Two Podcasts Four Cups - CourtneyR.dev's 2nd Brain

Two Podcasts Four Cups ![rw-book-cover](https://img.youtube.com/vi/pKlDPwIOoQ8/maxresdefault.jpg) > [!note]+ Document Notes > --- > tags: #devrel #programming #community #ai #career-development #professional-development > author: Two Podcasts Four Cups > source: YouTube > reference: https://www.youtube.com/watch?v=pKlDPwIOoQ8 > date created: 2024-12-19 16:00 > --- > # 🏷️ Title: Two Podcasts Four Cups - Programming Career Development and AI Evolution > ## 🧰 Action Inbox > - [H1] Research OpenClaw setup for automating development workflows — 05:23 #ai-tools > - [H2] Evaluate proper AI mental models for developers — 38:45 #ai-adoption > - [H3] Investigate entry-level hiring trends in AI-augmented development — 35:12 #career-development > ## 🧭 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 > - [Two Podcasts Four Cups](https://www.youtube.com/watch?v=pKlDPwIOoQ8) > - [nerdy.dev](https://nerdy.dev) > ## 📝 Summary > - Leon Noel and Danny Thompson discuss the evolution of AI in software development and its impact on entry-level careers > - Key topics include git workflows (rebase vs merge), AI tool adoption patterns, and changing interview processes > - Both emphasize that AI creates more work opportunities rather than eliminating jobs, especially for junior developers > - Discussion covers practical AI implementations including OpenClaw automation and cloud agents > - Strong focus on community building and helping underrepresented people break into tech > ## 📖 Main Content > Leon Noel (100 Devs, Resilient Coders) and Danny Thompson (Elastic) join the show to discuss how AI is transforming software development careers. They argue that rather than eliminating jobs, AI tools are creating demand for more developers who can effectively guide and manage AI-powered workflows. The conversation covers technical topics like git workflows and practical AI implementations, while emphasizing the continued importance of human guidance and taste in the development process. > Both guests are heavily involved in helping underrepresented communities break into tech, providing perspective on how the industry landscape is evolving for newcomers. They discuss specific examples of companies reopening junior developer hiring and the need for new interview processes that focus on AI interaction skills rather than traditional coding challenges. > ## 🗒️ Notes > #### ⏱️ [00:00](https://youtu.be/pKlDPwIOoQ8?t=0) — [H1] Introduction and Japanese Whiskey Tasting > - Key idea: Setting context with Matsui Sakura single cask whiskey review > - Evidence: "100% malted barley, 96 proof and aged in Japanese cherrywood" > #### ⏱️ [03:45](https://youtu.be/pKlDPwIOoQ8?t=225) — [H2] Toast to Inclusive Technology > - Key idea: Focus on building accessible and inclusive technology > - Evidence: "To building tables where there were once only walls and architecting spaces that leave no one behind" > #### ⏱️ [08:30](https://youtu.be/pKlDPwIOoQ8?t=510) — [H1] Git Rebase vs Merge Debate > - Key idea: Danny advocates for git rebase to maintain clean linear history > - Evidence: "I should never merge multiple opinions into one conversation unless it's from the person that's giving that opinion" > #### ⏱️ [12:15](https://youtu.be/pKlDPwIOoQ8?t=735) — [H2] AI Commit Message Generation Discussion > - Key idea: Tension between AI automation and developer control > - Evidence: "Oh, I pushed a release for you with a new tag. You're welcome." And I went, "Fuck what? No, no, don't do that" > #### ⏱️ [18:45](https://youtu.be/pKlDPwIOoQ8?t=1125) — [H1] AI Evolution in Software Development > - Key idea: Major shift in AI usefulness over past year > - Evidence: "There was just one day where it was just not the worst thing to use anymore" > #### ⏱️ [22:30](https://youtu.be/pKlDPwIOoQ8?t=1350) — [H2] OpenClaw Automation Setup > - Key idea: AI agents working autonomously during downtime > - Evidence: "When I wake up in the morning, I don't look at my social networks anymore. I wake up and I look at what OpenClaw did" > #### ⏱️ [28:15](https://youtu.be/pKlDPwIOoQ8?t=1695) — [H1] Changed Developer Workflow > - Key idea: Adam's job has fundamentally changed to technical steering > - Evidence: "My job has changed more in the past few months than it's ever changed in my entire career" > #### ⏱️ [35:12](https://youtu.be/pKlDPwIOoQ8?t=2112) — [H1] Junior Developer Demand Increasing > - Key idea: Companies hiring 3x more junior developers due to AI productivity > - Evidence: "IBM saying they're going to hire 3x the amount of juniors they have in the past" > #### ⏱️ [38:45](https://youtu.be/pKlDPwIOoQ8?t=2325) — [H1] Interview Process Evolution > - Key idea: LeetCode becoming obsolete, focus shifting to AI prompting skills > - Evidence: "Hacker Rank even said that they see the entire interview process changing and that leak code is now going to be defunct dead" > #### ⏱️ [42:30](https://youtu.be/pKlDPwIOoQ8?t=2550) — [H2] Developer Mental Model Problem > - Key idea: 97% of developers using AI incorrectly with transactional approach > - Evidence: "Every single AI tool that I use... said I was in top 1%" > #### ⏱️ [48:15](https://youtu.be/pKlDPwIOoQ8?t=2895) — [H2] Corporate AI Adoption Failures > - Key idea: Companies buying licenses without training or context > - Evidence: "I bought 700 licenses for our engineers and Danny the only thing I did was create a seven-figure bill" > #### ⏱️ [52:45](https://youtu.be/pKlDPwIOoQ8?t=3165) — [H2] AI Infrastructure vs Models > - Key idea: Success comes from infrastructure around models, not just models themselves > - Evidence: "The way things shift is the infrastructure around the model that manipulates the prompt and the response" > #### ⏱️ [56:30](https://youtu.be/pKlDPwIOoQ8?t=3390) — [H1] Claude vs Pi Comparison > - Key idea: Trade-off between high floor/low ceiling vs trainable systems > - Evidence: "Claude has a high floor... However, it has a small ceiling... Pi has a lower floor... but as you work with it, as you teach it your way" > #### ⏱️ [60:15](https://youtu.be/pKlDPwIOoQ8?t=3615) — [H2] Cloud Agents and Long-Running Tasks > - Key idea: Evolution toward multi-hour autonomous development sessions > - Evidence: "They reduced they put out long-running agents. So now it's kind of like a Ralphie or a Ralph loop but in the cloud for you that can run seven, eight plus hours" > ## 📚 References > - [nerdy.dev](https://nerdy.dev) - Adam's blog about token dashboard and OpenClaw setup > - [100 Devs](https://100devs.org/) - Leon Noel's free software development program > - [Resilient Coders](https://resilientcoders.org/) - Leon's organization helping underrepresented communities > ## 🗣️ Paraphrase > - AI is creating more development work rather than eliminating it, leading to increased demand for junior developers who can guide AI workflows effectively > - Most developers (97%) are using AI incorrectly with a transactional mindset rather than building collaborative relationships with AI tools > - The interview process is evolving away from algorithmic challenges toward assessing AI interaction and prompting skills > - Successful AI adoption requires proper mental models and company-wide training, not just purchasing licenses > ## 📰 Newsletter > 🤖 **AI Is Creating More Dev Jobs, Not Fewer** > Contrary to doom predictions, AI tools are driving demand for junior developers. Leon Noel and Danny Thompson share insights from the trenches: IBM is hiring 3x more juniors, and companies are realizing they need human guidance for AI workflows. > **Key takeaway**: The bottleneck isn't coding speed—it's having skilled developers who understand how to direct and collaborate with AI systems effectively. > **Action items for WordPress developers**: > - Experiment with AI tools like Cursor, OpenClaw, or cloud agents for your workflow > - Focus on developing "AI collaboration" skills alongside traditional coding > - Consider how AI might enhance rather than replace your WordPress development process > The future belongs to developers who can art-direct and technically steer AI tools, not those who compete with them on raw coding speed. > Resources: [Learn WordPress](https://learn.wordpress.org/), [Developer Docs](https://developer.wordpress.org/), [WP-CLI](https://wp-cli.org/) > ## ✅ Action Items > #### DevRel > - Developer advocacy: Create content around AI-assisted WordPress development workflows > - Education: Develop workshops on proper AI mental models for developers > - Developer experience: Evaluate AI tools integration in documentation and tutorials > - Community management: Facilitate discussions around AI adoption best practices > #### WordPress > - Core: Assess impact of AI-assisted development on contribution workflows > - Community: Create guidance for meetups on AI tools and career development > - Training: Develop lesson plans incorporating AI collaboration skills > - Documentation: Update developer guides to include AI-assisted development patterns > #### Community > - Contributors: Establish mentorship programs pairing AI-experienced developers with newcomers > - Sponsors: Fund training programs focused on AI collaboration skills > - Ecosystem: Coordinate cross-platform discussions on AI adoption patterns > ## 🧪 Hypotheses > - AI adoption success correlates more with mental model training than tool selection > - Companies investing in AI collaboration training see higher productivity gains than those just buying licenses > - Interview processes focusing on AI interaction skills better predict job performance than traditional coding challenges > - Junior developers with AI collaboration skills have higher hiring rates than those without > ## 📊 Methodology > - Survey companies on AI tool adoption patterns and training approaches > - A/B testing of interview formats (traditional vs AI-focused) and subsequent job performance > - Longitudinal study of developer productivity before and after AI collaboration training > - Analysis of job posting requirements evolution over time > ## ❓ Questions > - How can WordPress core development processes adapt to AI-assisted contributions? > - What specific training curricula produce the most effective AI collaboration skills? > - How do we balance AI efficiency with the learning needs of junior developers? > ## 🔮 Future Research > - Long-term impact of AI-assisted development on code quality and maintainability > - Evolution of senior developer roles toward technical steering and architecture > - Development of standardized AI collaboration competency frameworks > ## 💡 Implications > - Practice: Development workflows shifting toward AI collaboration and technical steering > - Policy: Need for updated hiring practices and competency frameworks > - Education: Curriculum redesign to include AI collaboration alongside traditional coding skills > - Community: Mentorship programs must evolve to include AI tool guidance > ## 🧠 Learning Styles Assessment > Reading level estimate: 12th Grade > #### 👀 Visual learners > - Present: Workflow descriptions, tool comparisons > - Improve: Architecture diagrams of AI-assisted development processes, visual tool comparison charts > #### 🔊 Aural learners > - Present: Podcast format with conversational explanations > - Improve: Transcripts with timestamps, audio summaries of key concepts > #### 🗣️ Verbal learners > - Present: Rich analogies (puppy training, art directing), technical definitions > - Improve: Glossary of AI development terms, story-driven case studies > #### 🧩 Kinesthetic learners > - Present: Hands-on tool examples, workflow demonstrations > - Improve: Step-by-step AI tool setup guides, interactive exercises > #### 🧠 Logical learners > - Present: Systematic comparison of tools and approaches > - Improve: Decision trees for tool selection, metrics for measuring AI collaboration success > #### 🧑‍🤝‍🧑 Social learners > - Present: Community-focused discussion, collaborative insights > - Improve: Discussion guides for development teams, peer mentoring frameworks > #### 🧘‍♀️ Solitary learners > - Present: Personal workflow optimization strategies > - Improve: Self-assessment tools for AI collaboration skills, individual learning paths > #### 🌿 Naturalistic learners > - Present: Organic evolution of AI adoption patterns > - Improve: Ecosystem thinking about AI impact, sustainable development practices > ## 📊 Accessibility Assessment > #### Screen reader semantics > - Present: Clear heading structure, descriptive content > - Improve: Enhanced alt text for technical concepts, structured data markup > #### Keyboard navigation > - Present: Standard video player controls > - Improve: Transcript navigation shortcuts, topic jumping functionality > #### Contrast and readability > - Present: Standard video player interface > - Improve: High-contrast transcript options, adjustable text sizing > #### Media and captions > - Present: Audio content available > - Improve: Professional captions with technical term accuracy, downloadable transcripts > ## 🌍 DEIB Assessment > #### Cultural breadth > - Present: Focus on underrepresented communities in tech, global perspective on development > - Improve: Examples from diverse geographic regions, multilingual considerations > #### Inclusive language > - Present: Respectful discussion of career challenges, accessible explanations > - Improve: Gender-neutral examples, plain language alternatives for jargon > #### Representation > - Present: Two experienced professionals from underrepresented backgrounds sharing insights > - Improve: Broader range of career stages and backgrounds in examples > ## ✅ Logical Validity > #### Strong reasoning > - Deductive strength: Clear progression from AI adoption patterns to career implications > - Inductive strength: Multiple company examples supporting hiring trend claims > - Causal reasoning: Logical connection between AI productivity and increased hiring needs > #### Fallacies > - Survivorship bias: Focus on successful AI adopters may miss broader challenges > - Hasty generalization: IBM example extrapolated to industry-wide trend > #### Tone of voice > - Overall tone: Enthusiastic and optimistic about AI potential while acknowledging challenges > ## 🛡️ DARVO Analysis > - No DARVO patterns detected - discussion maintains positive framing focused on opportunity and growth > ## 🎯 Conclusion > The conversation reveals a fundamental shift in software development where AI tools are creating more opportunities rather than eliminating jobs. Success depends on developing proper mental models for AI collaboration and companies investing in training rather than just purchasing licenses. For WordPress developers and the broader community, this