [436:26] they're not taken care of, our users can [436:28] suffer. [436:30] This can be juggling bug reports, making [436:32] sure releases are in time, making sure [436:35] there's user documentation and developer [436:37] documentation. [436:38] But it can also be things like community [436:40] management and actual code [436:42] contributions. [436:44] Essentially, maintainers are just [436:46] drowning in this mechanical repetitive [436:48] work. They don't have the time. [436:53] I've been playing around with this [436:54] concept recently where I picture the [436:58] project as a diamond and despite my best [437:01] efforts I couldn't get chatt to ignore [437:04] diamond meaning the shiny gem that [437:06] everybody likes but is expensive but if [437:09] you think of like the rhombus style [437:11] diamond and you make a line in the [437:13] middle you you separate the sides and I [437:16] think that that's kind of where we're at [437:17] right now with the WordPress project. We [437:20] have a ton of people in the middle that [437:21] know how to contribute. They spend their [437:23] time contributing, helping to organize [437:25] events like these. We have at the top, [437:28] we have maintainers like core committers [437:30] and project leadership. And the funnel [437:32] is closing up there because there's so [437:34] much to do and there's not enough of [437:36] them. And at the bottom, it's restricted [437:39] because it's not easy to get involved [437:41] contributing to WordPress all the time. [437:44] There's a lot going on. There's a lot to [437:46] consider. And it's like I I equate it to [437:50] jumping into a moving train. Often if [437:53] you don't have someone there to guide [437:54] you, see the maintainers and and [437:57] experienced contributors who don't have [437:59] the time to do everything that needs to [438:01] be done, then you might not get new [438:03] contributors at all. And we depend on [438:07] this. We need new contributors to look [438:10] out for WordPress for years to come. [438:17] And I I another way to describe this is [438:19] your bottom funnel is how easy is it to [438:21] get started. How easy is it for you to [438:24] use and understand the software but also [438:26] contribute back to it. [438:28] And the top funnel is all about freeing [438:30] maintainers and having enough so that [438:33] they can focus on what matters. Making [438:35] the decisions that need to be made, [438:37] dealing with people. All of these things [438:39] can't be automated. [438:44] Automation doesn't necessarily make [438:45] those things disappear though. What [438:48] actually happens is they just move to [438:50] the background. [438:52] The goal should be to free the focus of [438:54] the humans that we need to focus on hard [438:57] problems and those actually require [439:00] judgment. [439:01] These are design decisions, how to best [439:04] architect things, community health and [439:06] of course mentorship. [439:11] There's another aspect to the automation [439:13] ratio that really stuck out to me when I [439:15] read it this time through. Um I it had [439:18] probably been a year or so since I had [439:20] read this chapter, probably two. And I [439:23] think things become a little bit more [439:25] interesting today when we look at how he [439:27] explained he was approaching the concept [439:29] of automation in the book. He says that [439:32] it should broadly be considered as any [439:35] sort of technical infrastructure that [439:37] assists humans. [439:39] And it made me realize that this [439:41] automation ratio also applies to uh one [439:45] of the most single important [439:46] technologies of our lifetime. [439:51] At Word Camp Canada, I spoke about how [439:54] we should approach integrating new AI [439:56] tools into our community. People come to [439:59] our community because they find a home. [440:01] They feel like they belong. They have [440:02] experiences. They make friends for a [440:05] life. And if we overimplement AI, then [440:09] that goes away. And our community is not [440:11] as attractive to participate anymore. [440:14] But we're naive to say that AI is not [440:17] reshaping how we collaborate, how we [440:19] make decisions, and how we maintain [440:21] trust, even in uncomfortable ways. Often [440:25] this is especially true in large [440:27] projects like WordPress. [440:29] So, how do we utilize AI without [440:31] eliminating these low friction paths for [440:34] new contributors? If we overautomate, [440:37] then there's less ways for people to get [440:40] involved. [440:42] How do we remain deliberate as we're [440:44] trying to solve problems and implement [440:46] faster? We do need to move faster. We [440:49] need to keep up with these new models. [440:51] Um, this week there was an announcement [440:53] that Enthropic has a new model that [440:56] they're not going to give to anybody [440:57] because it's been finding so many [440:59] security vulnerabilities and we have to [441:02] find new ways to prepare ourselves for [441:04] how this may affect our project in our [441:06] surrounding ecosystem. [441:09] But a part of as a part of that talk at [441:11] work Canada, I went through an exercise [441:13] looking at the widely accepted [441:15] definitions of some of the related [441:17] concepts like language model or large [441:20] language model. [441:21] And what really hit me in the face was [441:23] just how much AI is at the core just [441:27] math. [441:29] And math is very easy to automate [441:31] because it's very deterministic. You get [441:33] the same outcomes when you add the same [441:35] numbers together every time. And so AI [441:38] is essentially just another form of [441:40] automation. [441:42] So a lot of this automation ratio and [441:44] this concept I I found myself applying [441:47] it to how we're implementing AI. [441:50] And I'd argue that the original ratio [441:52] that Carl Fogle discussed is likely far [441:55] higher now and probably more like 100 to [441:58] one or a thousand times. [442:03] Like anything though, it's not always [442:05] smooth sailing and there are some things [442:08] that are very important to be aware of [442:09] as you begin to explore automation and [442:11] implement it. [442:14] With any tool or process that you [442:16] implement, there's always potential [442:18] costs associated with that. These are [442:20] obvious and nonobvious. So sometimes you [442:22] need to go looking. [442:24] In the case of GitHub actions, as an [442:26] example, they are free for public [442:28] repositories, but there's no guarantee [442:31] that that will remain free forever. [442:34] Actions for private repositories incur a [442:36] permanent cost of a fraction of a cent [442:38] each. [442:40] from the previous unit of test example. [442:43] The difference between this bill and the [442:46] cost it would would take to have an [442:48] engineer running tests full-time and [442:50] constantly evaluating the software still [442:53] makes this worthwhile even if all your [442:55] repositories are are privates most of [442:57] the time. However, if you implement [443:00] automation poorly, you can inadvertently [443:03] introduce very high volume and it could [443:06] be very costly at the blink of an eye. [443:09] Think contributor day or before a [443:12] release deadline when many people are [443:13] trying to get final changes in or try [443:17] their hand at at contributing. [443:19] We could you could very quickly rack up [443:21] a very very expensive bill. [443:26] Over time, automation will become one of [443:29] the very uh critical foundations of your [443:33] codebase. [443:34] You wouldn't buy a car from a company [443:36] that ceased to exist over 10 years ago. [443:39] So, you shouldn't find yourself in a [443:41] situation where you choose to use a tool [443:43] that has been abandoned and is [443:44] unmaintained. [443:46] There's no recourse when a bug is [443:48] discovered or you need a new feature. [443:51] You would be forced to dive in and adopt [443:53] that tool on your own, which would [443:55] lessen the the benefit of the investment [443:58] you