A New Era The Future of the Intelligent Web

A New Era: The Future of the Intelligent Web ![rw-book-cover](https://img.youtube.com/vi/YaIEdEj4Wuw/maxresdefault.jpg) 🔗 **Source:** [A New Era: The Future of the Intelligent Web](https://www.youtube.com/watch?v=YaIEdEj4Wuw) > [!note]+ Document Notes > tags: #artificial-intelligence #intelligent-web #wordpress #vector-databases #semantic-search > author: Theory Müller > source: WP Engine > reference: https://www.youtube.com/watch?v=YaIEdEj4Wuw > date created: 2024-12-19 14:30 > > # 🏷️ Title: A New Era: The Future of the Intelligent Web >  ## 🧰 Action Inbox - [H1] Implement semantic search using vector databases for better content discovery — 15:45 #semantic-search - [H2] Prepare websites for agentic interactions using MCP protocols — 25:30 #ai-agents - [H3] Evaluate AI infrastructure needs beyond traditional PHP/MySQL stack — 20:15 #infrastructure ## 🧭 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 - [WP Engine Decode Conference](https://wpengine.com/decode/) - [WordPress Developer Resources](https://developer.wordpress.org/) ## 📝 Summary - WP Engine VP Theory Müller presents a framework for transitioning from the "classic web" to the "intelligent web" powered by AI - Key principle: "Human first, AI native" - keeping humans at both ends while AI powers the middle layer - Traditional content lifecycle (creation, distribution, consumption) is being disrupted by AI agents and new interfaces - Infrastructure challenges: PHP/MySQL stacks need external AI capabilities like vector databases and GPU resources - Solution involves decoupled AI infrastructure that integrates seamlessly with existing platforms like WordPress ## 📖 Main Content The presentation introduces a fundamental shift from the traditional web to an "intelligent web" where AI mediates between human creators and consumers. Müller argues that while AI feels overwhelming due to rapid changes, a clear guiding principle can provide stability: "Human first, AI native." This philosophy means maintaining humans at both ends of the content journey (creator to consumer) while using AI to power everything in between. The complete journey is "human to agent to agent to human" rather than simple agent-to-agent interactions. The traditional digital content lifecycle consists of creation, distribution, and consumption. AI is transforming each segment: creation tools are becoming agentic builders, distribution involves new agentic interfaces, and consumption happens through AI assistants rather than direct website interaction. A major challenge is the "M by N integration crisis" - connecting multiple content sources to numerous AI destinations while maintaining compatibility with traditional search engines. This creates four friction points: creation paralysis, discovery black holes, agentic incompatibility, and website experience gaps. The solution involves external AI infrastructure that handles intensive workloads (like vector databases and LLM processing) while integrating seamlessly with existing platforms through APIs. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/YaIEdEj4Wuw?t=0) — [H1] Introduction to the Intelligent Web - Key idea: AI is fundamentally rewriting human online interactions - Evidence: "The digital landscape is really going through a massive structural change" #### ⏱️ [02:30](https://youtu.be/YaIEdEj4Wuw?t=150) — [H2] Human First, AI Native Philosophy - Key idea: Humans must remain at beginning and end of the journey - Evidence: "Regardless of the technology in the middle, the beginning and the end of the journey must stay human" #### ⏱️ [05:45](https://youtu.be/YaIEdEj4Wuw?t=345) — [H1] Digital Content Lifecycle Framework - Key idea: All content follows creation, distribution, consumption cycle - Evidence: "Every piece of digital content follows this path" #### ⏱️ [08:15](https://youtu.be/YaIEdEj4Wuw?t=495) — [H2] AI Transformation of Creation - Key idea: Creation splits between integrated AI features and autonomous agents - Evidence: "There's a new completely world of agents that is emerging" #### ⏱️ [11:30](https://youtu.be/YaIEdEj4Wuw?t=690) — [H2] Evolution of Content Consumption - Key idea: Users expect semantic search and capable chatbots on websites - Evidence: "People are expecting search bars that understand human language" #### ⏱️ [13:45](https://youtu.be/YaIEdEj4Wuw?t=825) — [H3] AI Interfaces and Training Data - Key idea: AI interfaces answer from training data but lack real-time website connection - Evidence: "Those answers are really kind of static, they aren't necessarily connected to the live website" #### ⏱️ [15:30](https://youtu.be/YaIEdEj4Wuw?t=930) — [H1] Agentic Commerce Challenges - Key idea: Identity verification and liability remain unsolved for agent purchases - Evidence: "How does a site verify that an agent is authorized by a person to spend money?" #### ⏱️ [17:00](https://youtu.be/YaIEdEj4Wuw?t=1020) — [H2] M by N Integration Crisis - Key idea: Industry faces chaotic connections between multiple sources and destinations - Evidence: "Imagine if every light bulb in your house required a different plug" #### ⏱️ [19:15](https://youtu.be/YaIEdEj4Wuw?t=1155) — [H1] Four Critical Friction Points - Key idea: Creation paralysis, discovery black holes, agentic incompatibility, experience gaps - Evidence: "Without the right principles being applied, content may not be visible" #### ⏱️ [22:30](https://youtu.be/YaIEdEj4Wuw?t=1350) — [H1] Infrastructure Limitations of PHP/MySQL - Key idea: Traditional web stacks aren't designed for AI workloads - Evidence: "Running an LLM requires significant GPU resources" #### ⏱️ [25:45](https://youtu.be/YaIEdEj4Wuw?t=1545) — [H2] Vector Databases as Website Memory - Key idea: Vectorization provides context and understanding for AI features - Evidence: "If a website doesn't speak vector, it isn't really speaking the language of AI" #### ⏱️ [28:00](https://youtu.be/YaIEdEj4Wuw?t=1680) — [H3] Agentic