suggests focusing on AI collaboration skills while maintaining the human elements of taste, guidance, and technical steering that remain irreplaceable. > ## 🧱 Refactor Appendix > ### 📚 Literature Split Pad > ### [H1] AI adoption requires mental model training not just tool access > > [!quote] [Two Podcasts Four Cups](https://www.youtube.com/watch?v=pKlDPwIOoQ8) by Two Podcasts Four Cups via YouTube. Published 2024. Accessed 2024-12-19. > > "97% of people using AI tooling they're not using it correctly" > Danny Thompson's research reveals most developers use transactional approaches with AI rather than collaborative mental models. > [view highlight](https://youtu.be/pKlDPwIOoQ8?t=2325) #xref #articleseed ID:rw_ai_mental_models > ### [H2] Companies hiring more junior developers due to AI productivity gains > > [!quote] Two Podcasts Four Cups > > "IBM saying they're going to hire 3x the amount of juniors they have in the past" > AI creates more work opportunities rather than eliminating them, requiring human guidance for effective implementation. > [view highlight](https://youtu.be/pKlDPwIOoQ8?t=2112) #xref ID:rw_junior_hiring_trend > --- > ### 🔖 > --- > ## Full Transcript > [00:01] Howdy y'all. [00:03] Hey. Hey. We have two very important [00:06] guests with us today. Adam, could you [00:07] tell the folks at home who we've got on [00:09] the show today? Oh, absolutely. Both [00:11] these cats coming to us from the [00:13] programming podcast. In this corner, [00:15] Leon Noel, managing director of [00:17] engineering at resilient coders, founder [00:19] and leader of 100 devs, distinguished [00:21] faculty at general assembly, lecturer at [00:23] Harvard Business School, [00:25] guest lecturer at MIT Sloan [00:27] entrepreneur, educator, and turbo [00:28] booster of hundreds of underrepresented [00:30] folks looking to break into web tech, [00:31] and soon to be my Pokémon Go friend. I [00:34] as long as you're still playing. And I [00:36] wonder if you play Monster Hunter. [00:37] Questions for later. And in this corner, [00:39] we have Danny Thompson, senior developer [00:40] advocate at Elashian. Speaker, who's [00:43] done over 200 talks. What the [ __ ] 200 [00:45] talks? That [ __ ] is crazy, dude. [00:46] Workshop runner, community grower, event [00:48] thrower, meetup maker, known to go from [00:50] gas stations to billboards, is another [00:52] rad turbo booster of hundreds of [00:54] underrepresented people looking to break [00:56] into web technology. And did I stumble [00:58] upon a book that you're writing? You [00:59] better tack on. Author to that list. [01:01] Welcome to the show, y'all. Stoked to [01:03] have you. [01:04] >> Thanks for having us. I'm stoked, too. [01:07] Let's go. [01:08] >> I was actually scared he was going like [01:10] and here's Danny Thompson like after [01:11] that. [laughter] [01:15] >> Oh, that would have been a good joke. [01:17] >> That's [laughter] literally where I was [01:18] going. I was like I was so ready for [01:20] that. But uh thank you, man. Did some [01:23] research. I I got to get [01:25] >> I appreciate it. [01:27] >> How y'all doing? [01:29] >> Yeah, doing good. Doing good. So, as the [01:31] name of this podcast entails, we always [01:34] start with whiskey. We've got a Japanese [01:38] one today. The Metsui single cask [01:41] Sakura. [01:43] Uh it is 100% malted barley, 96 proof [01:46] and aged in Japanese cherrywood. Uh [01:50] which I guess is Sakura. I don't know. [01:51] I'm uneducated. Um and maybe X bourbon [01:54] barrels as well, depending on where you [01:57] ask on the internet. So [01:59] >> So I just opened the box. I didn't get [02:01] any notes of anything. Is that supposed [02:03] to happen? [laughter] [02:05] Yeah, you got to open the bottle first. [02:07] >> It's not like a a letter from junior [02:09] high girlfriend or whatever that like [02:11] perfume on it, you know. [02:13] >> I wouldn't know about that either, [02:14] unfortunately. But regardless, [02:15] [clears throat] [02:16] let's see. [02:18] >> Uh some funny other backtory here. Also, [02:20] it's Sakura, by the way. Uh [02:22] >> whatever. No, it's cool. I like how you [02:23] said it, too. [02:24] >> Okay. [02:25] >> Um [laughter] yeah, this is a a [02:28] redemption arc. This company was uh [02:30] before like a lot of Japanese whiskey [02:32] apparently is imported, blended in Japan [02:34] and then exported. So, it's not even [02:36] Japanese whiskey. And so, these folks [02:38] have been teased for years about doing [02:39] that. They just slap a a label on [02:41] something that's not actually Japanese [02:43] whiskey. So, anyway, this is their [02:44] redemption arc. Uh so, now that they get [02:47] to be like, "Oh, yeah. Well, now we're a [02:48] distillery at the foot of a snowcapped [02:50] mountain called Mount Dyson, and these [02:52] are single cherrywood casks, hence the [02:54] name Sakura. and they have ultra soft [02:57] ground water that's filtered through [02:58] volcanic soil. Uh the same aquafer is [03:01] what softens their whiskies and lets the [03:03] Sakura influence shine instead of [03:06] turning overly bitter. There's no added [03:08] color. So anyway, [03:09] >> [ __ ] Pete. Yeah, [03:10] >> you know how I know this is an expensive [03:11] bottle? There's there's actual ink from [03:14] a handwritten pen at the bottom. [03:16] >> That's like extra. [03:17] >> You got to pay extra for that for that. [03:19] >> Oh, I got 290 out of 300. [03:21] >> I got 289. Sucks to be you. 294. [03:25] >> Closer to the original 156. [03:28] >> Whoa. Hey, [03:30] >> he went. [03:33] >> You got a toast for us, Adam, before we [03:34] drink? [03:34] >> I do have a toast. To the craft of the [03:37] code and the warmth of the poor. Here's [03:39] to building tables where there were once [03:41] only walls and architecting spaces that [03:43] leave no one behind. May the web be [03:45] built uh may the web we build be as [03:48] accessible, open, and inviting as a [03:50] shared bottle amongst friends to [03:52] breaking down barriers, debugging our [03:53] biases, and lifting up the users who [03:55] need it most. Cheers. [03:57] >> Cheers. [03:58] >> Cheers. [03:59] >> Yeah. [04:01] >> Smells really good. [04:07] >> I smell limes. [04:10] >> Limes? [04:11] >> Yeah. I don't know. [04:12] >> Sharp at first. I like it. [04:13] >> Oh, yeah. pretty stringent. I was [04:14] expecting more like vanilla. [04:17] >> They said golden honeylos. [04:18] >> Oh, were we not supposed to drink it [04:19] yet? My bad. [04:20] >> Don't do it, dude. [04:21] >> No, you can. [laughter] [04:23] >> Whatever you want. [04:23] >> I heard everybody still talk about the [04:24] smells like Oh, no. [04:25] >> You go through phases. It's okay. [04:28] >> Oh, the taste is much different. [04:30] >> Tastes a lot scotchier. Got that classic [04:33] Sharpie taste. [04:35] >> I like it. The scotchy peiness. Give it [04:38] to me. But it's sweet. It's got a [04:40] sweetness to the pee. Sweet pee. [04:45] Yeah, [04:46] >> it's alcohol. [04:47] >> It sure is. [04:49] >> Looks like like a cheaper mullen. [04:51] >> Yeah, [04:53] >> it's [snorts] smoother mullen. [04:56] It's It's smoother than I was [04:58] anticipating, which is good. Um, [05:01] the one thing that I find interesting is [05:03] like I always look for I don't know if [05:05] there's like a special name for this, [05:07] but when I breathe in again, does it [05:08] feel like a dragon is there? And it [05:10] does. [05:11] So, like there I'm sure that's [05:13] sophisticated um liquor talk right [05:15] there, but it actually tastes pretty [05:18] decent. [05:19] Uh I'm I'm thoroughly surprised. [05:22] >> Chuck calls that the hug. If you can [05:24] feel it burning you and it's the hug, [05:27] >> but wait, I think I know what he's [05:28] saying though, which is different. Like [05:29] sometimes I'll take a sip and yeah if [05:31] you breathe in oxygen over top of it [05:33] there is an aromatic it and it fills [05:36] your mouth and your lungs and it can be [05:38] kind of burny but also kind of like the [05:39] hug where you're like the hug is nice. [05:41] There's a uh a vaporized hug uh type of [05:46] vibe. I like it. The fire dragon is very [05:48] nice. [05:50] >> Yeah, this is good. [05:53] >> Yeah. Um, okay. So, we have a very [05:57] stringent rating system on this show. It [05:59] is zero to eight tentacles. Zero being [06:02] the worst. Wouldn't even clean your [06:04] toilet with it. Eight being the best. [06:06] Uh, you don't want to drink anything [06:07] else ever again. Four and a halfish is [06:11] um average. Middle of the road. [06:13] >> Um, [clears throat] [06:14] >> what has been an eight for you in the [06:15] past? [06:16] >> Good question. [06:17] >> We've only had like three eights, I [06:19] think, and I don't know. I couldn't tell [06:21] you what they are. Um, but [06:23] >> they were so good they were unmemorable. [06:26] >> Well, no. I mean, they're memorable, but [06:27] we've had 240 [06:30] whisies or whatever. I'm like, [06:31] >> was Lug Villain uh 16 that recently? Did [06:34] I give that an eight? [06:36] >> I think it was high, but I don't think [06:38] you've given anything an eight yet. I [06:39] think it was like 7.75. [06:41] >> Where does Jack Daniels end up on this [06:43] rating scale for you? [06:44] >> Not high. Not high. [06:46] >> Four. [06:47] >> Three. Three, probably. [06:49] >> Whoa. Okay. Okay. Okay. It's funny. [06:52] >> Yeah, we've had zero. [06:53] >> I'm on roller skates, you know, but [06:54] other than that, I don't want it. [06:58] >> Yeah, there was this one that was um it [07:00] was called the Daringer, and I bought it [07:03] only because it had a pistol cork, which [07:05] I thought was really cool, and it was [07:07] the worst whiskey I've ever had. It is. [07:09] >> I believe that. [07:10] >> Yeah. Terrible. [laughter] [07:13] >> Anyway, Adam, you want to show us how [07:15] the the rating goes? [07:17] >> Uh yeah. So, I'm I'm I'm really loving [07:20] scotches right now. I've been drinking a [07:22] lot of them. So, it's kind of uh [07:24] interesting to have one that's from [07:25] Japan. Uh literally from Japan, so it's [07:28] not just blended there. And I think [07:30] they're hitting on all the notes. [07:32] There's like a sparkliness to it. And I [07:34] I like the natural color, the natural [07:36] sort of flavoring. It feels very [07:37] authentic. Um I'm going to give this a [07:40] seven. [07:43] >> Seven. Wow. Nice. It's pretty good. Uh [07:46] yeah. Who wants to go next? [07:48] Well, I'm kind of glad that I I was [07:50] afraid we're going to do this like [07:51] Planet Poker and send like our ratings [07:53] in the chat just so there's no bias. [07:55] Glad that didn't happen. [07:57] >> I do like this. [clears throat] [08:00] It isn't the best thing that I've ever [08:01] had. It is very smooth compared to what [08:05] based on the original smell I thought it [08:07] was going to be. Uh the dragon fire [08:10] breath is definitely there. So, your [08:13] first inhale after taking a drink will [08:15] feel sharp. So, I'm not I like it, but I [08:18] also dislike how strong it is. And it [08:20] feels a little bit warmer, a little bit [08:24] longer than what I would like for each [08:25] sip. Uh, it almost feels like you're [08:27] taking like some really strong cough [08:29] syrup that's like opening up your chest. [08:31] And I kind of like the warm feeling, but [08:33] not to that effect. To me, it's like a [08:36] five and a half. I would definitely have [08:38] it again. I think it's above average. I [08:41] do think like if somebody had it, I [08:43] would definitely drink it with them. I [08:45] don't know if I would feel compelled to [08:47] go out of my way to go buy this bottle [08:49] though. [08:52] >> Yeah, that's fair. I feel similar. [08:56] I big fan of Japanese whisies. I was [08:59] telling Off Pod Hiki Harmony is one of [09:01] my favorites for a long time. And so [09:03] there's definitely some notes of like [09:04] the Hibiki harmony coming through. [09:08] I think I I would put it above a four. [09:11] or so I'm feeling like a strong five [09:13] just because it's something that I would [09:15] recommend people try especially if you [09:17] haven't tried any of the Japanese [09:18] whisies. It's definitely worth the [09:20] drink. Um definitely the slight I don't [09:25] like the aftertaste on it as much as I I [09:27] thought I would. The smell is very [09:29] different than the taste too. I'm not [09:31] getting any of like the vanilla that I [09:33] thought I would get. Uh very [09:36] very mild. But overall I enjoy it. I [09:38] would say a strong five. Definitely if [09:40] you see it, I would try it. Um worth [09:42] worth a taste. [09:45] >> Yeah. Um I'm kind of similar in that [09:48] like it's above average, but it's [09:51] definitely not my favorite. Um [09:54] I don't know. I think I'm going to Yeah, [09:55] I think a five is appropriate. Um I [09:58] don't know. Don't can't tell you why. [09:59] Don't super love it. [laughter] [10:03] >> Dragonfire is real, as you call it. [10:05] [laughter] [10:06] I didn't have a more sophisticated name, [10:08] so that's what I'm rocking with. [10:09] >> What scotches have you been drinking [10:10] recently, Adam, that you've liked? [10:13] >> Some Highland ones. Uh, and then the [10:15] Logan. So, the Logan being super smoky [10:18] and then the Highland ones from H. [10:20] What's the area? There's an area. [10:22] Anyway, they have very similar [10:24] qualities, similar color, similar [10:26] tingliness on my tongue. uh peey and [10:30] bright I would say as opposed to like [10:32] syrupy or bourbony or sugary and I've [10:36] just kind of like I'm a very earthtoned [10:38] flavor palette type of person. I drink [10:40] my coffee black. Um I I drink tea just [10:43] as it is. Like if you give me a tea I'm [10:45] like I want to taste the like leaves and [10:47] [ __ ] like don't don't mask this. And so [10:49] it's kind of what I like about scotches [10:51] is they bring me back to that sort of [10:53] like I'm drinking something green. I'm [10:55] drinking something earthy. Uh, I want [10:57] the bitter and natural flavors. I don't [11:00] need any disguises, you know? I'm like, [11:01] just give me as you are. I want to I [11:03] want to discover your journey on my [11:05] tongue. That's weird to say, but Danny's [11:08] nodding like I [laughter] made sense. [11:10] That's great. Thank you. I appreciate [11:11] that. [11:11] >> I I'm in a similar light. I I like uh [11:14] especially when it comes to beer, I like [11:15] the more bitter