had on that particular automation. [444:05] Likewise, you should always be careful [444:08] about the tools that you choose. Even [444:10] when they are appear to be active, that [444:13] may not always be the case. You should [444:16] also be cautious of proprietary tools or [444:18] any that rely on non-standard formats. [444:22] It's not clear on the onset, but these [444:24] can also be very expensive to you. And [444:27] this cost is not fully realized until [444:29] much much later when you outgrow a tool [444:31] or you choose to migrate to a new one. [444:34] So look after your future self. And when [444:36] you're evaluating the tools that you [444:38] want to use, don't just ask how easy it [444:41] is to adopt the tool now. You should [444:43] also look at how hard it would be to [444:45] move away from that tool in the future. [444:51] There's been a lot of research that [444:52] shows people tend to follow the advice [444:55] of a computer more than a human. And we [444:59] don't really understand this yet. Um, [445:01] and this is even true when the advice is [445:04] identical and when the imperfections in [445:07] the AI or the computer are made [445:10] explicit. [445:12] It's very important to [445:15] avoid burying the results of your [445:17] automation. [445:19] The results should be loudly broadcast [445:22] so that when something is wrong, people [445:24] take note and they address it. [445:26] But you don't you want to find a balance [445:29] because too much noise will result in [445:31] everybody just ignoring it because oh [445:33] it's that silly automation that's always [445:36] failing for no reason, right? So too [445:38] much noise and false positives. It it [445:41] just equals a learned behavior of [445:42] ignoring it entirely and that decreases [445:45] that benefit of that investment that [445:46] you've made in automation. [445:51] One of the barriers in the context of [445:53] WordPress would be making it harder to [445:55] get started at all. [445:58] With the previous example, if you are [446:00] not reporting results in a way that [446:02] people can understand them, then it [446:04] becomes difficult to know why it's [446:06] failing or what you need to adjust in [446:08] order to get it to pass and be [446:09] acceptable as a contribution. [446:13] And as I stated before, we want to find [446:15] ways to open that funnel and make it [446:17] easier for people to get involved. [446:23] Tools also reflect the bias of the [446:26] people that build them. These could be [446:28] simple things like your preferences and [446:30] tooling or how you think somebody should [446:32] be using a specific type of automation. [446:36] But it could be more complex and more [446:37] dangerous like cultural or economic [446:40] bias. [446:42] For example, perhaps a specific tool [446:44] requires a lot of resources to run and [446:47] it's expected that that tool is run many [446:50] times throughout the course of uh [446:52] building something. [446:55] This could result in it being very [446:57] expensive and people maybe in developing [447:01] countries who have less powerful [447:02] hardware would not be able to run it as [447:05] much or it's more costly to them because [447:07] they don't have as good internet. [447:09] And so those are things that we always [447:11] need to be conscious of and make sure [447:12] that we're not excluding people through [447:15] our unintentional bias in our tools. [447:24] As I was making this talk, I realized [447:25] that a good analogy to automation is the [447:29] game of chess. [447:32] In chess, there are good moves, there [447:34] are bad moves. There's a seemingly [447:37] infinite number of strategies. [447:40] You can recover from most bad mistakes [447:42] with enough skill and and thought, but [447:46] you can also blunder a very good [447:47] position to the point of failure. [447:51] In automation, there's a mind-numbing [447:53] number of tools and libraries that you [447:55] could use to accomplish the same task. [447:58] There are multiple strategies that you [448:00] could use to solve problems or surface [448:02] the results or make them actionable for [448:04] your your community. [448:08] And there's others that introduce [448:10] crippling amounts of technical debt or [448:12] costs that you'll have to be will be a [448:15] burden to you in the future. [448:20] Ultimately, automation and AI cannot [448:22] solve the people problems that we deal [448:24] with, especially in a large community [448:26] like WordPress where the human element [448:29] is very important and valued. [448:32] The goal should never be to remove [448:34] remove human judgment. [448:37] We shouldn't aim to remove people [448:39] entirely from the equation. [448:42] Instead, we should be able to easily [448:43] override it or interject as necessary. [448:49] Good automation enables people to do [448:51] their best work and it focuses on the [448:54] work focus them on the work that humans [448:57] are uniquely uh empowered to to handle [449:00] and qualified. [449:03] It can make your code more enjoyable to [449:04] read, can be easier to become involved. [449:08] It can grow your project. [449:11] In the context of that diamond analogy, [449:14] the best automation widens both funnels. [449:17] It doesn't do so at the disadvantage of [449:21] another group. Doesn't make things [449:22] easier for maintainers but harder to get [449:24] involved. Everybody should be considered [449:27] when you when you work through [449:28] automation. [449:29] And sometimes that can take extra time [449:32] because you need to understand who your [449:33] community is. How are they using your [449:36] product? How are they contributing to [449:37] you? [449:38] What are the various tools that people [449:40] are using? [449:43] In some ways, automation can be a path [449:45] to freedom. It can free your time and it [449:49] will allow you to say, "I don't have the [449:51] time." Instead of saying, "I don't have [449:52] the time." [449:54] I want you to ask next time, what am I [449:57] wasting time on? [449:59] What are even the small things that I'm [450:01] doing repeatedly that doesn't require [450:04] any special skill or doesn't really [450:06] require my experience or my critical [450:09] thinking to come up with a good answer [450:13] and then of that I want you to consider [450:15] how can I reclaim some of my time my [450:18] company's time contributor time or even [450:22] within my communities [450:25] I want you to do that and then make an [450:27] effort to tackle something you've always [450:30] wanted to do but felt that you didn't [450:31] have the time. Maybe that's learning a [450:33] new language or a new technology or [450:36] contributing back to WordPress more. And [450:38] then hopefully you report back to me and [450:41] say what ways you've automated your [450:43] workflows in your life and you realize [450:46] that there's so much more that you can [450:49] focus on that's more worth your time. [450:52] Thank you. [451:02] Thank you Jonathan. Uh it was a very [451:05] powerful insight about the automation. [451:07] Being a QA engineer I can relate with [451:09] the topic and yeah it's also important [451:12] to keep a human in loop that is the very [451:14] important point. We do have some time if [451:17] you have any questions there is a [451:19] microphone provided in the aisle. Uh we [451:22] have we can take few questions. [451:33] No questions. I You are all motivated to [451:35] go reclaim your time. [451:39] >> Okay, then. [451:40] >> All right. Thank you. [451:42] >> Thank you, Jonathan. We would like to [451:43] give you a gift. [451:45] >> Sure. [451:50] >> Thank you. Thank you. [452:18] We are starting the next sessions in few [452:20] minutes. Uh please be seated and please [452:24] consider talking to everyone and gra [452:26] maybe grab some te. Thank you. [455:07] out. [458:41] Hey, [458:51] hey, hey. [459:52] everywhere. You [460:16] Hey, [460:22] hey, hey. [460:51] Heat. Heat. [461:25] feel. [461:33] Hey [465:26] Hey. [465:30] Hey. Hey. [466:06] I [468:04] Oh [468:17] my god. [468:47] Hey. Hey. Hey. [470:38] Down. [472:02] Hey