Interaction Protocols - Key idea: Traditional PHP workers can't handle long-running AI sessions - Evidence: "They handle a request and they're pretty much done" #### ⏱️ [30:15](https://youtu.be/YaIEdEj4Wuw?t=1815) — [H2] WP Engine's Intelligence Cloud Solution - Key idea: External AI infrastructure integrates with WordPress through APIs - Evidence: "We make these AI capabilities directly accessible through turnkey products" #### ⏱️ [32:30](https://youtu.be/YaIEdEj4Wuw?t=1950) — [H3] Semantic Search Implementation - Key idea: Smart search uses meaning rather than keyword matching - Evidence: "If a user types footwear but the site only knows sneakers, search comes up empty" #### ⏱️ [34:45](https://youtu.be/YaIEdEj4Wuw?t=2085) — [H2] Future Expansion Plans - Key idea: AI capabilities will be available to sites hosted anywhere - Evidence: "We will expand smart AI availability to sites hosted anywhere, even outside WP Engine" ## 📚 References - [WP Engine Intelligence Cloud](https://wpengine.com/ai/) - [Model Context Protocol (MCP)](https://modelcontextprotocol.io/) - [WordPress Core](https://wordpress.org/) ## 🗣️ Paraphrase - The web is evolving from simple human-to-human connections to AI-mediated interactions that maintain human purpose while leveraging machine intelligence for processing - Traditional content creation, distribution and consumption are being transformed by autonomous AI agents that can operate independently of traditional user interfaces - Existing web infrastructure built on PHP and MySQL cannot handle the computational demands of AI features like vector databases and large language models, requiring external cloud-based solutions ## 📰 Newsletter The intelligent web is here, and it's reshaping how we build and interact with WordPress sites. Theory Müller from WP Engine shared a compelling framework: keep humans at both ends of the content journey while letting AI power everything in between. The key insight? Traditional PHP/MySQL stacks need external AI infrastructure to stay relevant. Think of it like image CDNs - you don't process heavy images on your web server, and you shouldn't run AI models there either. Vector databases and semantic search are becoming as essential as responsive design once was. For WordPress developers, this means preparing for agentic interactions where AI assistants will browse and interact with your sites on behalf of users. Your content needs to be machine-readable and your APIs need to support long-running AI sessions. Resources: [Learn WordPress](https://learn.wordpress.org/), [Developer Docs](https://developer.wordpress.org/), [WP-CLI](https://wp-cli.org/) ## ✅ Action Items #### WordPress - Core: Evaluate infrastructure needs for AI integration, vector database compatibility - Core Performance: Assess impact of AI features on site performance budgets - Core AI: Develop semantic search capabilities, agentic interaction protocols - Documentation: Create guides for AI-native development patterns - Meta: Plan .org roadmap integration with intelligent web standards - CLI: Add commands for AI infrastructure management and vector operations - Plugins: Establish guidelines for AI-compatible plugin architecture - Themes: Design patterns for conversational interfaces and semantic markup #### OSPO - Policy & Governance: Develop AI integration standards for open-source CMSs - License Compliance: Review AI model training data usage and licensing - Security Management: Assess vector database security, API authentication - Software Inventory: Track AI dependencies in SBOM strategies - Community Engagement: Fund research into decoupled AI infrastructure #### DevRel - Developer advocacy: Promote semantic web standards, AI-native development - Education: Workshop on vector databases, semantic search implementation - Developer experience: API documentation for AI infrastructure integration - Developer marketing: Case studies of successful AI-native WordPress sites ## 🧪 Hypotheses - Websites without vector database integration will lose discoverability in AI-powered search interfaces - Traditional PHP/MySQL performance will degrade significantly when attempting to run AI workloads locally - Semantic search will become a minimum viable feature for competitive websites within two years - Agentic commerce protocols will emerge as new web standards similar to how REST APIs evolved ### 📊 Methodology - Performance testing: Compare PHP-based AI processing vs external cloud infrastructure - User behavior analysis: Track search success rates between keyword vs semantic search - Market research: Survey WordPress site owners on AI feature adoption timelines ### ❓ Questions - How will WordPress core balance AI infrastructure needs with platform simplicity? - What authentication standards will emerge for agent-to-website interactions? - How will smaller WordPress sites access enterprise-grade AI capabilities affordably? ### 🔮 Future Research - Development of open-source vector database solutions for WordPress - Standard protocols for agentic website interaction beyond MCP - Economic models for shared AI infrastructure in the WordPress ecosystem ### 💡 Implications - Practice: WordPress developers need to learn vector databases and semantic technologies - Policy: New standards required for AI agent authentication and authorization - Security: API security becomes critical as agents interact with sites autonomously - Education: Training programs needed for AI-native WordPress development - Funding: Investment required in decoupled AI infrastructure for open-source platforms ### 🧠 Learning Styles Assessment Reading level estimate: 12th grade (Flesch-Kincaid) #### 👀 Visual learners - Present: Framework diagrams for content lifecycle, infrastructure comparisons - Improve: Architecture diagrams showing AI infrastructure integration, visual API workflows #### 🔊 Aural learners - Present: Full keynote video with clear narration - Improve: Podcast-style summary, audio-only technical deep-dives #### 🗣️ Verbal learners - Present: Clear definitions of technical terms, analogies (light bulb/adapter example) - Improve: Glossary of AI terms, comparison tables between old vs new approaches #### 🧩 Kinesthetic learners - Present: References to hands-on building, keyboard interaction - Improve: Step-by-step implementation guides, sandbox environments for testing #### 🧠 Logical learners - Present: Clear framework structure, cause-and-effect explanations - Improve: Decision trees for choosing AI solutions, performance metrics comparison #### 🧑‍🤝‍🧑 Social learners - Present: Community conference context, collaborative building references - Improve: Discussion forums, peer implementation review opportunities #### 🧘‍♀️ Solitary learners - Present: Individual reflection on implementation choices - Improve: Self-assessment checklists, personal learning path recommendations #### 🌿 Naturalistic learners - Present: Evolution metaphors for web development - Improve: Ecosystem thinking about platform interdependencies ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading structure in presentation, descriptive content - Improve: Alt text for technical diagrams, ARIA labels for interactive elements #### Keyboard navigation - Present: Standard video player controls - Improve: Transcript navigation shortcuts, searchable content sections #### Contrast and readability - Present: High contrast presentation slides - Improve: Customizable transcript display options #### Text resizing and reflow - Present: Responsive video player - Improve: Scalable transcript layout, adjustable text sizing #### Media and captions - Present: Full video with spoken content - Improve: Professional captions, downloadable transcript with timestamps #### Forms and interactions - Present: Basic video interaction controls - Improve: Interactive transcript with jump-to-time functionality ### 🌍 DEIB Assessment #### Cultural breadth - Present: Global web perspective, universal technology access - Improve: Examples from different markets, consideration of global infrastructure disparities #### Inclusive language - Present: Technical jargon explained with analogies - Improve: Plain language summary, multilingual key concepts #### Neurodiversity support - Present: Clear presentation structure, chunked information - Improve: Multiple information processing pathways, visual + auditory + text options #### Representation - Present: Focus on democratizing advanced technology - Improve: Diverse use case examples, consideration of different organization sizes #### Psychological safety - Present: Acknowledges AI anxiety and complexity - Improve: Clear guidance for different skill levels, reassurance about gradual adoption ### ✅ Logical Validity #### Strong reasoning - Deductive strength: Clear premise that AI requires different infrastructure, logical conclusion about external solutions - Inductive strength: Good pattern recognition from image CDN analogy to AI infrastructure needs - Causal reasoning: Well-established connection between vector databases and semantic understanding - Analogical strength: Light bulb adapter analogy effectively illustrates integration complexity #### Fallacies - Appeal to novelty: Some emphasis on "new" without sufficient evidence of superiority - Hasty generalization: Claims about "most sites" being static need more data support #### Tone of voice - Overall tone: Professional and educational with appropriate technical depth ### 🛡️ DARVO Analysis - Deny: No significant denial patterns observed - Attack: No attacks on opposing viewpoints or critics - Reverse victim/offender: No victim/offender role reversals - Emotional cues and rhetoric: Positive framing focused on opportunity rather than threat ## 🎯 Conclusion The transition to an intelligent web requires a fundamental shift in how we architect WordPress sites. The "human first, AI native" principle provides a stable foundation while everything else evolves rapidly. Traditional PHP/MySQL stacks need external AI infrastructure to remain competitive, just as image processing moved to CDNs. The key is preparing for agentic interactions while maintaining the human-centered purpose that makes the open web valuable. WordPress developers should focus on semantic markup, API design for AI agents, and integration patterns that leverage external AI capabilities without sacrificing platform simplicity. ## 🧱 Refactor Appendix ### 📚 Literature Split --- ## Full Transcript [00:11] Welcome back. In this keynote, we will [00:14] hear from my colleague Theory Müller, VP [00:16] of product at WP Engine on the future of [00:19] the intelligent web. In part one of the [00:22] keynote, we talked about how WP Engine [00:25] is solving today's disruptions, laying [00:27] the groundwork for tomorrow's [00:29] intelligent web solutions. [00:31] In part two of this keynote, Theory will [00:34] cover the future of digital experiences, [00:37] and how to stay competitive in a rapidly [00:40] changing landscape. Over to you, Theory. [00:44] >> Hello everyone. It's so great to be here [00:46] at Decode with the community that's [00:48] actually building the future of the web. [00:50] Today's session is going to focus on how [00:53] AI is fundamentally rewriting how people [00:56] interact online. I think by now everyone [00:59] understands that is not just a set of [01:01] new tools. The digital landscape is [01:04] really going through a massive [01:05] structural change. A fundamental shift [01:08] is really happening, moving us away from [01:10] what many call the classic web to what [01:13] is now being defined as the intelligent [01:15] web. This new era is built on this deep [01:19] understanding between human intent and [01:22] machine intelligence and trying to find [01:24] the balance between the two so that it [01:26] works well together. Now, you know, it's [01:29] clear that