beers than I do um the [11:18] lighter tasting ones. Uh when it comes [11:21] to liquor, I'm on a similar plane as [11:23] well. Like I I like the I guess for the [11:26] lack of a more sophisticated terms, the [11:28] earthier tasting ones versus the sweeter [11:32] ones. Um but like then I say that and [11:35] then it's like I like Jack Daniels, but [11:37] I guess that's also the Memphis [11:39] Tennessee side kicking in with that one. [11:43] >> Yeah, Lugaboon 16's definitely one of my [11:46] core three in rotation. [11:48] Um, yeah. [11:50] >> But that earthiness is what my family [11:52] and friends label as the band-aid [11:54] flavor. [laughter] [11:56] >> Really? [11:56] >> I love it. I love it. [12:00] >> Yeah, the 16 was pretty good. Um, a lot [12:03] of them are just really smoky or really [12:06] pey and I feel like that was just nicely [12:08] balanced. [12:10] >> Although that one was made as like a [12:11] smoky joke and then everyone was like, [12:13] "Bring it on. [laughter] [12:16] >> [snorts] [12:18] >> All right. [12:18] >> Delicious. [12:20] >> So, [12:20] >> well, I will say just for argument sake, [12:23] I decided to throw some coconut. [12:25] >> I can't even taste the liquor. [12:27] >> Dang. [12:27] >> Uh, I guess the sweetness of it mixes [12:30] really well with it. I can't taste it. [12:32] And I'm not like one of those like, "Oh [12:34] my god, Bikardi tastes like Sprite." [12:35] It's like, "No, Bartiardi tastes like [12:36] Bikardi." But like this legitimately, I [12:39] can't tell that there's alcohol in this [12:41] right now. [12:43] >> Well, hey, good mixer then, I guess. [12:45] >> Yeah. All [12:47] right. So, we have not done hot takes in [12:50] a while, and all of them are kind of [12:53] lukewarm at this point, but just going [12:56] to do a couple of the favorites just for [12:58] fun. All right. Uh, get rebase or get [13:01] merge. [13:03] Oh, no. I've heard your takes on the I [13:05] but Well, I don't know if I heard [13:07] everybody's takes on this. I for I think [13:09] I joined one live episode that you did [13:11] and you were discussing this. I'm a get [13:13] rebase kind of person. [13:15] And it's actually I'm one of those [13:19] weirdos that actually looks at the [13:20] history and so that makes sense why. But [13:26] one of the best engineers that I ever [13:28] worked with in the entirety of my life. [13:31] He is now retired and he is actually one [13:35] of the strongest reasons why I truly [13:37] understand Golang at a very deep level [13:39] where I was very like proficient with [13:42] it. Um, I was like architecting some [13:44] amazing solutions with Golang and a lot [13:46] of that came because of his guidance and [13:48] he instilled rebasing into the entire [13:51] team and I was so against it because I [13:54] was a get merge kind of guy and once I [13:56] finally felt okay I see that I see the [13:59] light at this point and the light to me [14:01] is this [14:03] I should never merge multiple opinions [14:06] into one conversation [14:09] unless it's from the person that's [14:10] giving that opinion. Right? Then I can [14:12] take the nuance. I can take all the [14:13] information. If I'm just merging your [14:16] intelligence of a take with my lack of [14:19] intelligence on what your nuance should [14:21] be, that's a bad thing to mix together, [14:23] right? Or [14:26] if I rebase it, [laughter] [14:32] I have a liquor mixer that I can't even [14:36] taste. And so I get all the history. I [14:38] get all your information. I get [14:39] everything that you wanted. I could [14:40] squash it if I need to, but rebasing by [14:43] itself, I think I'm big on keeping that [14:45] linear history, and I know the take a [14:47] lot of people have is like, who really [14:48] looks at the history? You might not look [14:50] at it today, but when you finally do [14:52] need to go back and find that commit [14:53] that broke something and then fix [14:54] everything, you would be very [14:56] appreciative of having a history that is [14:58] in line. You know why things are [15:00] happening and it's easy to follow. And [15:02] so for that, I'm on the rebase train. [15:05] >> I don't know if it's the the dragonfire [15:07] speaking, but whatever 4.6 six or codeex [15:10] tells me to do is what I do these days. [15:11] Who gives a crap? [laughter] [15:13] >> Oh, man. [15:13] >> 4.6 loves the rebase. So, that's what [15:16] we're going with. [laughter] [15:17] >> Well, wait. You let your Here's the new [15:19] hot take. I feel like Stop me if I'm [15:21] wrong, but do you let AI get commit for [15:23] you? This is a question. [15:25] >> Not a dinosaur. What are you doing? [15:28] [laughter] [15:28] >> This is a debate that we've been having. [15:31] >> Yeah, I think it really depends how much [15:33] I trust what it's doing at the time. Um, [15:37] a lot of times I'm not wanting it to and [15:39] then it like most of the time when I [15:41] don't want it to, it decides that that's [15:43] when it's going to do it. It's like when [15:44] I actually don't care is not the time [15:46] that it's going to do it. But when I'm [15:47] like experimenting with a thing, it's [15:49] like, "Oh, like I was working on a some [15:51] open source project. I forget what it [15:52] was." And it was like, "Oh, I pushed a [15:54] release for you with a new tag. You're [15:55] welcome." And I went, "Fuck what? No, [15:58] [laughter] [15:59] >> no, don't do that." [16:01] >> Um, so yeah, it was uh [16:04] >> it can be bad. So, make sure you let it [16:06] know right up front what you would [16:08] prefer. [laughter] [16:09] >> Let I have a hot take that we can go [16:11] into after this, but I'm curious on Adam [16:13] and Robbie's take on the get rebase, get [16:16] merge. [16:18] >> Robbie, take it away first. I'll I'll [16:20] follow up. [16:21] >> Uh, yeah, I'm rebase for the same [16:23] reasons that you said, Danny. like um [16:26] you know the people that say the history [16:28] doesn't matter have never looked at 500 [16:31] commits in a row of fix this change that [16:33] do this save work in progress and I [16:36] don't want that so yeah I always want [16:38] each PR should be one commit and I don't [16:41] even want the second commit that says [16:42] you merged this commit so I want just [16:45] that one and that's why I rebase [16:47] >> this is why he was one of our earlier [16:49] guests on the podcast Robbie I see you [16:51] Robbie okay [laughter] [16:52] >> no pressure now based on that I feel [16:53] like Adam is about uh like making you [16:56] feel heartburned. But let's continue. [16:58] >> No, I'm I'm very much a rebaser. But I [17:00] will I will admit though that I was [17:02] like, "Hey, what's the shame of like a a [17:04] get merge if you're just going to squash [17:05] me anyway?" So look, if you're going to [17:07] take my PR and squash it into a single, [17:09] who gives a [ __ ] if I merged or not? [17:11] Like, right? Like, no one cares if [17:14] you're talking about a clean history. [17:15] The history is clean on the main. But [17:17] like, hey, my work in progress, you got [17:19] to see how I worked. Hey, I pulled in a [17:20] merge. I merged right now. You all [17:22] merged, too, when you rebased. you just [17:24] struggled through it more, right? You [17:26] were just like, I need to merge, but [17:28] make it look like I didn't merge. I'm [17:30] going to rebase. Uh, this is how I do it [17:32] by acting it's I'll deal with all the [17:35] conflicts which are really complex, but [17:36] I'll just do it anyway. Anyway, I'm [17:38] still on rebase. Um, but I have recently [17:40] been very much on the team like Work [17:42] Tree and um stack team uh whether it's [17:46] through like Git Butler or Graphite as [17:49] these kind of two like meta scenarios on [17:52] top of Git. And I'll admit in the [17:54] beginning I didn't like it very much [17:56] like rebase treated me but over time I'm [17:59] all on on like team stacks uh and team [18:02] work trees. These are really ways to [18:04] like liberate your way with git. They're [18:05] just abstractly a little bit more [18:07] intense. And so uh consider me rebase [18:10] but also consider me uh dabbling in [18:13] stacks and work trees. How about y'all? [18:19] >> I don't know if I care enough about that [18:21] one. Uh, I don't want to say I don't, [18:24] but I also haven't really given it much [18:25] thought. So, that's interesting. You're [18:27] you're making me think through it right [18:29] now. And I don't have a take on this at [18:32] this very moment, but give me a couple [18:34] more sips and maybe I will, but right [18:36] now, [18:37] >> yes, the heat, the dragon heat, it'll [18:38] hit you. [18:39] >> The dragon heat. [18:39] >> I feel like work trees were really [18:41] important to me for like one month. I [18:42] saw I talked to Danny about for like at [18:43] least a solid month. Work trees, work [18:45] cheese, work trees, and then cloud [18:46] agents kind of took over for me and now [18:49] they're just on their own machine. So, [18:50] it's like, "All right, you can go do [18:51] whatever the heck you want to do, and [18:52] then we'll figure it out later." Which [18:54] is why like I don't mind giving up the [18:56] rebasing decision when I have four [18:58] different cloud agents coming all back [19:00] together with random stuff. [19:01] >> That'll make sense. Wow, that's an [19:02] awesome take. Very modern. Love [19:05] You know, it's interesting where this [19:08] conversation is going because 11 months [19:10] ago, we had a guest on our podcast and I [19:13] we were talking about how the CEO of AI [19:17] company said that 90% of code will be [19:19] written by AI um tooling. [19:22] >> Did you hear Hot Pockets joke? [19:25] >> A what? [19:26] >> The Hot Pockets joke. The CEO of Hot [19:28] Pockets says that in 2026 all meals will [19:31] be Hot Pockets. [laughter] [19:33] M [19:34] >> true facts. [19:35] >> But like looking back on this 11 months [19:38] later, do you still feel that figure is [19:40] off [19:43] >> that are not all the meals or [19:47] >> Well, yeah, obviously that one like [19:49] >> Yeah. No, I think we've been talking [19:51] about this like, you know, we've talked [19:53] almost exclusively AI for our last like [19:55] I don't know how many episodes because I [19:58] think in the past [20:00] >> I don't know 3 to six months. It's hard [20:01] to say exactly the time frame or what [20:04] because it all moves so fast, [20:05] >> but there was just one day where it was [20:07] just not the worst thing to use anymore [20:10] >> and I was like, "Huh, okay. I can [20:13] actually use this." And we talked to uh [20:15] Zach from Warp about this where it was [20:16] like he had this uh post on LinkedIn [20:19] where he was like, "Oh, um how come my [20:21] seniors will never start with AI first? [20:23] They just want to code it. They don't [20:24] want to like ask any AI." And we were [20:26] all kind of like, "Oh, cuz it's fun. I [20:28] like to code. I I can do it better. I'm [20:30] not like I don't need it." Now it can [20:33] code circles around me all day. So I'm [20:36] like, why would you not use it? Everyone [20:38] should be using it. You should be doing [20:39] like Leon was saying, like throw a ton [20:41] of agents in the cloud, let them go to [20:42] town. Like um I think the game has [20:46] changed a bit and maybe it hasn't [20:48] changed a ton but like the little ways [20:49] it has changed has made it more [20:51] approachable for agentic workflows and [20:54] like for seniors to really like take [20:56] that on and own that and like it's like [20:58] managing a team almost and you can [21:00] really build stuff that you before like [21:02] I would just be fighting with it all day [21:04] like do this and it's like cool I did [21:06] it. It's like no that's that's not it at [21:07] all. Like that's completely wrong. Oh [21:09] cool I did it. Oh no that's that's still [21:10] completely wrong. Now it's like it's [21:12] greater than half the time. It's just [21:15] almost flawless. And I'm like, "Huh, [21:17] okay." [21:19] >> Well, that's so interesting cuz on the [21:20] podcast you said this, "Produce and play [21:22] the clip." [laughter] [21:24] >> But that that was the arc of [21:25] legitimately our podcast. That was the [21:27] arc. It was like October when we first [21:29] started. I was like, "I have no idea [21:30] what you're talking about, Danny. This [21:31] is all nonsense. Like, I've never gotten [21:33] anything worthwhile from this. You're [21:35] absolutely out of your mind." And then [21:36] like 3 months later, I was like, [21:37] "Actually, wait a minute. Hold on now." [21:40] And then I think January was a big like [21:41] I don't know like December January [21:43] everything changed again for me like it [21:44] became way more useful again and so I [21:47] just keep seeing these like ratchets [21:48] that have been happening like every 3 [21:50] months where something kind of changes [21:52] and I becomes a bigger part of my [21:54] workflow than it was before. Now [21:55] >> are you open clawing? Are you got an [21:57] open claw? [21:59] >> So on purpose. [22:01] >> So here's something you have to know [22:01] about the next fun thing. It's crazy. [22:03] It's it's the best thing I've done in 15 [22:06] years of coding. It's the weirdest, [22:07] [clears throat] most addictive thing [22:08] I've done in a while. Anyway, [22:10] >> I'm really susceptible to scams. I know [22:13] this about myself. [22:14] >> I I I I know this about myself. I just [22:17] like like uh I don't know, e-commerce [22:19] stuff. Like I know it's it's got me my [22:22] whole life. [22:22] >> Yeah. You're like deflection deflection. [22:24] >> So that that's been my hesitation with [22:26] open claw. It's just like if all the [22:28] people that got me in the past are now [22:30] doing open claw, [22:31] >> man. It's the one to let in, [laughter] [22:32] dude. It's it's the one that actually [22:34] works for you. But [22:35] >> I'll take your advice. I'll stop for a [22:37] sec. [22:37] >> So, this can buy a Mac Mini to do it. [22:40] Just throw it [clears throat] in a [22:40] server on the cloud for a few bucks and [22:42] try it out. [22:43] >> We're saying Railway has it now. Like [22:45] that's that was really cool. Yeah. [22:47] >> No, we This can lead into like our next [22:49] hot takes. We just did an episode on [22:52] this recently and Leon and I just to [22:56] show by the way how far Leon has come. I [22:59] have been pro AI for a very long time. [23:01] So much so that