everybody, hey [472:32] The program is about to begin. Please [472:35] take your seat and silence your mobile [472:36] phones. [473:10] Hello again everyone. Thank you for [473:12] being here and I hope everyone is [473:14] enjoying and catching up more insight [473:16] from different session and enjoying the [473:18] session so far. Up next is a panel [473:21] discussion that will be AI panel [473:23] discussion and the topic will be how AI [473:26] is restructuring traditional and modern [473:28] marketing methods [473:30] and in the panelist we have Himmani, [473:32] Julian, Karishma and Si. I I would like [473:36] to introduce the moderator, Alexander, [473:39] um for this session. [473:44] He's based in Tokyo. He's a partnership [473:47] and community manager for APAC Instin. [473:51] Please join me to welcome Alexander, our [473:53] moderator. [473:55] >> Thank you. Thank you for that [473:56] introduction. Well, let's uh have our [473:58] panelists come on stage. [474:10] Yeah. Uh, please welcome Himmani on the [474:13] stage. [474:16] Then we have Julian. [474:25] Next we have Karishma. Should we follow [474:27] her? [474:30] >> Please welcome Sundep on the stage. [474:39] Okay. Hey everyone, how's day one of [474:43] Word Camp Asia in Mumbai? [474:49] >> Thanks very much to our organizers as [474:51] well. This has been a ton of fun. Uh I [474:53] hope you've all had a chance to see some [474:55] of the sessions, to visit the sponsor [474:56] booth. The food has been pretty good [474:58] too. Um, as thank you for that kind [475:02] introduction. I think we'll do just a [475:03] quick introduction of who's standing [475:05] here before you. Uh, my name is Alex. [475:07] I'm originally from Los Angeles, [475:09] California. So, from Hollywood to [475:10] Bollywood. [475:14] Thank you. Thank you. Uh, I'm based in [475:17] Tokyo, Japan. I have lived in Japan [475:19] about 11 years. I was in Singapore two [475:21] and a half years. Uh, I currently work [475:23] for Kinsta. Uh, I've been at Kinsta for [475:26] about five and a half years. and my [475:28] current role is partnerships and [475:29] community manager for all of Asia [475:32] Pacific. So I manage Kinsta's uh [475:34] partnerships with agencies and uh [475:36] interacting with the community, [475:38] representing the company at events. It's [475:39] a lot of travel and I've have the great [475:42] fortune of uh getting to know so many [475:44] people in the community connecting and [475:47] seeing what people are talking about, [475:49] partners, vendors, competitors, what's [475:52] going on in the industry in WordPress [475:54] and beyond. So recently I've attended [475:56] quite a lot of events in various [475:58] countries in Japan, in Singapore, [476:00] Australia. I've been to uh most of the [476:03] flagship word camps. Uh so Manila, [476:06] Taipei, uh I was in Basil, Torino, uh [476:10] word camp US. So it's a lot of travel [476:12] and I have the great chance to see what [476:14] people are talking about and AI is on [476:16] everyone's mind. So that is the topic of [476:18] today's discussion and if I could ask [476:22] our panelists to introduce themselves [476:24] please. [476:25] >> Thank you. Thank you so much. Uh I'm [476:27] Himmani Kankara and I am from uh [476:30] Ahmedabad. Uh and uh I run an organic [476:34] marketing agency which is missive [476:35] digital and we work with um B2B SAS and [476:38] tech companies all from early stage [476:40] startups to unicorns and super unicorns [476:43] as well. And uh our core strength has [476:46] always been uh uh you know the content [476:50] content stuff where we actually utilize [476:53] the customers stories from the product [476:55] and the sales team and present it to [476:58] them itself so that they can learn [477:00] better and they can build better brands. [477:03] So yeah that's how it is. [477:05] >> Oh thank you. [477:06] >> Hi. Hi. I'm Julian Song from Malaysia. [477:09] Um I run the agency in in this Malaysia [477:12] and uh I handle client from small to to [477:16] enterprise. [477:19] >> Hi, I'm Sadep from Bangalore. I run a [477:22] marketing agency and um I've been [477:25] building tools with AI over the last uh [477:28] 6 to 12 months and uh yeah looking [477:31] forward to sharing some of those stories [477:34] here today. [477:35] >> Hi, I'm Karishma. [477:37] I'm I think the only local on this [477:40] panel. I'm from Mumbai. Um I have headed [477:45] uh content teams with mostly IT [477:47] companies, SAS companies. So uh AI has [477:51] been a mainstay for the last two and a [477:53] half years or so and I'm excited to talk [477:55] about this on the panel today. [477:58] >> Awesome. Thank you very much. Well, so [478:00] AI has been on everyone's mind if you [478:03] looked at the schedule for today and for [478:05] tomorrow. um AI is kind of dominant and [478:09] this has really been a theme in events [478:11] globally. Uh but I've really seen a [478:14] shift in a lot of the events that I've [478:15] been attending where last year it's been [478:18] a lot it's a lot of discovery, a lot of [478:20] experimentation, a lot of questions, a [478:22] lot of listening. What can we do with [478:25] AI? What is possible with AI? What is [478:27] the current state or status of AI tools? [478:31] And it's so different now even compared [478:33] to six months ago. We were talking [478:35] earlier when we were preparing for this [478:36] panel that now the conversation has [478:40] shifted where it's not so much about [478:43] experimentation but people really want [478:46] value out of these tools. Um there's [478:48] been a shift I think in global uh [478:51] corporate [478:53] situation uh that people are doing more [478:56] with less. They want to see what can we [478:58] actually do with AI tools. And so I'd [479:00] love for today's conversation to really [479:03] focus on practical real life [479:06] applications, case studies, what is your [479:09] actual experience? And then we'll open [479:11] it up to questions at the end. So rather [479:13] than big sandbox ideas, we'll have some [479:17] big ideas of course, but let's get deep [479:19] into what's real because there's kind of [479:24] a conflict. I think we as we were [479:26] discussing earlier. So amid this shift [479:30] where everyone is looking for [479:32] some reality perhaps but some value to [479:36] be certain [479:37] as a quick icebreaker question to the [479:39] whole panel. What is something that you [479:42] have started to do differently [479:44] in the last six months [479:49] with AI in your work uh as it pertains [479:51] to marketing to WordPress to martekch? [479:54] um what are what is AI doing in your [479:57] daily life in your workflows what are [479:59] you doing differently [480:00] >> so I think uh you know the first thing [480:03] that we automated using AI is the review [480:06] process since uh uh you know content has [480:09] always been a core for us and that has [480:11] helped a lot of uh of our clients build [480:14] their entire uh success and the growth [480:16] journey right and in that case uh what [480:20] was taking a lot of our time was the [480:22] review manual review process [480:24] Uh so what we did is that uh we have a [480:27] 368 pointer checklist that reviews uh [480:30] the compliance part of the content the [480:32] accuracy the fact check and all of those [480:34] stuff. So for that we created uh you [480:36] know an AI agent internally and now we [480:39] review most of our part which was taking [480:41] earlier uh you know an editor's time [480:44] like say 9 to 10 hours maybe and now [480:46] that has reduced to just 10 minutes. So [480:48] what has happened is now every writer [480:51] knows what the editor or uh you know [480:52] what the SEO manager is going to expect [480:55] out of a particular content and then [480:57] they check it themselves they score it [480:59] themselves and then they optimize it in [481:01] a way so that you know the manual to and [481:04] fro the optimization or the you know [481:08] proofreading and rewriting has reduced [481:10] quite a lot. I'd just like to highlight [481:12] something