AI can feel incredibly [01:32] overwhelming and even creating some [01:34] anxiety. It's really complex. It's fast [01:37] moving, probably faster than anyone can [01:40] track really. Uh and but one thing [01:42] that's really exciting is that it's [01:44] undeniably super powerful. It feels like [01:47] every morning we wake up there's a new [01:49] model, there's a new protocol, there's a [01:52] new game changer that we have to deal [01:54] with. And I find myself actually uh deep [01:58] into technical challenges every day as I [02:01] think that we should all apply some of [02:03] the things that we learn. But very often [02:06] I feel this need to zoom out a bit. this [02:09] need to essentially understand the why, [02:12] understand why we building these [02:14] applications, these products and putting [02:16] this AI tooling um and applying them [02:19] across our implementations. [02:21] So this talk is not going to jump right [02:24] into technical topics although we will [02:27] but first as a starting point we're [02:29] going to zoom out uh with a very human [02:32] centric and easy to follow view before [02:35] we dive in. It really feels like a [02:38] bridge needs to be built between the [02:40] simple things that we understand today [02:42] and the more complex solutions that's [02:45] being created for tomorrow. So to make [02:47] sense of this change, a clear guiding [02:49] principle is needed. And I believe that [02:52] principle is actually really simple. [02:54] Human first AI native. It makes sense to [02:57] start by discussing why the human [03:00] element has to remain the priority. And [03:03] the core idea is that regardless of the [03:06] technology in the middle, the beginning [03:08] and the end of the journey must stay [03:10] human. This is the ultimate goal and the [03:13] main focus. And by human first, we [03:16] really mean by humans for humans. So [03:19] this is our foundation. Think about the [03:21] creator, the company uh with a vision, [03:24] with a story to tell, with a product to [03:27] sell. They are the ones that have been [03:29] breathing life into the web for decades. [03:32] And then on the other side, think about [03:33] the consumer, the person looking for an [03:36] answer, a connection, a product or a [03:38] solution. And this is really what we [03:40] mean by human first is what it [03:43] represents. Now AI native represents [03:46] everything that happens in between. If [03:49] human first is the goal, AI native is [03:52] really the engine that makes it works. [03:54] And in this context when we mean AI [03:56] native we mean all the AI technologies [03:58] that we see on the market agents the [04:01] platforms the interfaces essentially the [04:04] logic that happens between that humanto [04:07] human journey building today uh needs to [04:11] happen with AI as a foundation rather [04:14] than treating it as an afterthought. Uh [04:16] in this new world, everything between [04:19] the creator's brain and the consumer's [04:21] needs uh essentially has to become [04:24] intelligent. Interfaces are becoming [04:27] much more conversational. Uh platforms [04:30] are becoming agentic and the content [04:32] itself is becoming a lot more machine [04:34] readable and ready for agents to action [04:37] on it. This philosophy is really what [04:40] can help us and provide us to the [04:43] stability to stay grounded while the [04:46] technologies move at lightning speeds. [04:50] So to bring it all together, we hear a [04:52] lot this concept of agent to agent and [04:55] to me this captures the middle layer, [04:59] the AI native, but if we want to capture [05:01] the full journey, it's really human to [05:04] agent to agent to human. [05:07] All right. So now we have this top level [05:10] philosophy that is clear something that [05:12] we can anchor on and this is another [05:14] concept that I would like to share with [05:15] you that helps me and help us aligning [05:19] all the things that we do on a daily [05:21] basis and also understanding the depth [05:24] of the complexity in this space. This [05:27] framework is really simple is the [05:29] digital content life cycle. A very [05:32] simplistic way to look at it is to [05:34] essentially divide it in three segments. [05:37] Creation, distribution and consumption. [05:40] At its core, really every piece of the [05:43] digital content follows this path. So [05:45] now let's look at it individually. [05:47] Creation is the first segment. A good [05:50] tooling example to facilitate it is the [05:53] CMS's platform content management system [05:56] which is still largely applicable today. [05:58] This is essentially where posts are [06:00] written. This is where media is managed. [06:03] This is where pages are designed. The [06:06] second segment is distribution. This [06:09] essentially involves how the content [06:11] moves from a database or static page out [06:14] into the world. And for so many years, [06:17] you know, this has been mostly relying [06:19] on web technologies like HTML, CSS, and [06:23] JavaScript to serve pages into a [06:26] browser. And finally, there's the [06:28] consumption segment. And until now, this [06:31] has been mostly human activity. A person [06:34] sits down, open a browser, and interact [06:37] with the content on a website. This [06:39] cycle has been the foundation of our [06:41] entire industry. And really, these three [06:43] segments of the framework uh remain [06:46] immutable. It doesn't change today. But [06:48] the rise of the intelligent web, the [06:50] rise of AI is rewriting how things are [06:54] being done across the board. Really [06:56] changing the how behind this segment. [07:00] So looking closer at this life cycle [07:02] now, diving a little bit deeper, the [07:04] reality is that while the goal is still [07:07] serving people, the in between is [07:09] becoming incredibly