someone on our last [23:03] podcast episode said after I referenced [23:06] the 90% they said this. I think what [23:08] you're saying makes a lot of sense and I [23:09] would equate it to like 90% of art is [23:11] made by toddlers. Right. [laughter] [23:14] Who is that? [23:17] Amazing quote because it stuck in my [23:19] head so much I had to go pull pull it [23:20] back. But Leon has gotten to the point [23:24] now where in our last episode or I think [23:28] it was two episodes ago, he talked about [23:29] how he wants to be at a stage where he [23:31] could be more handsoff with the code [23:35] where I'm of the mindset and matter that [23:38] I actually want to be super handson [23:41] while using AI as a productivity [23:44] enhancer, a force multiplier, but I'm [23:47] very very very much handson. He wants to [23:50] get to a point where he's like, "I love [23:52] this so much that I'd love on like a lot [23:54] of my side projects, whatever it may be, [23:55] where I could be more handsoff." [23:58] >> For y'all, what is your positioning on [24:00] this? Are you the hands-on? Are you the [24:03] hands off? Are you the type where it's [24:05] like, I would just love to voice to text [24:06] from my phone and be done with the [24:09] thing? [24:10] >> It depends. If I'm going to look at the [24:12] code ever again, like if it's a domain [24:14] that I have knowledge in, then I want to [24:17] be more hands-on. But whereas like I [24:19] spent a good three weeks building a um [24:23] color picker magnifier thingy in Rust. [24:27] >> Oh yeah. [24:28] >> And I don't know any Rust. So have at [24:30] it. I don't never need to see the code [24:31] as long as it works. Rust has great [24:33] errors. So like it's not going to be [24:35] broken if it compiles. So like um it [24:39] yeah like in that situation I just need [24:41] it to work. Write some tests that can [24:43] show me it will continue to work and I'm [24:46] cool. But if it's like my day-to-day [24:48] JavaScript TypeScript code that I'm like [24:51] going to look at a lot and that's like [24:53] what I live in, then I care a little [24:55] bit. I want it to be the way I would [24:57] write it is. Like it doesn't have to be [24:58] exactly the way I would do it, but it [25:00] can't be using like completely [25:01] non-standard stuff and just kind of like [25:05] throwing crazy stuff out there. [25:08] I'm uh navigating this space as I feel [25:12] like anyone else is where I'm I'm a I [25:15] have many domain expertises. I have a [25:17] lot of opinions. I have a lot of [25:19] vocabulary. I have a lot of taste and [25:22] I'm finding myself more and more um [25:25] finding myself more and more like [25:27] [snorts] [25:28] um art directing techni technical [25:31] steering. Um my job has changed. My job [25:34] has changed more in the past few months [25:35] than it's ever changed in my entire [25:37] career. And I'm not sad about it. Uh I [25:40] used to be sad about it. Last year you [25:41] would hear me crap on AI all the time. [25:43] It tried to do a thing and and then I I [25:46] evaluated and it sucked and it did suck. [25:48] And these days I'm like I evaluated the [25:50] thing. I'm like, "Oh my god, it's good." [25:51] Uh I'm at I'm at the point right now [25:53] where I'm I have there's two revelations [25:55] that have happened to me that I can't [25:57] deny. One of them is while I'm making [25:59] coffee in the morning, French press [26:00] takes about 12 minutes. I [26:02] [clears throat] can I can open and close [26:04] a PR. I can find a GitHub issue, voice [26:07] it, be like, "Hey, there's a GitHub [26:09] issue. Here's the thing. You should just [26:10] do this strategy, and then by the time [26:12] my coffee is done, I've reviewed the [26:15] code and merged it in [ __ ] GitHub in [26:18] the app on my phone." And I'm like, [26:19] "Well, that [ __ ] is crazy." The other [26:21] thing that's driving me crazy is that [26:23] OpenClaw is um now out coding me while I [26:26] sleep. And so, like, this is why I'm [26:28] like, Leon, you got to try this out if [26:30] you're into like the cloud thing. Like [26:31] this is cool. I have a I call it the [26:33] burn window, which is uh I'm paying I'm [26:35] I'm cheap. I'm paying for Claude $20 a [26:38] month, which means I have a limited [26:39] amount of things I can do every 5 hours. [26:41] The thing is is OpenClaw has cron jobs. [26:43] So I'm like, "Hey, before the 5 hour [26:45] window hits, if I haven't done [ __ ] with [26:47] my tokens, spend that [ __ ] [26:50] >> Spend them." And I have a whole task [26:52] list of stuff it does. So when I wake up [26:53] in the morning, I don't wake up and I [26:55] don't look at my social networks [26:56] anymore. I wake up and I look at what [26:57] OpenClaw did. I wake up and I'm like, [27:00] "What [ __ ] got beat?" Dude, I have a [27:02] whole bunch of apps. I have apps right [27:03] now. So, like, can you check my blue [27:04] sky? [27:05] >> Your whole home screen? [27:07] >> Yeah. My whole home screen is covered in [27:08] PWAs that are being made by OpenClaw [27:11] right now. I have a banjo tracker. I [27:12] have a whiskey tracker. I have a beer [27:14] tracker. I have a birthday tracker. I'm [27:16] kind of like taking data that used to [27:18] exist in all these silos and all these [27:19] walled gardens and I'm like, "That's [27:21] just mine." And I'm doing it with [27:22] OpenCloud. I'm like, "Hey, OpenCloud, [27:24] open up a browser. I've already [27:25] authenticated you into Untappd. I've [27:28] already authenticated you into guitar [27:30] tabs or ultimate guitar and I'm like go [27:32] steal every single one of my tabs and I [27:34] want you to take them put them on my [27:36] local hard drive and I want you to do it [27:37] with scripts and next time I don't even [27:39] have to use tokens. I just want to spend [27:41] script time on my computer, my desktop [27:44] computer up here which is a gaming [27:46] machine which I'm a dad and I have two [27:48] kids that are seven and nine and I don't [27:49] game anymore. So use those [ __ ] [27:51] [laughter] GPUs and make me some apps. [27:54] And dude, I wake up and I review code [27:56] that's and I wake up and I and I'm also [27:58] building a Shopify store. I wake up and [28:00] I'm like this thing farted my job. [28:04] That's this is where I'm at right now [28:05] with like this stuff is I used to spend [28:07] hours and hours sweating details about [28:09] the craft and I still love the craft and [28:11] the craft has enabled me to empower this [28:13] thing to do the work but dude in in like [28:16] five minutes this thing farts out hours [28:19] of my life into code that's well [28:22] executed. And so this is what I mean. [28:24] I'm like I used to care that my job was [28:27] uh sweating the details in terms of code [28:29] and now I sweat the details at a higher [28:31] level as an art director, as a technical [28:33] steer. [28:34] >> Um I I it's crazy. I like I'm in love. [28:39] I'm in love in a weird way because in [28:40] one way I'm in love and another way I'm [28:42] like I'm ruining the world and water is [28:44] boiling and I'm torturing poor souls in [28:48] this country that have to live next to [28:50] these machines. or maybe and then I even [28:54] that I'll read I'll read counterpoints [28:56] that are like hey that actually that [28:58] data center um recycles all of its water [29:00] so it's actually not consuming a single [29:02] drop outside of itself and I'm like oh [29:03] [ __ ] really then I'm just like reading [29:05] headlines like a dork oh god I am [29:07] reading headlines like a dork I need to [29:09] read so I'm going to stop talking now [29:10] but that's how I feel about [29:11] >> keep talking what what's your setup [29:12] [laughter] are you using a mini are you [29:14] in the cloud what models does it have [29:16] access to like what are you [29:17] >> no this particular one up here okay so [29:19] dude I'm I'm going to love telling you [29:20] about this This this is like a regular [29:22] old gaming machine. [29:23] >> But the thing is is like I installed [29:25] vanilla openclaw on it and I said, "Hey, [29:26] openclaw in the burn window. The first [29:28] few things I want you to do is add [29:30] security and I want to make sure that [29:32] anytime you scrape a website, you're [29:34] immune to prompt injection because I [29:36] know that like these are things you're [29:37] susceptible to. So when you have [29:39] downtime, downtime. Can you believe the [29:41] downtime of the brain of my computer is [29:44] spent improving itself? It does [29:46] self-healing and self-improvement when [29:48] it's just chilling. chilling with my [29:50] extra tokens. And so those are the first [29:53] few things that I had to do was build [29:54] guards, build rails. And then the third [29:56] thing I had to do was um make sure that [29:59] when I ask it to do something that did [30:01] cost tokens, it makes scripts that make [30:03] these things replayable without tokens. [30:05] And then the fourth thing I had to do [30:07] was build a dashboard about where are my [30:09] tokens being spent because I want to [30:11] know where you're doing something [30:12] stupid. And I just blogged about this [30:13] over the weekend. It's at nerdy.dev. can [30:15] go check out how my my token dashboard [30:18] informed me that everything being posted [30:20] in my Discord channel and the general [30:21] channel was being bundled up into one [30:23] Bing context. Every request and I said, [30:26] "Cut that [ __ ] out, OpenClaw. Your goal [30:29] is to take my request in the general [30:31] channel, open up a thread, and that is [30:33] the context. It's the single message." [30:35] And if you're in a thread and I text [30:37] you, then I want you to crawl only up [30:39] one message and look to see for some [30:41] context. Maybe two messages. crawl up [30:43] until you feel confident that what [30:45] you're doing is the most minimal and [30:46] most applicable thing that dude. So Pi, [30:49] I don't know if you've used Pi yet [30:50] versus Claude, but Pi is a trainable [30:53] puppy and I have a puppy up here on my [30:56] gaming machine that is now my pet and I [30:59] call it Punkass, which is short for punk [31:01] assistant. Uh anyway, I thought that was [31:03] clever. Anyway, so like this is this is [31:06] the coolest thing I've ever done with my [31:07] computer in my whole life and I'm trying [31:09] not to freak out about it. Um and be the [31:12] Not very excited about it at all. [31:14] >> Yeah, I'm not excited. I'm I am going to [31:16] stop talking this time. Even though you [31:17] aged me on, Leon, you did this. This is [31:18] your fault. [laughter] [31:22] >> Yeah. Um, so clearly there's a lot that [31:25] has changed. You can do a lot of stuff [31:28] from, you know, wherever easily, which [31:30] really kind of changes uh the dynamic [31:33] that all programmers have on how like [31:36] our jobs are, what the the landscape is [31:38] in general. So, you know, as since you [31:40] guys do a lot of helping other people [31:42] break into the tech space, get jobs, [31:45] etc. I was curious how you think this is [31:47] affecting things. Like, are you still [31:50] having a lot of demand for that sort of [31:52] thing or is everyone kind of doom and [31:54] gloom and like, oh, there's no way [31:55] anyone can break in anymore or what are [31:58] you feeling like? [31:59] >> We just talked about this literally an [32:02] hour ago [laughter] and just nerded out [32:04] deeply about it. Look, you have folks [32:06] like IBM saying they're going to hire 3x [32:08] the amount of juniors they have in the [32:10] past and uh we're just kind of starting [32:13] to see the beginning of this, right? [32:14] Like we thought that AI originally could [32:16] maybe do more work, but what it's doing [32:18] is creating more work. Uh [32:20] >> the the pace that what you're able to [32:22] develop, you just rattled off, Adam, 10 [32:24] apps you're working on right now, right? [32:26] Like like at one time, every developer [32:28] can do that now. And the speed at which [32:31] we can move has just completely changed. [32:34] you need bodies to help guide that [32:36] process. And so we're starting to see [32:38] the the maybe the resurgence of the [32:41] junior developer or the entry- level [32:42] engineer um like in a way we haven't [32:44] seen before. And literally just got out [32:46] of a meeting with a bangle company who's [32:49] literally opening back up the [32:51] recruitment practices for entry- level [32:52] talent. Like they they they see the [32:54] value of having engineers guiding this [32:58] process and they realize they might need [33:00] a lot more than they originally thought. [33:01] So that's kind of what we're seeing just [33:02] play out right now from partners and [33:04] things like that. But also just the [33:07] usefulness of it, right? Like we can see [33:09] that we're able to do more, but it still [33:11] needs our taste. It still needs our [33:14] input. It still needs our guidance. It [33:16] still needs someone to point it in the [33:17] right direction. Uh and and and I think [33:20] that's going to be the the shine of [33:21] entry- level talent going forward. [33:24] >> Yeah. So what about the interview [33:25] process though? Like so we're struggling [33:27] with this of like how do you even [33:29] interview someone because [33:31] >> what's the point of being like hey could [33:33] you sort this array my dude because like [33:35] who cares because you can just throw it [33:37] in AI. [laughter] [33:38] >> No, we actually just record the episode [33:41] is coming out. I guess sneak peek for [33:43] everybody that's here live. The [33:45] episode's coming out on Thursday, but [33:47] Hacker Rink even said that they see the [33:50] entire interview process changing and [33:51] that leak code is now going to be [33:55] defunct dead. Um, what's funny is I've [33:57] been talking about this for over a year, [34:00] >> but what they're doing is they've made a [34:03] web client that's an IDE that you're [34:07] given a task and it's judging your [34:09] prompting, the number of attempts that [34:11] it takes you to potentially get to a [34:13] result, how you're um building your [34:16] prompts, constructing it, things of that [34:18] nature, which has like kind of led to a [34:20] different angle. [34:22] And one thing that I've been very vocal [34:24] on is the biggest struggle with the [34:27] average software developer today is they [34:29] actually have the wrong