you said Himmani. Um you [481:14] mentioned that this process in the past [481:16] was taking nine to 10 hours and now it's [481:19] 10 minutes. [481:21] >> Yeah. Exactly. So what has happened is [481:22] that we have three layer review process. [481:24] So there is a review that happens from [481:26] the you know uh critical or the [481:28] technical side of the content. One [481:30] review is that second review that [481:32] happens which is from the SEO and the [481:34] branding point of view and then the last [481:37] review happens from the brand reputation [481:39] point of view. Right? So u then we have [481:42] that different processes and the [481:44] different people who does this checks. [481:46] uh since for example if you talk about a [481:48] content which is into fintech space [481:50] right it is very critical to get your [481:53] content approved by the compliance team [481:55] of a company like uh you know like say [481:58] Phone Pay or any other such large uh uh [482:00] fintech companies right in that case we [482:04] always have uh the accuracy of getting [482:07] that approved in one go right and that [482:10] is the reason why we had to always [482:11] manually review the content a lot more [482:14] times I'd like to highlight that as well [482:17] if I may. The manual review aspect. This [482:19] is something that I think we're going to [482:20] keep coming back to in this discussion [482:23] >> is what is the human element when we're [482:25] using AI tools and how important is [482:28] that? Where does it come in? [482:29] >> So, we'll come back, but I want to just [482:31] continue with the icebreaker question if [482:32] I may. Uh Julian, something that you've [482:34] been doing differently in your [482:36] workflows. [482:36] >> Oh, okay. Because we when we receive the [482:39] lead from clients kind of things, we [482:41] when we prepare proposals. So in the [482:43] past we always do a lot of research and [482:45] write a lot of proposals for the clients [482:47] and now we AI is so much to do those [482:50] proposals so much easier. What we do is [482:52] we actually ask AI to criticize our [482:54] proposal. So ask them to criticize as [482:56] much as possible for us to further [482:58] optimize to to make the proposals very [483:01] solid and to get a much uh win the high [483:06] win percentage. So uh in fact I think it [483:10] works very well. It works very well. [483:14] All right, thank you. And let's continue [483:16] down the line. Um, [483:18] Sep. [483:20] >> Yeah, as a marketer, I didn't expect to [483:22] spend this amount of time in the [483:25] terminal, but I seem to be spending [483:27] almost all of my time uh in any one of [483:31] those uh you know uh CLIs, whether it's [483:35] clot code or CEX building things. Um, [483:38] and now I'm encouraging my team to do [483:40] the same thing. That's another thing I [483:42] think that's going to keep coming up is [483:44] having more time as Simani mentioned as [483:47] well. um what happens with that time or [483:50] what do you do with that time now and [483:53] Julian as you mentioned as well as [483:55] what's that new process looking like u [483:58] as we shift our workflows [484:02] what what are we shifting to in the [484:04] future uh so last if if we could [484:06] continue with the icebreaker karishma [484:09] >> so what we have done is we are a team of [484:12] uh any content team is built of various [484:15] skills whether it's a writer an editor [484:17] an SEO person, somebody who may also [484:20] have some um anme come in for domain [484:23] expertise or what. Uh my goal as [484:27] administrator was to make sure that I [484:30] could make checklists of everybody's [484:33] skills and then program AI to take over. [484:36] So um one key stat that we had is that [484:40] when we started using AI we were doing [484:42] two articles per writer per week and we [484:45] can now shrink that entire process down [484:47] to two wellressearched well thought out [484:50] well-written articles that uh convert [484:53] for us per day per writer. So that's [484:56] like a 5x output right off the bat and [485:01] we're working towards adding more and [485:02] more capabilities as we go forward. [485:05] So the process is often reduced here. [485:08] But what I'm not hearing and I think [485:10] this is very important to mention. What [485:11] I'm not hearing in these introductions [485:13] is that AI is replacing some part of the [485:16] process. It's changing some some part of [485:18] the process. So this is an important [485:21] distinction I think. Um we were [485:23] discussing this in the other room [485:25] earlier about when and how we use AI as [485:28] a tool but not as a replacement. Um, we [485:32] discussed this a lot earlier, but maybe [485:35] as a as a kind of a follow-up question, [485:37] something that you've seen that didn't [485:40] work. And I I mentioned an example [485:42] earlier where I saw an an agency that [485:46] was advertising their services in in the [485:48] US. Um, [485:51] but it was clearly an AI generated slop [485:54] site. And so we're talking about the [485:56] idea of if you are putting out slop then [486:00] that becomes your product. Your product [486:01] is slop. So maybe just a couple of [486:05] examples of something that that really [486:07] didn't work. Do you have a something [486:10] that you've seen that maybe left a bad [486:11] impression? Because I think it's really [486:13] important to to mention this as well. [486:15] Like we were discussing in the other [486:16] room. Anyone from the panel? What what's [486:19] an example that you've seen recently [486:20] that made you think [486:24] like this is not the right way to use [486:26] AI? [486:27] >> Right. Um I can go. Um, we had a little [486:31] bit of a scare uh recently [486:35] um when the AI um agent slash engine [486:40] that that we've built for uh a client [486:42] that we work with sudden suddenly [486:45] flagged the fact that the title tags on [486:48] the site is completely missing. Right? [486:51] this is a large site with 10,000 plus [486:53] pages and I ran the audit came back and [486:57] said all these are okay but none of the [486:59] pages have title tags and that took me [487:02] by surprise but when I could saw this [487:05] website for myself I could see the title [487:08] tags so why is AI telling me this turns [487:12] out that the title tags were inserted [487:14] using JavaScript and AI couldn't read [487:16] JavaScript and so for it basically it [487:20] considered decide to have no title tags. [487:23] >> You have to trust your eyes in that [487:25] moment, [487:26] >> right? [487:27] >> There's no there's no substitute for for [487:29] human experience and I think that is [487:32] especially important to to learn. I I [487:34] think um uh you were also discussing [487:38] earlier about in the content uh creation [487:41] process. [487:42] >> Yes. um AI can't replace our specific [487:47] niche domain expertise especially in [487:49] highly controlled and compliance uh [487:52] related industries. So [487:54] >> we are in WordPress security for example [487:57] and uh I can ask AI to write a article [488:01] maybe on login security and it'll pick [488:04] up 100 200 sources condense the [488:07] knowledge but since there is so much of [488:10] misinformation and there's so much of um [488:15] there are so many things that people are [488:17] expected to do or told to do which don't [488:19] have the value that is associated with [488:22] it. us coming in to say no this is wrong [488:26] and an example of this which is very [488:28] specific to WordPress is changing the [488:29] login URL it's very commoning common [488:32] hardening recommendation that's given to [488:35] improve WordPress security and we know [488:38] for a fact that this is not not good [488:40] practice for many reasons so if I ask AI [488:44] without any guardrails without any kind [488:46] of quality check to write an article [488:48] about login security the first thing [488:50] it's going to say is change your login [488:52] URL [488:53] and us coming in from our background and [488:57] knowing