complex. On the [07:11] creation side, AI features are still [07:14] appearing in existing tools, but there [07:16] also been a major uh market shift [07:19] towards agentic builders like lovable, [07:22] vzero, and replets. These are [07:24] essentially changing how content is [07:27] created and managed, moving the industry [07:29] away from this manual work towards [07:32] automated orchestration. Regarding [07:34] distribution, standard interface such as [07:37] website and mobile apps are shifting [07:39] with the rise of the agentic web. On the [07:42] consumption side, interfaces like chat [07:44] GPT and Gemini are increasingly becoming [07:47] the new way that people browse the web. [07:50] Users aren't just looking at the site, [07:52] they are asking models to summarize the [07:55] value for them and they are interacting [07:57] with agents more and more. So the [07:59] content life cycle is no longer a [08:01] straight line. Not that it was a simple [08:04] straight line until now. There's always [08:05] been fragmentation, but it has become [08:08] more complex and disrupted ecosystem for [08:11] every user with the rise of AI. So now [08:13] the core philosophy is set. We are we [08:16] have a framework to anchor on with a [08:18] simple understanding on how these three [08:20] segments are being impacted. Let's dive [08:23] a level deeper. It is worth looking [08:26] closer at how the content is actually [08:28] being created and managed in this new [08:31] environment. [08:32] creation is currently split into two [08:34] very different paths. The first one is [08:37] familiar interfaces being updated with [08:40] more AI features. I think we all [08:42] familiar with this new little star icon [08:45] which has become a universal sign of [08:47] like hey AI help me do something in this [08:50] interface. A simple example of that is [08:53] an email app with a helper to clean up [08:56] the writing or to create a draft or even [08:59] suggest a response. These are [09:01] essentially the integrated AI features. [09:04] AI living inside the software is already [09:07] being used that most users are really [09:10] familiar with. However, there's a new [09:13] completely world of agents that is [09:15] emerging and these agents are [09:18] multi-purpose agents. They are usually [09:20] separated from the existing software [09:22] interfaces and I think recently we saw [09:25] with open claw uh it has sparked an [09:28] explosion of these agents and the [09:30] fundamental difference is that instead [09:32] of these agents uh being embedded in an [09:35] existing UI and having the user clicking [09:38] through the UI an instruction is given [09:41] through a simple interface or even a [09:43] chat app uh like telegram for example. [09:47] These agents have the ability and the [09:50] skills to go out and make changes [09:52] directly in the software without the [09:54] user even looking at the interface at [09:57] any point of time. So these agents are [10:00] growing at massive rate, right? [10:01] Automating workflows and pulling reports [10:04] completely outside of our traditional [10:06] interfaces. And this is really happening [10:09] across the industry including in [10:11] open-source distributed software like [10:14] WordPress where agents can be plugged in [10:16] to interact with the application [10:18] directly. Turning to the consumption, [10:21] consider how users are actually [10:22] interacting with content today. So like [10:25] the creation, consumption is also [10:27] splitting into two very different paths [10:31] uh in somewhat a very similar pattern [10:33] than creation. First, let's talk about [10:36] website UIs that everyone is familiar [10:38] with. And I think on uh existing [10:41] websites, the expectations have moved [10:43] towards much richer experiences. People [10:46] are expecting search bars. They [10:48] understand human language, also known as [10:51] semantic search. Essentially, the site [10:53] understands the intent and deliver [10:56] results based on meaning and uh not on [10:59] words and keywords only. Users also [11:02] expect chat bots that aren't just script [11:05] menus but assistant that can perform [11:08] tasks uh on their behalf. So users [11:11] expectations on websites are becoming [11:14] much higher. That covers existing [11:16] websites but at the same time there's [11:18] this ever growing breed of AI interfaces [11:21] like chat GPT Gemini and cloud in which [11:25] users increasingly consume and interact [11:28] with content. And I really want to [11:30] double click on this one for a bit. So [11:32] the AI interfaces mostly answer [11:35] questions based on training data. If a [11:37] site is included in that training data, [11:40] they provide the answer. But those [11:42] answers are really kind of static, [11:44] right? They aren't necessarily connected [11:46] to the live website in real time. [11:49] However, there are also ways to connect [11:52] a site directly in these AI interfaces. [11:56] But a big topic is how agent find a [11:58] website and how do they interact with [12:01] it. And this opens this major discussion [12:04] about discovery often referred as AO or [12:08] SEO but also connectors that allow users [12:12] to interact with a site directly as [12:14] opposed to only getting the data from [12:17] the training set. So related to this [12:20] topic, I think another topic that we [12:22] can't ignore when it comes to taking [12:25] action uh from an AI interfaces is [12:28] commerce. Discovering, comparing and [12:31] purchasing product inside an AI agent is [12:34] a massive topic. In agentic commerce, [12:36] there are several layers of complexity [12:38] and you know this really deserve its own [12:40] session which I would love to talk more [12:42] about. But to take a few example, the [12:44] first one would be identity. How does a [12:47] site verify then