mental model [34:31] when interacting with AI. And this is [34:34] where a lot of them are struggling. And [34:36] the thing that I've constantly said to [34:38] people is it's it's actually made me [34:40] like question everything. This happened [34:41] over Christmas by the way. Like I'm [34:43] still mentally strugg like Leon knows [34:45] when this like epito like epiphany [34:48] scenario happened. I questioned [34:50] everything for weeks because [34:54] I, you know, at the end of the year last [34:56] year, uh, CH chatbt and OpenAI, they [35:00] sent out like this, um, yearly review of [35:02] like what your arc type is and this and [35:05] that and all these things. And what was [35:08] interesting to me was about like [35:10] Juneish, I switched from Chat GPT being [35:12] like one of my daily drivers for like [35:14] email auditing and things like that to a [35:16] completely different tool. [35:18] and it said I was like the top three% of [35:21] users and I was like well how is that [35:24] like how is that possible [35:26] and I kind of audited um what these [35:30] archetypes meant all these things but [35:32] one thing that got very clear to me was [35:34] 97% of people using AI tooling they're [35:38] not using it correctly and [35:42] I I then took it a step further recently [35:44] and over the like the last several [35:46] months I've been doing a ton via [35:47] prompting and other things and I asked [35:49] every single AI tool that I use, the [35:50] idees that I use, all that good stuff. I [35:52] said, "What like, if you were to rank [35:55] me, where would you rank me as far as [35:57] like the average user or the type of [36:00] user that I am against like the world?" [36:02] Every single one of them said I was in [36:03] top 1%. At first, I thought there was [36:05] this like blowing smoke, but then I had [36:07] like a ton of other developers do the [36:09] same thing, and they got varied results, [36:11] right? And that led me to believe that [36:14] if you were to literally ask AI, how do [36:16] people use it? you'd be so shocked cuz [36:19] it almost rants. It's like they keep [36:21] using me like I'm just a Google search [36:23] and I'm like that's so interesting and [36:27] it it it leads me down this entire path [36:29] that the average software developer [36:31] doesn't have the right mental model and [36:33] because of that that's the beginning of [36:35] the dominoes falling in the wrong [36:37] direction. The mental model that the [36:39] average software developer has is a very [36:41] transactionary model and that's I need [36:45] this thing, give me this thing. I need [36:47] you to do this, do this, build me this, [36:49] create this. And once I started [36:52] realizing that, and the reason why I [36:54] realized this, by the way, was because [36:55] of that um end of year wrap-up. So, my [36:59] theory was [37:01] I'm going to do like a my LinkedIn [37:02] series that's kind of popular. Um I'm [37:05] going to do a LinkedIn series for AI and [37:08] we're I'm just going to record a bunch [37:09] of software developers using AI to teach [37:11] the average person how they should be [37:12] prompting, right? What I wasn't [37:15] expecting is for zero of the software [37:18] developers to use AI correctly. And then [37:20] every single recording session ended up [37:22] with me teaching them how to use AI. And [37:25] that's when I realized we have a bigger [37:26] gap than I walked in here knowing. Um it [37:29] made me question every AI take because [37:31] whenever somebody was saying AI is bad, [37:33] I'm like, I agree with you. If I got the [37:35] results that you got, I'd feel like it's [37:36] bad, too. I would I would feel like it's [37:38] a terrible scenario to be in. But that's [37:41] when I started realizing where the gap [37:43] was and I did an AI master class about [37:45] two weeks ago or maybe three weeks ago. [37:48] And the interesting thing about that was [37:53] last week when we did a meetup, I had [37:55] six different people run up to me to [37:57] show me their prompts because they're [37:59] like, "Oh my god, this changed so much." [38:02] Last night in my Discord, someone said [38:04] my monitor stopped working and I was [38:07] about to prompt it to find out what was [38:09] going wrong. And then I instantly said, [38:11] "Hold on, let me try it." The way Danny [38:13] said, he said, "The level of information [38:15] that I got, I've never gotten that [38:17] quality of before." And that just went [38:20] to show. He even said it's like it's my [38:21] mental model shifting. Like it totally [38:23] changed everything that I was getting. [38:25] This now lets me see in real time how [38:28] many software developers were in that [38:29] wrong mental model and how many are now [38:31] leveraging it in the right way. And I [38:33] think like over the next three months [38:35] what does that even look like? How much [38:37] are they going to unlock that they [38:38] didn't even know was there previously? [38:40] And I think that's like the big reason [38:42] why we even see stats right now where [38:44] companies are saying we brought all [38:46] these licenses for co-pilot which come [38:49] on but other tools too and they're like [38:52] we saw no productivity output. Like I [38:55] was talking to a senior vice president [38:56] of engineering for this Fortune 500 [38:59] company. I won't say which but he said I [39:01] I only I created I bought 700 licenses [39:04] for our engineers and Danny the only [39:06] thing I did was create a sevenf [39:08] figureure bill. AI is a joke. And I'm [39:10] like, just out of curiosity, when you [39:12] handed them the licenses, did you give [39:14] them a reason to use this besides [39:16] saying, "Hey, we bought this thing." And [39:18] he just stared at me like, "This was the [39:19] craziest question I ever asked in my [39:21] life." And then he realized like, "Yeah, [39:24] I didn't do that." But I just thought [39:26] they would be into it. It's like, why [39:27] would they be into it if you gave them [39:28] no reason to do the thing? [39:31] >> Yeah. If they didn't know what it could [39:32] do, why would you try it randomly? For [39:35] sure. But yeah, [39:36] >> I think most companies right now, [39:38] >> we've had that too. Like we had a [39:40] meeting where uh someone up the chain, I [39:42] don't remember exactly what position or [39:44] whatever, was like, "Hey, yeah, like [39:45] only 25% or less of people are using [39:48] these co-pilot licenses that we got. [39:50] Like why? Like we should all be using [39:52] this all the time." And I don't think it [39:55] really made anyone think, "Oh, I should [39:57] use that." Because he still didn't tell [39:58] them why. He was just like, you [40:00] [laughter] know, you should use it. Um, [40:03] but yeah, it's like it's definitely um [40:08] completely different than like like [40:10] co-pilot 2, you can choose all your [40:11] models and stuff now. So, it's not like [40:14] not like the original co-pilot of like [40:16] it was just not that great. Now, it's [40:18] like you can use Opus, you can use [40:20] everything. You can like um you can [40:24] actually use that um license in open [40:27] code, which is what I do. And then you [40:29] can just like run everything through [40:30] open code. You don't even have to do [40:32] like co-pilot through like VS Code or [40:33] anything. So, I've been having good [40:36] results with Copilot personally. [40:41] >> Do I want to go down this path? [40:45] >> Hold on, baby. [40:45] >> Yes, that's what the spice is for, dude. [40:47] That's okay. This show This show, by the [40:48] way, everybody who's here, this is kind [40:50] of like Hot Ones where Hot Ones makes [40:52] people answer spicy questions and then [40:54] do you a little bit spicy. [40:57] >> You're under a little bit of duress [40:58] here. You got whiskey. We've given you [40:59] the Matsui Sakura Cask single malt. [41:02] Dude, let it flow through your veins and [41:04] execute whatever is in your brain, man. [41:07] Hit it. We're ready. [41:10] So, we're at the stage where a lot of [41:12] the models are starting to average out, [41:13] and they will. There are some clear [41:15] outliers of output of what you'll get [41:17] from certain models. Like, we just [41:18] actually recorded an episode right [41:19] before this on how we felt with 4.6 and [41:23] 5.3 CEX over the last month, right? and [41:27] how our initial takes have really [41:28] changed quite a bit. Um, especially [41:30] mine. Uh, like I thought 4.6 was the [41:34] bee's knees and uh, it's definitely not. [41:38] Adam, don't smile. This is going to make [41:39] me laugh. Anyway, [41:40] >> dude, I love it. No, this is great. You [41:42] said bees knees and I'm like that's [41:43] that's my generation. [laughter] [41:48] >> You didn't call it Anyway, there's like [41:49] cooler words that you didn't [41:51] >> There are there are, but like my mind [41:52] went to everything that's like not [41:54] family friendly. But anyway, um, And [41:56] we're not fans. This is not this is not [41:57] the family. [41:58] >> Oh, is this not the time to cut? I did [42:00] not get that. The whiskey should have [42:02] been a tell. The whiskey should have [42:03] been a tell, but uh it wasn't a clear [42:05] outlier to me. A lot of the models are [42:07] averaging out and they will continue to [42:09] do so. But the way things shift is the [42:12] infrastructure around the model that [42:14] manipulates the the prompt and the [42:16] response. So that is why certain tools [42:19] have better responses. It's not directly [42:22] the model, it's the way that it's [42:24] inferring things. And so I actually do [42:26] like Open Code. I've been uh testing it [42:29] quite a bit. I have even recommended to [42:31] a couple clients when we're in the [42:33] consulting space. Like it's a cool tool. [42:35] I think I think they did a really good [42:36] job. They did a bangup job. Um there's [42:38] other tools that I I do quite a bit [42:40] with. Um right now I'm a big fan of [42:42] anti-gravity. I actually I have very [42:45] very big ideas on like where I think the [42:47] field of AI is going and especially like [42:49] and around software development. And um [42:52] I think they align with a lot of that [42:54] thought. So, I I like supporting that [42:56] tool. I think it's really great. Um, I [42:59] like Cursor 2. Uh, but I do like that NI [43:01] gravity for the ultra plan. I think it's [43:04] like$2.49 for like 3 months or something [43:06] like they're doing something special [43:07] that I got. And so, you can max it out [43:10] non-stop and not pay any overages. So, [43:12] I'm pretty big fan of that. But the way [43:14] that their infrastructure around the [43:16] prompt, the output is pretty phenomenal. [43:18] And so, because of that, I like that. [43:20] And I think we're hitting the stage now [43:22] where infrastructures of tools will [43:24] manipulate the prompts in a way that and [43:27] the responses in a way that will change [43:29] the experience for the user. And we're [43:31] getting harness. Do you know what the [43:33] test like an AI harnesses? Okay, sweet, [43:35] dude. Then you're on point. [43:36] >> Yeah, the harnesses are different [43:38] amongst all these. Yeah. [43:38] >> Oh, different. But there there's another [43:41] stage that we're starting to get to that [43:42] you've already seen like cursor kind of [43:44] display where they have like this UI [43:46] element that will edit things because [43:49] the last 3 weeks I've kind of been [43:51] learning a lot more about this new area [43:54] that I didn't really pay attention to [43:56] and that's the product manager, the [43:58] TPMS, how they're leveraging tools like [44:00] Claude Code or other tools in that realm [44:03] because they are using it but they're [44:06] using in ways that a software developer [44:08] would never use and It's kind of really [44:10] unlocked a lot of the way that I'm [44:11] seeing things. That's why like a lot of [44:13] these reasoning models have been trying [44:15] to work on um doing more with less. [44:17] Meaning people that aren't prompting [44:19] well, it's trying to figure out what it [44:20] is that they're attempting to do so that [44:22] way it can give a better response. So [44:24] that way they have a better experience. [44:26] But it to a certain degree, they also [44:28] know that the average user uh that is in [44:32] that realm probably isn't doing [44:34] something heavily technical. so they can [44:36] deal with uh less higher quality prompt [44:40] uh or higher quality response so to [44:42] speak. [44:43] >> Yeah, I think that's the new era we're [44:46] in. Like it's another thing we talked to [44:47] Zach from Warp about is like I think [44:49] that's an implementation detail. Like I [44:52] should maybe not like me specifically. [44:55] I'm doing a little bit more with AI than [44:57] your average person, but um you know, [45:01] for the masses to use it, I think it [45:03] should still just be a text box. I don't [45:06] need to really care that much about what [45:08] I type in it. And the harness should [45:10] come into play of like, oh, we've [45:12] thought about how you would do this, [45:14] what results you might get, how to make [45:15] it better, like how many agents we need [45:18] to spin up for you. I shouldn't even [45:19] need to figure out, oh, I need a [45:20] TypeScript agent and a whatever agent [45:22] and a this one to do this and this one's [45:24] going to do this. I have this thing. [45:26] I've planned it. You figure out [45:28] implementation detail of the tool how [45:30] it's done. And I think whoever does that [45:33] the best is going to win this game we're [45:35] in. [45:35] >> Can they please call it Gandalf? Like I [45:37] feel like Gandalf, you know, like he's [45:39] like, "Oh, I took hobbits and dwarves [45:41] and a bunch of dorks [45:43] and we took down the ring. We took the [45:45] ring and we put it in the thing." like [45:47] [laughter] can can that be like what [45:49] it's like using a test harness where I'm [45:51] like hey here's my whack weird like [45:53] little request and then uh Gandalf [45:55] you're just going to make that happen [45:57] right and Gandalf's like fool of a tool [45:59] kit your prompt is stupid you know he's [46:01] [laughter] like let me create a team for [46:03] you and he creates a team [46:05] >> I mean that's why I like your open claw [46:06] Adam right like it's it's a it's a new [46:08] way of interacting with this stuff it's [46:10] it's