that this is not good and it's [488:59] kind of useless and can cause more [489:01] problems than it solves. Uh we have to [489:03] go in and say no no this is you can't [489:06] write that. We can't have that kind of [489:08] um [489:09] >> misinformation. I think these are great [489:10] examples here and we were all discussing [489:13] earlier in the other room about when [489:15] we're using AI we have to trust our own [489:17] experience as well and there's really no [489:19] substitute for that but also because [489:21] what AI is generating is derivative also [489:26] that information or whatever it's it's [489:28] pulling together its data from is coming [489:30] from whatever it's trained on so it's [489:33] going to be derivative it's not creating [489:35] anything new you always have to have [489:37] that human element checking at the end [489:39] so whether I'm looking at a potential [489:41] agency partner and I'm thinking, does [489:44] this company even exist that this [489:46] doesn't look real? Or whether it's [489:48] content that is not trustworthy, that's [489:50] giving people wrong advice. [489:53] Um, these are examples that really could [489:55] create a negative impression that we [489:57] really have to be careful of. So, I'd [489:59] like to now turn it around to the [490:00] positive side. [490:02] Uh, Imani, you mentioned some incredible [490:05] time saving. Uh Julian, we were talking [490:07] earlier about how conversion rates can [490:10] be so much higher with AI and we'll [490:12] we'll get into maybe uh AEO and GEO as [490:15] well. Um we've seen some negative. We we [490:19] started with some changes. We've seen [490:21] some negative. Now what about the [490:24] positive? What about the value? Because [490:25] I think that's what everyone really [490:26] wants to hear, right? Where can we find [490:29] some great value in using these tools? [490:32] Not just change, but what about the [490:35] promise for the future? [490:37] >> Julian, what what about maybe some [490:39] examples of of clients that that you've [490:41] worked with? [490:41] >> Okay. Um I what I do is like when I [490:44] prevent my proposals, I get AI to [490:47] criticize as uh we there's a something [490:50] called term called murder bot. So based [490:52] on that they will actually criticize the [490:55] proposal criticize whatever my [490:56] plannings. So they actually giving me a [490:59] totally different perspective that I [491:01] never thought of that and I also they [491:03] also tell me that there's a certain [491:05] things that we need to how need to uh [491:08] figure how to make it real or of course [491:10] we need to understand all the context [491:12] they provided. I think the gatekeeper is [491:15] very important. We need to make sure [491:16] that everything the written from this AI [491:19] are we we know how all the context is [491:21] whether usable or not and for that it's [491:25] how [491:27] it improve a lot of our workflows the [491:30] speed and everything and we also learn [491:32] something that we never thought of [491:34] because of the perspectives is from [491:36] totally difference. [491:37] >> Yeah. Yeah. So the workflows can develop [491:41] and you mentioned also Himmani earlier [491:42] about um the human element at the end. I [491:46] wonder if you can explain a little bit [491:47] more about what you mean and how that [491:49] fits into your workflow and why it's [491:52] important to ensure an a positive [491:54] outcome. [491:55] >> Right. So uh you know usually what [491:57] happens is that as Karishma clearly [491:59] mentioned about domain expertise which [492:00] is not something that AI can give. AI [492:02] have only those data which is available [492:05] on the web. What about the things that [492:07] are there on the product discussions? [492:09] What are the things that are there in [492:10] the sales teams right? What are there in [492:13] your BRDs, your RFPs as well, right? [492:16] Those discussions are the things which [492:19] AI no AI can give you, right? So even if [492:23] for example what change that we we even [492:25] brought is that we are now building a [492:28] content strategy which has a very u you [492:31] know different approach in terms of when [492:33] we are working with uh SAS products or [492:35] even now the service agencies as well we [492:38] have tried to make a combination of uh [492:41] you know like say middle of the funnel [492:43] content with bottom of the funnel [492:44] content and then top of the funnel [492:46] content along with that bottom of the [492:48] funnel content right so in that case [492:51] What we are trying to achieve is for [492:53] example for top of the funnel [492:54] information it can be 70 to 80% where [492:58] things are something which might be [493:01] people knowing right but then 20 to 30% [493:04] we try to bring the expertise from those [493:07] data right maybe we talk I mean at [493:11] missive we have this process that uh [493:13] every week instead of talking to the you [493:15] know just founders and the SEO team we [493:17] talk to the sales team and the product [493:19] team we understand what type of [493:21] requirements that we have got. Based on [493:23] those requirements, we build the content [493:25] topics. We curate the content and then [493:28] in that particular content we try and [493:31] ask the same sales team or the product [493:34] team that where and how are we going to [493:36] help such a requirement and then we [493:38] build a content around it. So those are [493:40] the areas where humanization becomes [493:43] really really important because such [493:45] domain expertise or you know even I [493:47] think Julian also me mentioned during [493:49] our conversation which is EAT right so [493:52] those expertise experience [493:53] authoritiveness and trustworthiness [493:55] right that only comes with such uh you [493:59] know expertise which is either in-house [494:02] or whether you go out ask experts and [494:04] put them in your content. So that is [494:07] something that you know uh humanization [494:09] that is where humanization becomes you [494:12] know non-negotiable [494:14] and of course that is part of content [494:15] but then when it comes to SEO as well [494:17] there are hell lot of areas uh which [494:20] even I think Sep mentioned in his [494:22] example that sometimes AI can be really [494:24] really wrong and that is the reason why [494:26] you have to be open um you know you have [494:29] your eyes open to see um I I would like [494:32] to highlight one example from our [494:33] Italian client and since they speak [494:37] Italian and they have their entire [494:39] website uh in this language and since we [494:41] are not uh you know that really thorough [494:44] with that language. So what we did once [494:46] is that you know uh they send us a huge [494:49] report of uh uh some uh data and then it [494:52] was really difficult for us to [494:54] understand what exactly that data talks [494:56] about right we use that PDF and we uh [494:59] you know with the client's consent we [495:01] put it on claude and we you know saw uh [495:04] what they actually mean and then we [495:06] asked claude that okay give give us some [495:08] recommendations what they are trying to [495:09] say and what they are trying to do and [495:12] uh a couple of things were really wrong [495:14] in terms of CTA, right? That they [495:17] recommended that CTAs are not in well [495:19] place or something like that. But when [495:21] we actually went and translated it via [495:24] Google Translate, it was like, you know, [495:26] we were like, "No, no, no. This is not [495:28] something that we clearly can accept." [495:30] And [495:30] >> that can be an embarrassing. [495:32] >> Exactly. It reminds me of something that [495:34] my high school English teacher said to [495:36] me that if you want to write creatively, [495:39] if you want to create something that is [495:42] meaningful and interesting, you have to [495:45] learn the rules first. You can't break [495:47] the rules effectively. You can't write [495:50] new rules unless you already know