that an agent is [12:51] authorized by a person to spend money? [12:54] The second one would be uh [12:55] responsibility. You know who is liable [12:58] if something goes wrong. So in the short [13:00] term this is about agentic readiness. [13:03] Making sure that product data is [13:05] structured so that an agent can find a [13:07] site. In the long term for this [13:10] industry, for our industry, it's about [13:12] building the standard protocols to [13:15] handle the full transaction. And we've [13:17] seen some flavors of that and we've seen [13:19] some protocols emerging on the market [13:21] like UCP and other protocols before. [13:24] There's a lot more to unpack in the [13:26] agentic commerce space. It's a [13:28] fascinating topic, a lot of complexity, [13:31] but I think that for now understanding [13:33] that if an agent can't find a site or [13:36] interact with a site, a huge part of the [13:39] future economy is essentially lost. So [13:41] here I've just scratched the surface and [13:43] we could dive much much deeper, but what [13:46] is clear is that the industry is [13:47] currently facing massive integration [13:50] challenges, especially as standards and [13:52] technologies are more fragmented than [13:54] ever and new things emerge every day. [13:57] This is often called the M byN [14:00] integration crisis. If you've watched [14:02] Ramadas keynote, he spoke about that and [14:05] I really want to reiterate some of these [14:07] friction points. So the industry is [14:09] forced to navigate a chaotic landscape [14:11] where connection must be made between [14:14] this large number of sources. sources [14:16] being websites, APIs, and other content [14:19] hubs to so many destinations like [14:22] Trajyp, Gemini, a bunch of agents while [14:26] still accounting for traditional search [14:28] engine. Okay, so to put it simply, [14:31] imagine if every light bulb in your [14:33] house required a different plug or a [14:36] different voltage. All the time would be [14:38] spent managing adapters instead of just [14:41] having lights. And this really feels [14:43] like the current state of the web. But [14:45] most critically, I think that there are [14:47] four critical friction points that are [14:50] causing a lot of complexity. The first [14:53] one is creation paralysis. How [14:56] essentially to build better, faster, [14:59] smarter uh with AI. And this really [15:02] often is kind of held back by the lack [15:05] of infrastructure capabilities or the [15:07] lack of integration and siloed AI tools [15:11] that don't essentially work together. [15:14] The second point is the discovery black [15:16] hole. Traditional search results tend to [15:19] be answered directly by AI interfaces. [15:21] We're seeing a lot of uh people talking [15:24] about the click-through rate uh which is [15:26] not necessarily affecting business goal [15:29] but a reality and without the right [15:31] traditional and new principle being [15:33] applied on the site, the content may not [15:36] be visible across the board. So [15:38] businesses really must maintain the [15:40] search visibility while ensuring that [15:43] they are surface in the AI interfaces or [15:46] appearing in the training data if they [15:48] intend to. The third one is the agentic [15:52] incompatibility. So most app and content [15:55] are essentially invisible to this new [15:58] agentic workforces because they lack [16:00] standardized protocols or implementation [16:02] solutions. And so if an agent can't uh [16:06] read a site, it probably won't recommend [16:08] it in the future. This is following the [16:10] same path as search engines. Search [16:13] engines uh you know have ranking and [16:16] algorithm for a good reason is because [16:18] they want to provide their users uh with [16:20] the best answers and best experiences. [16:23] The same will apply to uh AI interfaces. [16:27] probably the sites which are going to be [16:29] easy to interact with will be ranking [16:31] better and providing better user [16:33] experience. And the fourth one is really [16:35] this uh website experience gap. I think [16:39] even humans browsing a website now [16:41] expect much richer features the content [16:44] to essentially be alive and [16:46] individualized. Yet most sites today are [16:49] kind of static or fairly static and not [16:52] very individualized. [16:54] All right. By now I think a good [16:56] overview has been established on how AI [16:58] is rewriting some of the rules and some [17:01] of the industry challenges but really [17:04] these are actually exciting times and [17:06] you know with challenges come [17:08] opportunities. So let's speak a little [17:10] bit about these challenges um and then [17:13] we'll speak about solutions. So with all [17:15] of that, the big question is how is [17:18] relevance maintained still using the [17:20] tools and the platforms that are so [17:22] widely used and that we all love. First [17:25] the reality is that many open-source [17:28] softwares like Drupal, WordPress and [17:30] Jumla are built on PHP and MySQL. And [17:35] this has been an incredible stack. But [17:37] it is also important to consider that [17:40] the infrastructure and technology that [17:42] AI needs may be slightly different [17:45] especially at scale. The reality is that [17:49] PHP and MySQL are most often not the [17:52] right tools in the box for AI [17:54] capabilities just like so many things in [17:57] the past. Let's take a simple example. [18:00] Consider image CDN. In this case, large [18:03] images are not necessarily hosted on the [18:06] local web server, but rather external [18:09] cloud infrastructure handles the load, [18:12] the resizing, the multiformat, and all [18:15] the things that are needed for images at [18:18] scale. AI follows the exact same logic. [18:21] Take a large language model as an [18:22] example. Running an