the ultimate harness if I can just [46:12] do it in Telegram or Discord or whatever [46:14] and it figures out what to do you [46:16] It's better than that, dude. It is in [46:18] Discord. It's where you are and it's [46:20] your little puppy. So, here a way I like [46:21] to think about it is Claude has a high [46:23] floor and I that's God that's one of [46:26] those like oxymorons. Anyway, it means [46:27] that like your entry level into Claude [46:29] is you give it something stupid and it [46:31] does good stuff out. However, it has a a [46:34] small ceiling, meaning that since [46:36] they've pre-baked so much [ __ ] into it, [46:39] you don't actually get to go beyond the [46:41] ceiling of Claude. Now, Pi is not the [46:43] same. Pi has a lower floor. I mean, once [46:46] you start using it, it's like, I'm a [46:48] harness that does stuff and you're like, [46:49] cool, what do you do? And it's like, I [46:51] don't know. What do you want me to do? [46:52] And to a lot of people, that's really [46:53] frustrating. But as you work with it, as [46:56] you teach it your way, as you want to [46:58] rebase, I want to use stacks. I want to [47:01] use work trees. You tell it all the [47:02] things that are meaningful. It acquires [47:04] these and all of a sudden the floor [47:06] constantly rises. And now the clutch [47:08] thing happens with pi is it doesn't have [47:10] the ceiling. The ceiling is not there. [47:13] The ceiling is you and it and where you [47:15] want it to be. And that's I think where [47:17] we're moving towards is the battery [47:18] included Claude was really dope at [47:20] first, but the puppy that you got to [47:23] train over time and acquires and months [47:26] of months of your interactions grow the [47:28] ceiling higher than a ceiling could ever [47:30] grow inside a cloud. Now that is the [47:32] primo scenario. Um Leon, I want to hear [47:35] more about the way that you're using [47:36] cloud agents. Um [47:38] just tell me more about that. Yeah. No, [47:40] I mean I think probably around January I [47:43] switched to a very different workflow [47:45] than I've been using before. And one of [47:47] the big things was for me for a long [47:49] time I've been using cursor web. Um when [47:52] we talk about harnesses I find cursor [47:54] consistently gives me the best results. [47:55] I've I I'll use codeex 5.3 even inside [47:59] of codeex and get different results than [48:01] I will inside of cursor. And I think [48:03] there's some some bit of like I just [48:06] know how to use it and I've become [48:09] comfortable with it. So it gives me [48:11] better outputs. But inside of cursor, [48:14] I've always been using cursor web. So [48:16] I've been able to look at things just on [48:20] my phone. I was at the beach. I was on [48:23] my phone doing a little agent. I'm in [48:24] line at a grocery store [laughter] [48:27] little little web agent. And so in [48:30] January, early like late January, they [48:32] reduced they put out longunning agents. [48:34] So now it's kind of like a Ralphie or a [48:36] Ralph loop but in the cloud for you that [48:41] can run seven, eight plus hours and [48:44] solve really kind of meaty things for [48:46] me. And so before I go to bed, I have my [48:50] little to-do list. I go through it. I [48:52] add it to a cursor longunning agent. And [48:55] in the morning, like you're saying, I [48:56] wake up. I have my tea. I I read [48:58] everything that's been committed. I work [48:59] through it. I bring it in if I need to [49:01] bring it in. And it's completely changed [49:03] kind of how my day-to-day goes. I have [49:06] ADHD, like pretty extreme horrible ADHD. [49:10] And I finally feel like I have this [49:11] workflow where I have like five projects [49:13] that I'm just swapping in between. While [49:15] one's building, I'm doing something [49:16] else. And um so for me, the long running [49:19] agents were a big unlock because I had [49:21] been using Ralph and Ralphie pretty [49:23] pretty hardcore. So having that running [49:25] somewhere else now you have things where [49:27] like they can show you previews and like [49:30] the agent can like work through and [49:31] click through and it gives you like a [49:33] video of what it was able to do. Um so [49:35] that's a really interesting piece there. [49:37] Um and then for me just even just [49:39] day-to-day inside of cursor um really [49:43] heavily using plan mode that the way [49:46] that kind of before I was using Ryan [49:47] Carson's kind of uh their kind of plan [49:50] mode before it became part of Claude and [49:51] became cursor. But I really find that [49:53] cursors plan mode is really really how I [49:56] think and I'll go back and forth with it [49:58] for a while typically of 46 and then [50:00] I'll switch over to 53 to do all the [50:02] implementation details and it rips. It's [50:05] able to just build what I want it to [50:06] build. I have a bunch of other secondary [50:09] things. Um, one of the tools that Danny [50:11] showed me sync for making sure security [50:13] pieces are there. Bugbot is an I I I [50:17] could never not use Bugbot going [50:19] forward. Bugbot for folks that haven't [50:21] used it before is kind of like a code [50:22] review tool like a grap tile or [50:24] something similar to that. So once my [50:26] stuff is done, it gets pushed to it gets [50:29] pushed to um GitHub. I have just a [50:31] command to push it to GitHub. It does [50:33] really good kind of everything for me, [50:35] does everything for me. But then Bugbot [50:37] runs cleans up all my mess. Verselobbot [50:39] runs cleans up all my mess. Sync runs [50:41] cleans up all my mess. And I'm really [50:43] trying to get to this point to where [50:46] I trust it enough to not have to read [50:48] it. I still read everything. I still [50:50] like digging through. I catch stuff [50:52] every once in a while. But I want these [50:55] layers going, especially if I can have a [50:59] longunning agent that can fully test [51:02] stuff. Before I was using Playright, but [51:03] that just murdered my token count. Um, [51:06] but if I can have something that can go [51:07] endto-end test, I can have it do browser [51:10] testing, I can have it do all my [51:12] security audits, I can have it do all of [51:14] my code review, and I can get to this [51:16] point to where I can trust the output. [51:18] That's my ultimate goal. And so, I don't [51:20] think I'm there yet, but for like 80% of [51:23] my stuff, I am. And I feel confident [51:25] that I can start kind of Rick Rubin in [51:27] it and just lean into my taste, lean [51:29] into like what I like and don't like, [51:31] and then just keep it going, you know? [51:32] So, that that's where I want to get. [51:34] Yeah. So then if you do that, my my [51:37] follow-up question is like, you know, [51:39] that that's a utopia we're all like [51:40] trying to get to, I guess, of like I can [51:43] just trust it all the time, right? At [51:45] what point [51:46] >> do you audit it though? Like all these [51:48] tools are going to start like off the [51:50] going off the rails at some point and [51:52] hallucinating or getting worse, I would [51:55] think. Like um I just don't think it can [51:58] go forever of being good. I don't know [52:00] if you [52:00] >> Yeah. I mean I mean that and it's and to [52:02] be totally clear it is not there yet [52:04] like [laughter] like don't listen to [52:06] this and like go change over your entire [52:08] workflow like it's it's not there. It [52:09] runs in weird things. You're also [52:10] talking to somebody that like tests in [52:12] production loves feature flags. Like I I [52:15] want to I want to be in this world where [52:17] I can just ship stuff. If it breaks I [52:19] roll back. And I want the I want the AI [52:21] to know when it rolls back, right? And [52:24] sometimes you you you leak some security [52:26] keys. That's okay. Like we just keep it [52:28] going. And so it might not work for a [52:30] production [laughter] level works, but [52:32] for what I'm building, it's good enough. [52:34] And maybe one day we'll get there, you [52:35] know? [52:37] >> Yeah. [52:39] >> And again, it's it's it's not just to be [52:43] clear, it's it's not [52:45] perfect. It's not how I currently always [52:48] ship code. It's what I'm actively trying [52:50] to work towards, though. I want to get [52:52] to a world where I can do that and then [52:56] focus on product and focus on like [52:58] getting to what customers actually need [53:00] and and then also maybe spending time on [53:03] like the more nitty-gritty engineering [53:05] decisions that that one would have to [53:07] make and if I can have four cloud agents [53:10] running while I'm doing that doing the [53:11] UI work that I don't want to do like [53:13] great. [53:15] >> Yeah. The amount of stuff it lets you do [53:17] that you would have never touched before [53:19] because you didn't want to touch it is [53:20] so awesome. Like I was just like, "Hey, [53:24] uh, Shepherd is written in spelt." And I [53:26] tried to upgrade it to Spelt 5 and it [53:28] got way bigger for some reason. And I [53:30] went, "That's not good." So I was like, [53:32] "Hey, what if it was in solid?" And it [53:34] was like, "Boom, it's in solid." And I [53:35] was like, "All right, what if it was in [53:36] no framework?" And it was like, "Boom, [53:38] it's just TypeScript." And I was like, [53:39] "Sick, ship it." The amount like that [53:42] would have taken me months and months [53:43] and months of work to figure out how to [53:45] do that. And you can just ask it and it [53:48] just provides. So, it's like it's a [53:51] crazy world. [53:53] >> And asking is that like prompting [53:55] technique that people kind of mix up. [53:57] They're so used to telling it to do [53:59] something, but asking it how to do [54:01] something can give you phenomenally [54:03] better results, especially since it [54:04] knows more a lot of these verticals that [54:06] you don't know, you know. [54:08] >> Yeah. And like you said, start with plan [54:10] mode because if you just go into build, [54:12] it could go down the wrong track. But if [54:14] you do plan and you're like, "Your plan [54:15] sounds great," then you know it's going [54:17] to be doing the thing you wanted. [54:19] I want to capture y'all's uh hot tips [54:22] for using AI right now. So like and I'll [54:25] start um because I'm this is all new for [54:29] everybody, right? Models are now good. [54:31] This to recap where we're at everybody. [54:33] Models are now good. We're now realizing [54:36] that um the way that we interface with [54:37] them, we have to change the way that we [54:39] ask for them. They're not search engines [54:41] as Danny was talking about. So how do [54:44] you get the most out of it? Leon is [54:46] pitching plans. Plans are awesome. Um, I [54:49] want to pitch that you um have it [54:53] established the baseline like what is [54:54] today like what is what is the code now? [54:57] Where do I want it to be? I expect it to [55:00] be this state afterwards. And so one of [55:02] the things that I've been doing [55:05] is um having it track where it's at now. [55:09] Give me a test run a test that says this [55:10] is the performance of something right [55:12] now for example. Now I want you to go [55:13] make this change and I want you to [55:15] measure performance afterwards. If the [55:17] performance is not better, then this PR [55:19] is [ __ ] Okay, just throw it away. And [55:22] that's how we work. It's like, I want [55:23] you to track now and then. If then is [55:27] not better, then we're at a then we then [55:30] we'll move on. Um, so that's one of my [55:33] tips right now is try to orient this [55:35] thing about outcomes that you're [55:37] expecting. And if the outcome is not [55:38] better than it is now, it's trash work [55:41] and you can move on. So it should allow [55:43] you to move faster as you sort of [55:46] iterate through ideas that you want or [55:48] that you thought you want but then you [55:49] realize the outcome is actually not. [55:50] It's outcome focused outcome of what you [55:53] want outcome of what is today. So I'll [55:56] throw that to you all for sharing your [55:58] uh how to prompt AI to get the things [56:00] that you want tips. [56:05] >> Should I follow that? I don't know if my [56:07] response is going to be positive. [56:10] >> Do it. Let somebody else spice. Let the [56:12] spice out, dude. This the dragon. Let [56:14] the dragon out. You know, [56:16] >> you know, I appreciate that. Like you've [56:18] had good experiences with that. [56:20] >> I would almost challenge you the next [56:23] time that you use AI to try maybe [56:25] something slightly different. And the [56:28] way that I challenge people is um [56:31] especially in regards to code [56:34] to allow it to index itself first. This [56:37] is the problem that a lot of people [56:38] don't um it you're what you're [56:41] describing is that transactionary mental [56:43] model. I need this thing and if you're [56:46] not performing higher, this is trash [56:50] where the way that I think about the [56:52] prompt is it's almost impossible for it [56:55] to get for it to be wrong if it follows [56:58] the way that I've listed things out. One [57:00] of the ways that I list it out is by [57:03] number one allowing it to index [57:04] everything that it needs to index first. [57:06] So instead of giving it a task, what I [57:08] say is this is our potential theoretical [57:11] goal. Here is a suggested way that I [57:14] think we could do this. Tell me why this [57:16] is wrong. If it is wrong, challenge [57:18] this. I would also like you to find out [57:21] where this could potentially break or [57:23] what am I missing. That one question [57:26] alone allows us to do far more research [57:29] around secondary or third level [57:31] components to ensure that we're not prop [57:33] drilling, we're not breaking this down, [57:34] we don't have edge cases that we're not [57:36] aware of. And I have it do a full audit, [57:38] right? Prepare a markdown document that [57:41] I need need to know. It's not that I [57:43] really need to see it. Maybe sometimes [57:44] I'll look over it, but it's now that it [57:46] has a full picture of everything that [57:48] it's comprising because AI is amnesic by [57:51] nature. It's it needs to be amnesic. [57:53] needs to forget because it doesn't [57:55] necessarily know what problem you're [57:57] going to present it the first time [57:58] around or the next time around. And so [58:00] because