the [495:53] existing rules. [495:54] >> Yeah. [495:55] >> And that human check is so essential [495:58] along the way. You have to be [495:59] responsible for your own QA. [496:03] like you were saying um excuse me Sandy [496:06] you you're saying in the process of [496:10] checking the website's code you can see [496:12] with your own eyes that it's not what AI [496:15] is telling you you have to be [496:17] responsible for that [496:18] >> yeah absolutely I think having a human [496:20] in the loop is is essential u what I [496:24] think a lot of people misunderstand [496:28] uh this with AI is the idea of [496:32] handing off the responsibility [496:35] of of completing a task to AI. But I [496:38] think you can't outsource human [496:40] judgment. Uh that needs to be uh [496:43] retained. You can get AI to to run [496:46] audits and you know generate images and [496:49] videos and whatnot but at the end of the [496:52] day you still need to use your critical [496:54] eye to you know decide whether this is [496:57] good or good enough. [496:59] >> Yeah. uh [497:00] >> before it goes in front of the client. [497:02] >> And that's that's such a great point. Um [497:05] but maybe what about the other side of [497:06] things? So we get asked all the time as [497:09] professionals in marketing, how are we [497:12] making use of AI? And I get this [497:15] question all the time at Kinsta as well. [497:16] How is Kinsta using AI? And actually the [497:19] Kinsta CEO had a response to this that [497:21] it's not about using AI for the sake of [497:23] using AI, but how are we responding to [497:27] this shift? uh how are we helping our [497:30] customers to [497:32] be discoverable to be available to show [497:36] up in in in results uh to provide the [497:40] best service that we can. I heard [497:42] something wonderful from uh one of the [497:44] other speakers uh Luke Carbas I don't [497:46] know if he's in the room at the moment [497:48] but he gave a talk yesterday [497:51] uh where he was mentioning about in the [497:53] AI era uh competition becomes very [497:56] different because in the past maybe your [497:58] competitor could could launch a new [498:00] feature or develop some some new content [498:03] that was so great it would take you [498:05] months to catch up. Now, with AI tools, [498:08] you can create a a concept of that tool [498:10] in just a day. [498:12] >> Yeah. [498:13] >> So, what's left? What do we have to [498:14] compete on at that point? Well, Luke's [498:17] thesis, and I'm sorry if I'm I'm [498:18] paraphrasing, but your talk meant so [498:20] much to me. I I wanted to share it here [498:22] as well. Um, two things. Number one is [498:25] your brand, but maybe even more than [498:27] that is your user experience. And [498:29] regardless of what type of service you [498:30] provide, that user experience and making [498:33] sure that it's a good one is so [498:35] essential. So with that in mind, um I [498:38] wanted to come back to to Julian what [498:40] you were saying about uh how when you're [498:42] building client sites um making them [498:45] discoverable [498:47] um and maybe maybe this is a question [498:50] for for the whole panel perhaps um first [498:53] if we talk about AEO and GEO if you [498:56] could maybe just briefly explain to our [498:58] audience what's the difference I know of [499:00] Krishma you were also mentioning that as [499:01] well as this is very uh closely tied to [499:04] the content creation process and [499:06] delivery process that that you are so [499:08] familiar with. If you could just explain [499:10] what that is to us and then how can we [499:13] we respond because I think everyone in [499:15] this room has some interest in WordPress [499:18] and uh building effective websites [499:20] creating and and hosting or providing [499:23] tools for effective tools that benefit [499:26] everyone. So maybe if I don't know if [499:29] Julian or or someone uh on else on the [499:31] panel could explain to us what is the [499:33] difference between AEO and GEO and why [499:36] should we care first of all? [499:39] >> Yeah. [499:40] >> Okay. Um [499:42] uh so the way we see it I mean AEO GEO [499:46] has been somewhat used synonymously [499:49] not entirely well in the last few years [499:53] and the reality is that we are still [499:54] very fuzzy on what works and what [499:56] doesn't. Um AEO is within the search in [500:02] Google or in Bing or whatever it may be [500:04] AI overview. So it's answer engine [500:07] optimization and it it somewhat mimics [500:11] what uh how we used to use Google. You [500:15] have a search query, you're expecting an [500:17] answer um and then you do another search [500:21] query for more answers and so on. And it [500:23] is that answer uh key value pair as we [500:27] say. Whereas with generative engine [500:30] optimization, it's different. You're [500:32] part of a larger conversation which has [500:34] a lot more context. [500:36] And the way at least I have seen in a [500:39] very nent way and please don't quote me [500:41] on this um the way we appear as a brand [500:46] in GEO whether it's chat GPT or claude [500:50] or any of the other engines that people [500:53] use is that it's part of a larger [500:56] knowledge. So if at that point of time [501:00] if there's something uh that the engine [501:03] can use from our deeper insights which [501:06] are buried in our articles that shows [501:09] up. So insight shows up in GEO in my [501:12] opinion and uh if you have really well [501:15] ststructured articles which have query [501:18] answer query answer in a way that is not [501:21] only uh that that tends to show up [501:24] better in AEO. This is my opinion. [501:26] Again, please don't quote me on this. [501:28] >> Thank you. Thank you. So, so to [501:30] summarize, um, there are a couple of [501:33] ways that we, so earlier in the [501:34] discussion that we were talking about [501:36] how are we using AI tools and now I [501:39] think we're kind of shifting the [501:40] conversation over to how are we [501:41] responding to the ways that other people [501:43] are using a uh AI tools. Um, because we [501:47] have to be there for both both of these [501:50] sides of AI. In other words, how are we [501:53] using it and how are respond we [501:54] responding to it. It's all part of this [501:57] shift that is happening that we have to [501:58] be aware of and we have to use. So the [502:01] AEO and and GEO is about how uh your [502:04] content is discoverable. Um how and that [502:07] ultimately translates to how effective [502:09] your marketing is, your conversions and [502:12] ultimately your return on investment. [502:15] Um, so with that in mind, I'd love to [502:19] hear some maybe get to a a question for [502:22] the panel of kind of a case study, a [502:25] specific example maybe with a client or [502:27] partner or uh someone else on your team [502:31] uh of how you're able to respond to AI [502:35] in terms of what you're the content that [502:38] you're building uh or how you're working [502:40] with clients a specific project. um just [502:43] to share some practical examples with [502:45] our with our audience [502:47] >> right [502:49] >> you want to go first yeah sure [502:50] >> yeah sure so as I mentioned earlier [502:52] about the content combination that we [502:54] are now working on right so that [502:56] combination what has happened is when we [502:58] are using an one ICPdriven content where [503:03] there is one specific audience there are [503:05] one specific problem and there are one [503:07] specific solution when we are trying to [503:09] combine this kind of a format that has [503:12] you worked for us much better than the [503:15] content that we used to create for [503:16] search volume or something like that. So [503:18] now what happens is with such a [503:20] combination of content we get just one [503:22] visitor right but that one visitor [503:25] becomes our one customer straight away [503:28] which becomes that if there is one [503:30] visitor coming from charged we get 100% [503:33] conversion ratio and the ticket size is [503:36] also really great for that