LLM requires [18:25] significant GPU resources. Attending to [18:29] run that on a standard web worker will [18:31] cause the site to slow down or even [18:33] completely stop. Right? So the most [18:35] efficient pattern today is to use uh [18:38] cloud LLM. This is the core of external [18:42] infrastructure. It applies to LLMs but [18:45] it also absolutely applies to other [18:48] needs that AI depends on such as vector [18:51] database which we'll speak more about in [18:53] a little bit. Infrastructure relevance [18:56] is the real and present challenge. [18:59] Existing platforms, specifically [19:01] distributed softwares like CMSS often [19:04] lack the infrastructure needed to stay [19:07] relevant in this AIdriven landscape. [19:10] These stacks are designed for management [19:12] not really the highintensity workloads [19:15] that AI needs and modern AI really [19:18] requires these high resoning specialized [19:21] components like vector database for [19:23] memory and GPUs for reasoning that don't [19:26] exist on the standard web server. So [19:28] potential lack of AI capabilities in a [19:31] platform leads to difficult choices for [19:34] users. switching platform entirely or [19:37] even potentially even leave the open web [19:39] behind or you know maybe wait and risk [19:42] to become irrelevant which is something [19:44] that we should absolutely address as an [19:47] industry. The solution is to connect a [19:50] site to powerful external capabilities [19:52] regardless of the platform used uh or [19:55] where it's hosted and integrating this [19:58] gracefully into the preferred platform [20:01] of the user. So let's unpack this topic [20:04] a bit further especially in the context [20:06] of vectors. Think of vectorization as [20:09] the brain or the memory of a website. A [20:12] system that provides the context for AI [20:14] features. It understands the deep [20:17] relationship difference between the [20:19] content pieces of a website. It has been [20:22] noted that vectors are actually [20:24] fundamental way AI models understand and [20:27] process information. In fact, when [20:30] Google released Turboquent, they said [20:32] exactly that vectors are the fundamental [20:34] way AI models understand and process [20:37] information. So, if a website doesn't [20:40] speak vector, it isn't really speaking [20:42] the language of AI. Vectorization allows [20:45] for moving from isolated tools to a [20:48] centralized core that essentially [20:50] understands the meeting rather than just [20:52] keywords. Just as an image CDN handles [20:55] heavy media offserver, an AI site stack [20:59] essentially handles the heavy lifting [21:01] that vectors needs. The separate [21:03] environment allows software to plug into [21:06] an enterprisegrade AI capabilities [21:09] through a simple API. [21:12] Now let's talk about another common uh [21:14] change is agentic interaction with a [21:18] website. This is essentially how agents [21:21] interact with websites. uh which [21:24] protocols like MCPs are designed to [21:26] address. The needs of agents and how [21:29] sessions are handled must be addressed [21:32] at scale. And we have seen some websites [21:35] plug in into cloud for an admin user to [21:39] interact with the site. This is fine, [21:42] but at scale it becomes a lot more [21:44] challenging. Traditional PHP workers are [21:48] essentially transient. They handle a [21:50] request and they're pretty much done. [21:53] And they are not built for this long [21:55] running AI sessions or streaming chat [21:58] experiences. And once again, there is a [22:01] need to have the right infrastructure to [22:04] manage these long sessions and [22:06] structured workers without clogging up a [22:08] web server hosting a PHP MySQL website. [22:12] I mentioned MCP but protocols like MCP [22:16] are essentially currently used mostly [22:18] for cloud-based LLM interaction but the [22:22] rise of new tools uh favors autonomous [22:25] skills making common tools like CLI even [22:29] more relevant nowadays the [22:30] infrastructure needs or different some [22:33] requires a translator for models while [22:36] others requires persistent environment [22:38] but by separating this from the web [22:41] server [22:42] um the specific engineering needs for [22:44] every requirement is finally possible. [22:48] So it's clear that the industry is [22:50] seeing a strong shift towards that with [22:52] major hyperscalers enormously investing [22:56] in infrastructure capabilities and [22:58] making it available through APIs. [23:01] At WP Engine, we have invested deeply [23:03] into our intelligence cloud [23:06] infrastructure which sits on the side of [23:09] a WordPress website that we power. So to [23:12] take a simple example, one of our [23:14] infrastructure capability allows a site [23:17] to build a dedicated knowledge base. As [23:20] I mentioned earlier, this is essentially [23:22] a brain with the context of a website [23:24] that lives on a modern stack in a vector [23:27] DB under the hood. However, vector DBs [23:31] alone is just a piece of the puzzle. [23:32] It's an infrastructure. But capabilities [23:35] alone don't serve a purpose unless they [23:37] provide an integrated experience that [23:40] brings value. So, we make these AI [23:43] capabilities directly accessible in a [23:46] site built with WordPress through [23:48] turnkey products. to keep on the same [23:51] example track. While the knowledgebased [23:54] is the capability, our smart search AI [23:56] plug-in is the product that turns this [23:59] infrastructure capability into immediate [24:02] value. So, it isn't just making search [24:05] just a little bit better. It completely [24:07] rewrites how search works by using [24:10] semantic understanding to provide [24:12] accurate results. Traditional search [24:15] essentially has been all about keywords [24:17] literally matching character for [24:19] character. So if a user types uh [24:22] footwear but the site only use or only [24:25] knows uh sneakers the search often comes [24:28] up empty. Semantic search really changes [24:31] that by understanding the human intent [24:34] uh behind the words. It recognized the [24:37] goal of the user uh understanding that [24:39] the two parts mean the same thing and [24:41] then delivers result based on the [24:43] meaning rather than just spelling and we [24:46] want to offer more of these capabilities [24:49] and integrated solutions and make them [24:51] more universally available. To give you [24:54] a teaser, we will expand our smart AI [24:57] availability to site hosted anywhere [25:00] even outside of WP Engine. We really [25:03] want to empower users to stay relevant [25:07] independent of their choices of platform [25:10] and hosting providers by providing this [25:13] enterprise bleeding edge AIdriven [25:15] capabilities integrated directly into [25:18] their favorite platforms. [25:20] Thinking much further about this [25:22] capability and intelligence that we can [25:25] offer in the future. Um, when we combine [25:28] this external brain from a vector DB [25:31] with a cloud-based LLM, the results [25:34] become so much richer. We've actually [25:36] been playing around combining these and [25:39] the results are incredibly promising. We [25:42] would like to think that every website [25:44] on the open web can be supercharged and [25:47] showing that the open web remains one of [25:50] the most powerful platform in the era of [25:52] AI. [25:54] In closing, I want to reiterate the main [25:56] points. Uh we need a simple framework to [25:59] anchor on. We need a basic principle [26:02] that is human first AI native and AI [26:05] capabilities depend on decoupled managed [26:09] infrastructure that is reliable, secure [26:12] and essentially built for modern [26:14] technologies. And we are seeing this [26:16] transition towards this decoupled [26:18] managed AI infrastructure with built-in [26:20] intelligence powering products that work [26:23] right out of the box in a user preferred [26:26] software keeping it relevant in this AI [26:29] shift. This handles the complexity for [26:32] the user and it means that as new AI [26:35] models launch, as new pro protocols [26:38] emerge, a site is ready by default. This [26:41] is an incredible time to be building on [26:44] the web. I find it hard to take my hands [26:47] off the keyboard every day as the [26:49] process of building things is so [26:51] fascinating and exciting. I'm excited [26:54] about the work being done to make [26:56] opensource softwares and the open web [26:59] relevant and ensure it remains the [27:01] platform for the future of the [27:03] intelligent web. A last few words from [27:06] me. Remember, human first by humans for [27:10] humans. AI native everything that [27:13] happens in between. Thank you for your [27:15] time and I look forward to building this [27:17] future with the community.  📄 **Full text:** ![[Full Document Contents/Articles/A New Era The Future of the Intelligent Web.md|A New Era: The Future of the Intelligent Web by YouTube]] ## Highlights - new tools. The digital landscape is really going through a massive structural change. A fundamental shift is really happening, moving us away from what many call the classic web to what is now being defined as the intelligent web. This new era is built on this deep understanding between human intent and machine intelligence and trying to find the balance between the two so that it works well together. Now, you know, it's clear that AI can feel incredibly ([View Highlight](https://read.readwise.io/read/01kr1zbncx57py8az7n3x0g0cy)) - Tags: 👻 ai highlighted - has been mostly human activity. A person sits down, open a browser, and interact with the content on a website. This cycle has been the foundation of our entire industry. And really, these three segments of the framework uh remain immutable. It doesn't change today. But the rise of the intelligent web, the rise of AI is rewriting how things are being done across the board. Really changing the how behind this segment. ([View Highlight](https://read.readwise.io/read/01kr1zbndw8axhkb9b35ycd0ec)) - Tags: 👻 ai highlighted - So looking closer at this life cycle now, diving a little bit deeper, the reality is that while the goal is still serving people, the in between is becoming incredibly complex. On the creation side, AI features are still appearing in existing tools, but there also been a major uh market shift towards agentic builders like lovable, vzero, and replets. These are essentially changing how content is created and managed, moving the industry away from this manual work towards ([View Highlight](https://read.readwise.io/read/01kr1zbneej6mrqp5h149zcgds)) - Tags: 👻 ai highlighted - essentially the integrated AI features. AI living inside the software is already being used that most users are really familiar with. However, there's a new completely world of agents that is emerging and these agents are multi-purpose agents. They are usually separated from the existing software interfaces and I think recently we saw with open claw uh it has sparked an explosion of these agents and the ([View Highlight](https://read.readwise.io/read/01kr1zbng4ytkzz6n5m1rt9wfg)) - Tags: 👻 ai highlighted - better and providing better user experience. And the fourth one is really this uh website experience gap. I think even humans browsing a website now expect much richer features the content to essentially be alive and individualized. Yet most sites today are kind of static or fairly static and not very individualized. All right. By now I think a good overview has been established on how AI is rewriting some of the rules and some ([View Highlight](https://read.readwise.io/read/01kr1zbnjgyxxhr9jy070hs06q)) - Tags: 👻 ai highlighted