that we need to establish said [58:02] baseline in the parameters that we need. [58:05] Then once we start giving it tasks, it's [58:07] able to take it to such further levels [58:10] than it ever has prior. So because of [58:12] that, I don't personally like leading [58:15] with uh a a transactionary mindset or [58:19] conversation. What I like to do is [58:21] almost look at it as like a senior [58:22] developer on my team that's challenging [58:24] me like that a that a-hole that just [58:26] always wants to argue or he always wants [58:28] to fight or something along those lines. [58:30] I want that dude on my side [58:32] >> wants to merge [58:33] >> the devil. [58:34] >> He wants to merge. So he wants to get [58:36] mer wants to get merge when we're all [58:39] rebasing. Okay, we've already [58:41] established re this is a rebasing [58:43] universe over here and he wants to [58:45] merge. But because of that, he now has [58:48] enough information to where if I don't [58:51] necessarily know of a strategy, oh, this [58:53] is information I wasn't even aware of. [58:55] The problem that a lot of software [58:56] developers have is they automatically do [58:59] the mental work without having the full [59:02] picture. They have their determination [59:04] of a full picture, but they don't have [59:05] the actual full picture. And because of [59:08] that, there's edge cases that aren't [59:10] really edge cases that they miss. This [59:12] allows all that to be solved. And then [59:14] what happens with a lot of the output is [59:16] you find an edge case that wasn't really [59:19] an edge case but in your mind it was [59:22] that is now a bigger case. And because [59:24] this is not performant at the level that [59:26] you wanted it to be, you now question [59:31] everything the model did beforehand and [59:32] it throws it away. Whereas if we set it [59:34] up right the first time, it indexes all [59:37] of this like sphere of information to [59:39] where now it's almost impossible that [59:41] it's not more performant. [59:48] need more. [59:49] >> How are How are you? Uh how are you? [59:51] >> A couple on that [laughter] one. [59:52] >> That's good. I loved it. [59:55] >> Yeah. No notes from me. [laughter] [60:02] >> How about your processes? Like what are [60:04] what are some other things, Robbie, have [60:06] you been playing with that you've [60:08] noticed have been helpful? [60:10] Um well the yeah the first thing like [60:14] just picking tools and seeing how they [60:17] like are different like the harness we [60:20] talked about like [60:21] >> um I think open code does the best like [60:23] compared to um like copilot is not not [60:26] the best um warp is pretty good too but [60:30] open code tends to be a little bit [60:32] better. Um, yeah. I don't know. Like [60:36] I've been mostly experimenting with that [60:38] stuff, like trying to figure out, you [60:40] know, like Open Code's big thing was [60:42] they had um I always forget the the [60:45] freaking acronym. What is it? Uh, [60:49] someone [60:50] >> LSP. LSP. Yeah, I think is the word or [60:53] acronym, right? Like [60:54] >> language server protocol. [60:56] >> Yeah, I think like there that was their [60:57] thing is like we know what TypeScript [60:59] is, so it's better than like one that [61:01] didn't. And maybe that doesn't matter as [61:02] much on the latest models or whatever, [61:04] but like um picking tools you like. [61:06] There's so many different tools now. [61:08] There's the uh the one Adam I haven't [61:10] tried it yet, but it looked cool. The uh [61:11] like I think it was like CMX or whatever [61:13] or it was like [61:15] >> Oh, that was like um Arc and T-M had a [61:18] baby. Yes. [61:19] >> Yeah. [61:20] >> And then there's also Intent by Augment [61:22] which is uh Anyway, yes. There's so many [61:24] tools right now. Proliferation of tools. [61:27] >> Yeah. So, like that's that's when the [61:29] majority of what I'm doing is like it's [61:30] a new day, here's a new tool, let's try [61:32] something new, see how it works. Um, [61:35] because like I don't want to be too [61:37] granular. Like I said earlier, like I'm [61:39] I want them to figure it out for me. So, [61:41] I don't need to be like I want seven [61:43] agents that are have these specific [61:45] personalities. I'm not doing a a Ken [61:47] Wheeler openclaw where it's a whole team [61:49] of uh like super smart animals that like [61:53] they all have their own roles. Like, I [61:55] don't care that much. Um, but yeah, I [61:59] think like um, Open Code does a good job [62:01] too of like, oh, I'm going to spin up [62:03] sub agents for like, you know, these [62:05] three things for you. [62:07] >> Um, so you don't have to care. Um, [62:11] but yeah, also the uh the built-in thing [62:13] on GitHub is so compelling. Like you got [62:15] a little agents tab, it's like boop, and [62:18] you just go in and tell it to do stuff [62:19] and it's like [62:20] >> I've been rocking that. Yeah, I love it [62:22] being right there because like, you [62:24] know, I was telling when we first saw [62:26] it, it just kind of showed up. I was [62:27] telling Adam like usually when GitHub [62:29] has something show up, I'm like, like I [62:31] could I could do without whatever you [62:33] put in here. But I was like, this is [62:34] sick. I would like this is going to [62:36] change the game. People just won't leave [62:37] GitHub. They'll be like, here's this [62:39] issue. Implement it for me. Done. And [62:42] then like if it's not right, you just at [62:44] Copilot and [62:45] >> it's got the canban board built in. Have [62:46] y'all tried vibe canban yet? like any of [62:49] the flavors of it where like you [62:51] basically just create a task list of [62:53] things that it's like cards on a board [62:54] and all of a sudden AI moves them [62:56] forward. [62:57] >> That's cool. [62:58] >> I'm like that shit's amazing. [63:00] >> Yeah, you were just talking about this [63:02] where it's like the the tooling. [63:05] >> I mean, you have you've always had [63:06] developer experience stuff, right? [63:08] You've always had things that like made [63:09] our lives easier, but we're at the point [63:11] now where I'm making very serious [63:13] decisions based on how easy it is for my [63:16] agents to use these tools. [63:18] whole new databases I've never used [63:20] before, whole new orgs I never would [63:22] have touched with a 10-ft pole because [63:24] it makes it 3 seconds faster with my [63:26] agents, right? And so I think we're just [63:28] in this new world where even these [63:30] companies have been around for a while. [63:31] If you don't have good ways for my [63:33] agents to talk to your stuff and make it [63:35] easy for me to use, it's it's just over. [63:37] I'm not using your tool anymore. [63:39] >> Yeah. MCPs skills get it all. You got to [63:42] have it. Can't not have it anymore. [63:45] >> It's weird because [63:48] I'm with Leon in certain areas, right? [63:50] Like if my tool can't necessarily [63:51] communicate well um [63:54] with your tools in an agent flow like [63:57] that's a disconnect. [63:59] Prime example Leon mentioned Sneak ear [64:01] earlier. I'm a big fan of them. They [64:03] need to cut me a check by the way. I've [64:04] converted so many people to using this [64:06] tool. Like y'all need to cut me a check. [64:08] Can y'all sponsor this podcast? SNK. [64:11] Hold on. Hold on. [64:12] >> Yeah. SNK. [64:13] >> Okay. Yeah. Let me let me share it here. [64:15] >> Yeah. I I I like them and they they need [64:17] to sponsor both of our podcasts. How [64:19] about that? [64:19] >> Are they the one with the dog logo? Am I [64:22] Is that the right thing in the right [64:23] thing? [64:24] >> Is that the right thing? I think so. [64:26] Hold on. Yeah, it's the dog. Yeah. Um I [64:29] only had to type in the letter S and it [64:31] was the first thing that popped on my [64:32] browser. Uh [64:35] >> I got y'all. [64:36] >> Yeah. So, I do like uh tools that [64:38] communicate well, but unlike Leon, I'm [64:41] still like especially with like my side [64:44] projects. Like one project I've been [64:45] working on right now is called event [64:46] convo, right? Which is um my solution to [64:49] like meetup.com, Eventbrite, etc. I feel [64:52] like it's probably the right solution [64:54] for meetups, especially from the [64:56] organizer perspective, conference [64:59] speakers, um or just Meetup speakers [65:01] just in general. Um conferences and [65:03] actual attendees. It's solving problems [65:06] that people just have never thought [65:08] about and I don't necessarily know why. [65:10] Um, and like I think right now it's like [65:12] at 140,000 lines of code. So when I use [65:15] it for like these AI demos, it's no [65:17] longer going from zero to one. I feel [65:19] like everyone does zero to one. Nobody's [65:22] really interested in that. Like most [65:23] developers are working on something [65:24] that's pretty large. And so seeing AI [65:27] work within a large codebase is a very [65:29] very very um different scenario. And to [65:33] by the way, having a live chat is super [65:35] distracting for me. Uh like like [65:37] [laughter] I keeping [65:40] I'm so Twitch brain. This is my element. [65:42] I love it. [65:43] >> Oh, [laughter] see like I'm so like [65:45] hyperfocused. Like someone's like, "What [65:46] does Sneak do?" I'm like, "Hold on, let [65:48] me deviate right here." Uh, Sneak is a [65:51] tool. It's one of my favorite tools [65:53] actually. I use it a lot. And you [65:55] connect it to your GitHub repos or [65:57] whatever it may be, and it will do [65:59] audits and will find security [66:00] vulnerabilities if there are any, and it [66:02] will report it to you. uh once in a [66:04] while it has something that's not [66:05] necessarily like a legitimate [66:07] vulnerability, but it could be something [66:08] that you should pay attention to. So, [66:10] you're making the willful choice to [66:13] ignore said thing, but at least it's [66:14] bringing it to your attention. And so, I [66:17] think that especially as people are like [66:19] vibe coding or building tools without [66:22] like being hands-on, I think this is [66:24] like a really good safe way to at least [66:27] protect yourself a little bit, right? [66:29] Um, I'm more on the hands-on type, but I [66:33] like having an extra thing there in [66:36] order to confirm that I'm accurate. [66:40] Because one thing that I've talked about [66:41] on the podcast in the past, [66:44] the one thing that I'm really like [66:45] geeking out about in particular is SLMs [66:48] versus LLM, small language models versus [66:51] large language models, and orchestration [66:53] layers that are trained on certain [66:55] knowledge bases. And Matt, [66:58] >> we literally just recorded this by the [67:00] way. But [67:01] >> go go go go go go go dude do it. [67:03] >> One thing over the weekend I was [67:05] building a tool part of my role. I work [67:07] at Atlassian so transparent there but [67:09] I'm creating uh a lot of coding content [67:12] tutorials things like that. And one [67:14] thing that I was recording was I had the [67:16] idea that what if we had like a Discord [67:19] custom Discord bot that anytime that [67:21] someone got a C certification and [67:23] something Atlassian related, we can use [67:25] a Discord command and it would create a [67:27] Jira ticket because when they get [67:29] certified, a lot of us don't know. And [67:32] so it's not until they tag us in a post [67:33] thanking us for like helping them get to [67:35] XYZ goal do we finally realize, oh, you [67:38] got certified. But if we had this [67:40] automation there, it would create the [67:41] jury ticket and then we would know to go [67:43] say, "Hey, kudos to you for getting [67:45] certified in XYZ, right?" Cool tutorial. [67:48] I feel like there's a lot of use cases [67:50] for a lot of other communities as well [67:51] with this regardless. Fun little [67:54] automation. When I use CLA, I use Claude [67:57] and I use um 5.3 Codex to audit my [68:00] codebase to say, "Hey, did I miss [68:02] anything?" [68:03] Both of them said really solid, great [68:06] best practices here because I'm using [68:08] Atlassian Forge, which is JavaScript [68:11] using React components to create SAS [68:13] products. Um, and honestly, one of the [68:16] cool things about it is it alleviates [68:18] the reason why 99.9% of SAS products [68:21] die. Uh, which I don't know why more [68:23] people don't know about Forge. Uh, I've [68:25] been geeking out about it and it's [68:26] actually one of the reasons why I wanted [68:27] to join the team. But there are some [68:29] people making like some serious money on [68:31] it, by the way. like serious money, but [68:34] you could literally use JavaScript to [68:36] make a solution for like Jared [68:38] Confluence, etc. It goes into a um uh [68:42] marketplace that is served to 250,000 [68:44] enterprise level organizations and they [68:46] serve it there. You get their [68:48] infrastructure, you get their hosting, [68:50] you get Elassian security, which is the [68:52] reason why most SAS products aren't [68:53] allowed to get enterprise level [68:55] customers, you get their security [68:56] clearances and their security providings [68:59] just by using JavaScript. You don't need [69:01] to know everything else, just serve it. [69:03] Right? Regardless, I made this with [69:05] Forge and I was super proud of it. And [69:08] that's when I realized Atlassian has [69:10] roodev, which is an orchestration layer [69:12] that is trained only on Atlassian's best [69:14] practices. And it found three vulner [69:17] like issues. Codex missed it. Claude [69:21] missed it. But because this small [69:24] orchestration layer that's super [69:26] specific on Atlassian's knowledge base, [69:29] they were able to pick it up. Because of [69:31] that, I found three vulnerabilities that [69:33] honestly two of them I wouldn't want in [69:35] a tutorial. One of them I probably [69:36] wouldn't want in a production app, but [69:38] for a tutorial probably doesn't [69:39] necessarily matter, but it picked up all [69:41] three. [69:43] Claude missed it. OpenAI missed it, but [69:47] RoboDev picked it up. And that's because [69:49] SLMs, [69:51] they're going to be the way that as [69:53] these models evolve, that's going to be [69:55] the next wave. I've been talking about [69:56] this for over a year now. So, if you [69:58] listen to the podcast, you probably [69:59] already heard this one. But um I think [70:03] SLMs