I would like [503:38] to uh you know share an example that we [503:40] wrote uh uh an article on odo versus uh [503:44] uh quickbooks or something right we did [503:46] some kind of uh a comparison blog uh on [503:49] the same and we published it I think six [503:52] or seven months ago and you know uh the [503:56] usual quick KPIs as SEO we have is that [503:59] you know whether that p whether that [504:00] content has reached the page one or not [504:03] right but then the struggle was since 3 [504:06] days our content didn't reach page one [504:08] right so the internal struggle was that [504:10] why it is not reaching on page one why [504:12] it is not reaching on page one since we [504:14] put so much effort so much subject [504:15] matter expertise and everything right [504:17] But then on the third day we are getting [504:21] a call from the client that we have got [504:24] a lead from Germany which needs some [504:29] kind of uh you know development into the [504:32] odo space but uh they are not looking to [504:36] go for QuickBooks and we were like how [504:37] is that possible? I mean and then what [504:40] we did is we went to Microsoft clarity. [504:42] We watched the entire journey that how [504:44] they landed on the website and where [504:45] they reached and it was just one visitor [504:48] on that particular blog which is not [504:51] ranking on page one yet got the [504:54] conversion for us. [504:55] >> So I want to highlight what you just [504:56] said. We were talking earlier also about [504:59] how when someone is using GPT tools, AI [505:02] tools to do product research to find [505:05] solutions to find should I do this way [505:07] or that way the conversion rate is [505:10] actually vastly higher than traditional [505:14] search engines. [505:15] >> Yeah, that's right. And there are many [505:17] reasons for that which I mean you there [505:18] are specialized panels or books and all [505:21] kinds of videos and things that you can [505:22] read about why that is. But essentially [505:24] you're getting a warmer much warmer [505:26] lead. [505:27] >> Yeah. [505:27] >> Um and they're already targeted directly [505:29] at you. [505:30] >> Yeah, [505:31] >> that's a great great example. Thank you [505:33] Julian. How about you? What what's an [505:34] example of u some ways that that you [505:37] >> what I learned from that you've seen? [505:38] >> Yeah. from the client is we I think EA [505:41] is very important for this AOG kind of [505:44] things and the content you don't need to [505:47] be so complicated and long kind of [505:48] things you need to be is it answer the [505:51] questions [505:52] >> because a lot of people are answer [505:53] asking the question through change video [505:55] kind of things as long those query is [505:57] there the change pick up you have very [506:00] much high conversion rate is so much [506:02] higher compared to like uh SEO like [506:04] ranking factors kind of thing because [506:06] the traditional way is like everyone [506:08] want to go to page one. Now it's no [506:10] longer in page one. Now everyone want to [506:12] be in AI overview or in [506:14] >> you need to be trustworthy. You need to [506:16] be authoritative and you need to be [506:17] original perhaps. [506:18] >> Yes. You need to be original. [506:20] >> So that means not slop. [506:24] >> Yeah. [506:24] >> And we were talking earlier about bots. [506:26] Maybe um I don't know if Sandy if that [506:28] was kind of the story that you wanted to [506:29] to share. Um I'm curious though about [506:32] one of your perhaps uh case study [506:34] examples. [506:35] >> Yeah. Um I actually wanted to share um a [506:38] recent example of u how we've built out [506:44] a video editing pipeline [506:46] >> um in cloud code. So this is for a [506:49] client where we were doing uh their [506:51] admissions campaign and we were running [506:54] ads video ads on Google and meta. Um [506:59] the old school way of doing it was get a [507:01] video editor, put the clips together, [507:04] create a 30-cond video ad, uh which can [507:07] then go live. But we experimented with [507:10] uh with clot code by giving it the [507:12] source video files and giving it all the [507:16] right context and getting it to actually [507:18] edit and put together the final video. [507:21] And in the few runs that we've done, [507:25] I'd say about six out of 10 videos turn [507:27] out to be good. [507:28] >> So I think that's another great example [507:30] as well. But I want to highlight in this [507:32] example that um Sundep is not using AI [507:35] as a replacement for the video itself. [507:38] It's humanshot video of real situations, [507:43] >> correct? [507:43] >> Created by humans. What we're what [507:45] you're using AI for is the synthesis. [507:47] It's the editing. It's the structure. [507:50] And I think that's another great example [507:51] of not just throwing it at the AI and [507:55] using AI as a or attempting to use AI as [507:58] a as a replacement for human judgment, [508:01] but for synthesis, as a tool, as a [508:03] timesaver, and then you have the final [508:06] human check in there uh at the end as [508:09] well. So what goes in comes from a [508:11] human. The AI helps put it together and [508:14] then what comes out is checked by a [508:16] human. And so you can be sure that it's [508:18] original and that it's quality. [508:20] >> Correct. I think you need to provide the [508:22] right amount right kind of scaffolding [508:25] >> uh for agents to to be able to do uh for [508:27] them to do do a good job. Uh it's not [508:30] not just giving them a prompt and and [508:33] expecting you know amazing results. [508:35] >> That's a fantastic example. Thank you. [508:37] And uh Karishma [508:39] >> so um since I work with a product [508:41] company I'm my own client unfortunately. [508:44] So uh what we have done is built a lot [508:47] of tooling and what we have done is just [508:51] as people who aren't writers and want to [508:54] engage with audiences have been using [508:56] generative AI to write articles or [508:58] LinkedIn posts and so on. We have been [509:02] using as a marketing team as a content [509:04] team to develop tools for ourselves. For [509:06] instance, two of the brands I manage at [509:09] the moment, we have upwards of 500 [509:13] active articles on them and about across [509:17] both brands, we have 100 top articles [509:19] which drive revenue for us. Now to stay [509:23] on top of performance especially since [509:25] it has been fluctuating so much in the [509:27] last 18 months to two years is an uphill [509:31] battle for one person to go into Google [509:33] search console not only track uh [509:36] performance for the focus keywords but [509:38] also for the shoulder keywords and then [509:41] figure out and then figure out what went [509:44] wrong how what goes wrong all that and [509:46] that's a lot of processing manually that [509:50] to be very honest could have been solved [509:53] before by the engineering team but [509:54] engineering team works on product [509:56] marketing as secondass citizens so we [509:59] were like okay but we've been able to [510:01] solve that for ourselves [510:03] >> so one SEO person isn't spending her [510:05] entire week crunching data [510:07] >> getting uh CSV files uploading it to [510:10] Excel trying to find this needle in the [510:12] haststack that uh oh one competitor has [510:18] published an article they've added some [510:20] more some more insight And that's why [510:22] our article has dropped from 2 to 9 for [510:25] example. [510:26] >> Yeah. Yeah. And this is a lot and this [510:29] is not scalable which what we said and [510:31] with AI building those tools inhouse [510:35] um within the team itself removing the [510:37] dependency from engineering and product [510:40] to be able to do this is allowed us to [510:43] really build astronomical scale for um [510:47] pattern analysis and uh yeah hot fixes [510:50] that we needed to do [510:52] >> because you're closely invested in the [510:53] process and you make sure that the data [510:55] that goes in is good and then it's [510:58] checked properly. [510:59] >> So we were we we mentioned the phrase [511:01] earlier garbage in garbage out. If [511:04] you're using garbage then you're not [511:06] going to get much more than garbage [511:08] coming out. But if you control the [511:10] process going in you understand what's [511:12] happening along the way and you check it [511:14] coming out then you'll end up with [511:15] something that's quite good and you can [511:17] achieve some great benefits in terms of [511:19] conversion, in terms of time saving um [511:22] and maybe even in terms of quality as [511:24] well overall quality. [511:26] So, we have time for like maybe one or [511:28] two questions. If anyone has any [511:30] questions, uh we'd love to hear from [511:33] you. [511:36] >> Yeah, please. Uh if you could just share [511:39] your name uh where you're from and [511:41] what's your question? [511:43] >> Uh hello, Tinugala this side. I am from [511:47] Mumbai. I own a social media marketing [511:50] agency. So my question is like uh I have [511:56] an website my own uh sorry did was it [512:01] listenable? [512:03] >> Yes we can hear you. [512:03] >> Okay. Uh so I have my own website in [512:06] that I have integrated an AI chatbot. So [512:11] it is uh integrated with WhatsApp. So [512:15] whenever a lead comes so it uh is [512:19] directly converted in that AI chatbot [512:22] only. Uh but my uh question is when it [512:27] it is convertible in chats only but uh I [512:32] want in calling version. So is there any [512:36] software or tools because I have seen uh [512:41] mostly my leads uh convert in chat is 15 [512:47] to 20% and in call it is 35 to 60%. [512:54] So how can I overcome in calling [512:57] versions? [512:58] >> This is a great question. Sorry. Uh I [513:01] have also built a AI call bot also but [513:05] it's not very uh user friendly and not [513:11] converting that much percentage. [513:13] >> I have a great example that I just read [513:15] about if I may share this. So um the [513:20] price of Guinness beer in Ireland is [513:24] quite important. Uh it's the most [513:26] popular beer in Ireland. Uh it's very [513:28] popular in that country to drink a [513:30] Guinness and go to the pub and it's it's [513:32] part of the culture. Um but the price of [513:35] Guinness is not the same at different [513:36] pubs. [513:38] It what one thing that happened recently [513:40] is someone used AI tools to create a [513:43] calling bot that would call different [513:45] pubs around the country [513:48] >> and just ask them what the price of [513:50] Guinness is. What's the price of a pint [513:52] of Guinness? and it would listen to the [513:54] response and then create a database all [513:57] with AI tools. As a result of this uh AI [514:02] tool, this person one person was able to [514:05] create a database of hundreds of pubs [514:07] all over the country and that became [514:09] public knowledge and that lowered the [514:12] price of beer nationally. [514:16] So that's a massive effect. What you can [514:18] do is you could create something like [514:19] that with AI tools. Um, if anyone has [514:22] anything to add, uh, like a specific [514:25] tool that you might be able to use. [514:29] >> No idea. [514:30] >> It's possible though. Yes, you could use [514:32] AI tools. Yeah, thank you. [514:34] >> We're If we have time for one more [514:37] question, if is okay? [514:39] >> Yeah, please. [514:43] >> Could we get a microphone over there? [514:46] Thank you. [514:53] Yes. So I um run a marketing agency. Um [514:59] I want to know uh how do we use uh AI [515:03] training for our client projects maybe [515:06] something like cloud project maybe uh [515:09] chat GPT project uh building custom [515:11] skills on the cloud. Um and one um [515:16] discussion point that was there that um [515:19] uh input comes from human and output is [515:22] uh quality assessed by human. Uh we at a [515:25] times are in a situation where input [515:28] from the client is coming um which is AI [515:32] generated not verified not applicable. [515:35] Uh client just does a chat GP prompt get [515:38] some output and send it as a source. Now [515:42] uh we are not supposed to be subject [515:44] expert on every industry. So we take [515:46] clients word as their word. Uh how to [515:50] how do we navigate this situation? [515:53] >> That's a great question I think about [515:54] the the how do you train AI tools and to [515:57] to make them better use of them within [515:59] your own organization. Um [516:03] I am not quite so technical but perhaps [516:05] one of our panelists might might have [516:06] >> I would like to uh you know address this [516:08] on like say based on my experience. So [516:11] uh you know uh usually uh in such a case [516:14] what we try and go is we go with the uh [516:16] you know discovery session in that case [516:18] because a discovery session can be [516:20] recorded right uh in that case uh charge [516:24] won't help and won't come into the [516:26] picture right so we go onto the [516:28] discovery call and we record everything [516:31] that they say that they require right [516:33] and we use that recording to feed AI [516:37] right that okay this is what it is and [516:39] also um you You know sometimes even [516:41] client does not know certain things like [516:44] say for example yesterday I was having a [516:46] call with someone and then they they [516:48] were building a product but which they [516:50] they does not have that clarity that [516:52] okay whom they are targeting as their [516:54] ICP right they don't have a vision of [516:56] where they want to reach in the next two [516:58] to five years in that case it's us who [517:01] went into the deep uh dive discussion [517:04] and we help them understand that see the [517:07] product that you are building it already [517:08] has huge competition Right? So if you [517:11] will just go blind, you know, it will be [517:13] like just throwing arrows randomly [517:15] somewhere. Right? So in that case, our [517:17] own expertise would also help them uh [517:20] clarify, define that what they want to [517:23] build and then that part of information [517:26] we can feed to the AI so that AI can [517:30] have better clarity before they come up [517:33] with something which is uh you know way [517:36] our thing. [517:38] One thing also we've done at Kinsta as [517:40] well is we have uh corporate [517:42] subscriptions to tools like Claude, [517:44] Gemini, Chat GBT and different teams [517:47] within Kinsta use all of these tools. [517:49] With the corporate subscription, you're [517:51] able to create kind of a walled garden. [517:53] This is important for data privacy for [517:55] one thing. Um but also because you can [517:57] help train that tool on your own [517:59] content, your own manuals, your own [518:01] in-house materials. Um, and that will [518:04] make the AI tools more specifically [518:06] useful to your own organization as well. [518:11] Do we have time for one more question? [518:14] Is that possible? No, I'm sorry. We have [518:15] to wrap up, but we'd love to connect [518:17] with you afterward. Thank you very much [518:19] for listening and thanks to our [518:20] panelists and to the organizers. [518:28] >> Thank you. Thank you the panelist. Thank [518:30] you, Alex, for uh closing today's [518:33] program. Would uh like to offer [518:35] something for you. [518:49] And a quick reminder for everyone, uh, [518:51] our sponsor hall will open tomorrow at [518:54] 8:30 a.m. I request everyone to stop by [518:58] tomorrow and see how our sponsors are [519:02] working in WordPress community what they [519:04] are doing and uh like today uh we will [519:08] start tomorrow's session at 9:30. Uh on [519:11] the behalf of organizers I will say uh [519:14] good day and thank you everyone. [519:58] Hey, [520:06] hey, hey. [520:22] Hey. [520:40] Heat. Heat. [521:13] Hey, [521:21] hey, hey. [522:30] Heat. Heat. [522:36] Heat. Heat. [523:16] Folks, at this time we ask that you [523:18] gather your things and exit through the [523:20] doors behind you. We need to prepare for [523:23] tomorrow's sessions. Thank you, [523:24] everyone.