are going to be the way of the [70:04] future because more nuance into your um [70:09] prompts is going to be where the value [70:10] is. MCP has its value for sure, but SLM [70:14] is a closer level of where you would [70:17] want outside nuance to language specific [70:20] knowledge. [70:22] I want to back you up. So, Shopify um [70:26] just recently has said that if you're PR [70:29] merges and there wasn't an LLM that [70:32] reviewed it, you're busted. [70:35] >> So, Danny Danny is here trying to tell [70:37] you right now, [70:37] >> what if there's a real person reviewing [70:39] it? [laughter] Hey, real people hit the [70:42] button, [70:42] >> you know. [70:44] >> Open claw hired the person to then [70:46] review the code otherwise [70:47] >> human. [laughter] [70:49] >> Yeah. Right. So, [70:51] >> I mean that's that is a threat of what's [70:53] happening. All right. But like anyway, [70:54] it's like what the whole point here is [70:56] it's like you need a quorum. Have you [70:58] have you ever used an LLM scenario where [71:00] there's four LLMs that you pit against [71:02] each other? Have you ever pitted Gemini [71:04] against Chat GBT against Claude against [71:08] anybody? You get a quorum. So this [71:11] quorum then starts to debate about the [71:12] scenario. You can have them debate about [71:14] your PR. And all Shopify is trying to [71:17] say right now is if you didn't at least [71:19] bring one [ __ ] person into the [71:21] quorum, then you did it wrong. [71:23] >> Yeah. [71:23] >> You you need to have an intense [71:25] reviewer. Someone, you know, Danny's [71:27] also pitched the devil's advocate. You [71:29] can put that on your team in Claude, by [71:30] the way. You can create a new team in [71:32] Claude and you can have anybody that you [71:34] want. Put a devil's advocate in there. [71:36] It's always good to have someone in [71:37] there that's like a naysayer. This is [71:39] like, "Oh yeah, Murphy's law, [ __ ] Did [71:42] you did you know that this could show up [71:44] and really thwart your TypeScript [71:46] system?" And you're like, "Oh [ __ ] I [71:47] didn't think about that." So anyway, um [71:49] Shopify is all in on this concept of you [71:52] need to have LLMs, not just writing [71:55] code, but reviewing code. And they don't [71:58] have to be big language models. You can [72:00] use an SLM like Danny keeps pitching [72:02] pitching. And this is this is a healthy [72:03] way to move forward is to have as many [72:06] different checks and balances as you [72:07] can. So you've got your CI/CD, your [72:10] tests, your unit tests, your llinters, [72:12] your style, all these things. Plus you [72:14] have LLM popping in there. Um this saves [72:17] you from security issues. This there's a [72:20] number of things that this is saving you [72:21] from. And we're automating this. So, [72:23] this isn't like you're in a a gauntlet [72:25] trying to deliver something to a Boeing [72:27] 747 airplane where like you wanted to [72:30] change some little factor about the [72:32] chairs and you had to go through all of [72:34] this hullaloo. This is a meaningful set [72:37] of fast running evaluations that look at [72:40] what you're trying to do and stop you [72:42] before you ever ship it. And that's just [72:44] cool. I also like Leon strategy though [72:46] where he's like, "Hey, I ship it behind [72:47] a flag and yo, if it doesn't work out, [72:49] then that's all good." Oh [laughter] [72:51] >> yeah, you should do that with [72:52] everything. [72:53] >> Yeah, the other thing too is like I was [72:55] using c uh nineball in chat said that uh [72:57] good way to burn tokens and yeah, I was [72:59] using custom sub agents to do this for a [73:01] very long time, but now Bugbot Greile [73:04] are 40 bucks unlimited per month. Like [73:07] no reason why there shouldn't be some [73:10] level of code review for every single [73:12] commit, you know, [73:14] >> or PR at least. Yeah, [73:16] >> if you're open source, Code Rabbit is [73:17] free, by the way. Yeah, Code Rabbit has [73:20] been really good to us so far. Um, just [73:22] the quality of like the reviews are [73:23] really nice, actually. [73:25] >> But Gretile has a cooler name. I'm not [73:27] going to lie. Greile is a sick ass name. [73:29] That is really good. [laughter] [73:31] >> Yeah. [73:32] >> And Code Rabbit. [73:34] >> Yeah. [73:35] >> Yeah. Sorry. Go ahead. [73:35] >> Things too is like they they've caught [73:37] so much stuff that I never would have [73:38] caught. [laughter] [73:40] Like there's so many things that like [73:43] >> Yeah. This that never would I have saw [73:44] that in the first pass. And so it's just [73:47] it's also like a really cool learning [73:48] tool, right? Like I'm seeing all these [73:50] reviews coming in, thinking about things [73:51] I never thought about. So it's [73:52] accelerating my pace going forward. [73:54] That's the other big piece about this [73:56] that I really like, you know. [73:59] Yeah, Code Rabbit found stuff in that [74:01] like Rust stuff where I didn't know any [74:02] Rust. It was like, hey, um, you're [74:04] converting coordinates from a like cuz [74:07] it was like a 2x screen to like a not 2x [74:10] screen and vice versa and all like all [74:12] these different methods doing that back [74:13] and forth. And it was like, "This one's [74:15] off by like a few pixels. Like, you [74:17] probably missed that." And I'm like, [74:19] "How did you like you went through all [74:21] of this and found that?" Like, [74:23] >> so yeah, it's they're really good at [74:24] like it's I think another harness thing [74:26] of like they just tell it really good [74:28] stuff about being good at reviewing and [74:30] it checks that stuff that it like if you [74:32] just generically threw like, "Hey, [74:34] Claude, check [74:34] >> specialized models, [74:35] >> it would be not as good." Yeah. [74:38] >> Love it. [74:41] >> Dang. A little too tipsy there. Yeah. [74:43] Well, we're overtime anyway. So, [74:47] [laughter] [74:48] if he's not coming back, I don't know [74:50] what he's doing. Um, we could end it [74:53] here. I guess [74:54] >> I was calling for a round two anyway. [74:56] This was way too much fun. [74:57] >> A lot of fun. Yeah. Yeah. Anytime. [75:00] They did say one minute. So, they put [75:02] the finger up before they It wasn't like [75:04] a drop. [75:05] >> Sounds like a potty break, you know. [75:07] [laughter] [75:08] >> Yeah. He went to get another Coke to mix [75:10] with the uh [75:10] >> So, wait, Leon, I sent you my uh Pokemon [75:13] Go code or with your friends. Do you [75:15] still play Are you playing Monster [75:16] Hunter? [75:17] >> I've played I've played older versions [75:19] of Monster Hunter. I played I haven't [75:20] played any of the newer ones, [75:21] >> but the mobile Have you played the [75:22] mobile app? [75:22] >> No, I've not played the mobile game. [75:24] >> Oh my goodness, bro. It's really good. [75:25] It's like the only game I play on my [75:27] mobile phone where I'm like, "This feels [75:28] like it's a nextG mobile game." [75:31] >> Um, and it uses all the Pokémon Go like [75:33] backing. You got to be like physically [75:35] walking around. [75:36] >> Dope. I'll try it. Yeah. No. Anyway, [75:38] >> I still play Pokémon Go here and there. [75:40] Um, [75:40] >> same. [75:41] >> I'm a big TCG player, so I play PTCG [75:43] live. [75:44] >> You playing in the house, too? Yeah. [75:45] >> Yeah. Um, yeah. I'm I'm I'm at like [75:48] level 97 right now on that. So, [75:50] >> sick. [75:51] >> Yeah, that's my main one. [75:54] >> What's your favorite Pokemon? I know [75:56] that's a hard question, but do you know [75:57] it? [75:58] >> Like like favorite like cuz I grew up [76:00] playing this stuff is Bulbasaur. [76:02] >> You can tell me your childhood favorite [76:03] or your like [76:04] >> Bulbasaur. [76:05] >> Bulbasaur. [76:05] >> I put out a tweet the other day. I [76:07] showed I showed the wares. So, in 1999, [76:10] I won the first Pokemon tournament. [76:12] >> Like, I've been in this stuff since the [76:14] beginning. Yeah. I got the badges to [76:16] show it. Like, yeah, [76:17] >> there was one and you were the winner. [76:20] >> Uh, so there was like there's three [76:21] different cups that you could win in [76:23] like if before like you could like level [76:25] to 100, you could level to like 50, you [76:27] could like, you know, like and there's [76:28] different cups. So, I won my cup, but [76:30] like it wasn't like a Swiss tournament [76:32] that you have now. was like, "You had [76:33] you only had so many matches that you [76:35] won and if you won you could go play the [76:37] gym leader." So, I won my cup. I won all [76:40] my matches and I beat all the gym [76:41] leaders. So, I got all these like cool [76:42] badges in a hat. [76:44] >> Yeah. [76:44] >> Nice. Nice. [76:46] >> That's sick. [76:47] >> And I was the first person to play [76:49] Pokémon Go outside, which was another [76:52] interesting thing. Like live streamed [76:54] outside. Yeah. [76:55] >> Oh, I see. I see. Okay. [76:56] >> Mhm. [76:57] >> Yeah. [76:57] >> Probably not the first person to play it [76:59] >> outside. No, live [laughter] streaming [77:01] playing outside. Yeah. Yeah. [77:04] >> Yeah. I still I dipped out about like a [77:07] year or two ago. I started to slow down, [77:08] but I was Pokemon Go for a long time, [77:12] like consistently. [77:14] >> It sucks that it's not as popular [77:15] anymore cuz like it was so fun and then [77:18] everybody's just so fickle and like new [77:20] thing or oh, I'm I'm not on TikTok [77:22] enough. Let me stop this. Like [77:25] [laughter] I don't know what everyone's [77:26] doing. [77:26] >> The cool thing in LA is it's really [77:28] popular still in LA. Like you can go to [77:31] any of the hot spots. If you go to [77:32] Prominade in Santa Monica, you'll see [77:34] everyone just like glued to their phones [77:36] playing and so there's still like a [77:37] scene here which is really nice. [77:39] >> I love it when you go somewhere new and [77:41] there's just like lures on all the stops [77:43] and you're just like, "Oh man, people [77:45] are playing. I love it." [77:46] >> Um I go to Japan, I'm like, "People are [77:48] playing Monster Hunter, right?" And I'm [77:50] like, "Oh, I guess a little bit, but not [77:52] as much as Pokémon Go." Pokemon Go still [77:54] has lures everywhere. And I'm like, [77:55] "Okay." [77:56] >> So So same question to you, Adam. [77:58] favorite Pokemon and then favorite game. [78:01] >> Oh man. [78:03] >> Favorite Pokemon. Um it's the um [78:08] uh Oh [ __ ] dude. I'm going to sound [78:09] like a noob right now cuz like I have a [78:10] terrible thing with names. Um I don't [78:13] know what it is, but it's the it's the [78:15] toxicity. [78:16] >> Oh, there you go. [78:17] >> I just So like as you can see, [78:19] >> guitars and stuff and so well toxicity [78:22] is like this purple punk rock Pokemon. [78:24] It's basically got a mohawk that goes [78:26] all the way past his butt onto his tail. [78:28] >> That doesn't sound like an original 150 [78:31] to me. So, [78:32] >> it's not an original. No, this is [78:33] definitely a [laughter] modern one. [78:35] >> Um, for games that I really like, dude, [78:37] I I'm the Pokemon Go mobile app has been [78:40] by far the one I've put the most hours [78:42] in. I've played a lot of the ones on [78:43] Switch. They're really good. I feel like [78:45] a little bit grinding in those. Oh, [78:48] they're grinding on Switch. What am I [78:50] doing on mobile? Oh [ __ ] I'm grinding. [78:52] Um, but the mobile one's like a little [78:53] more casual as opposed to like I got to [78:55] sit down and play Switch ones. Um, but [78:57] Violet, I don't know. I really liked [78:59] Violet. Like once they modernized some [79:01] of the like interactions, I'm into that. [79:03] Um, I like it when technology kind of [79:05] takes the [79:06] >> model. Yeah. Did you? I I I really like [79:09] Scarlet and Violet. I um [79:12] >> I put I put 15 hours before the first [79:14] badge [laughter] [79:16] >> into the game. [79:17] >> Awesome. [79:18] >> And um early on there was like this [79:20] weird river I tried to cross. I went in [79:22] like followed it, found like a hole in [79:23] like the side of a mountain, went up [79:25] into this field and there's a huge [79:26] arcanine running around the middle. I [79:28] like lost my mind. I was like I [79:29] [laughter] was like jumping up and down [79:30] on my couch like Yeah. So, I think those [79:33] moments are really special and yeah, I [79:35] really enjoyed Scarlet Violet for sure. [79:37] >> Yeah, I did all the things where like [79:38] the mobile integrations with Scarlet and [79:39] Violet like with the Pokémon Go app. I [79:41] sat down and I me and my kids were like [79:43] sitting around. I'm like, "All right, [79:44] you need to hit that button right now [79:45] because I'm ready to trade and you're [79:47] kind of wasting my time. So, would you [79:48] hurt?" You [laughter] know, like [79:50] >> I like uh the gimme ghouls drove me [79:53] nuts, though. And like the whole the [79:54] gold lure thing. Did you actually get [79:56] the gimme ghoul to a thousand coins and [79:58] evolve him? [79:59] >> Hell no. [80:00] >> Okay, [laughter] dude. I knew people [80:01] that did. They're around me. They're [80:02] like in my They're like my neighbors. [80:04] And I'm like, "Y'all are crazy." I'm [80:05] like, I I admire you in a weird way. You [80:07] have no life and kids, but that's cool [80:09] that you did the give me goal thing. [80:11] >> Anyway, [80:13] >> uh Hacksurus was one of my other [80:14] favorite ones because I'm a coder and I [80:16] just love this name, Hacksorus. It's [80:18] just cool. Love it. [80:21] All right. Well, we probably should end [80:23] at this point. We're 20 minutes over [80:25] time. Um, we should definitely do [80:27] another one, though. This was fun. So, [80:29] yeah. Thanks for everybody hanging out. [80:31] Thanks for you guys coming on. Uh, what [80:33] do you guys want to plug or mention [80:34] before we end? [80:36] >> Anything. [80:39] >> Forge. [80:41] [laughter] [80:49] >> [laughter] [80:51] >> What does a plug mean? I don't know what [80:52] a plug is. [80:55] >> Here's a rubber circle. I would like to [80:57] >> this man I said the company I work for, [80:59] he says his mom [laughter] [81:06] uh you know what I will shout out? [81:08] >> Uh the Commit Your Code Conference. I'll [81:10] shout that one out. Uh if you listen to [81:12] the podcast, you should probably know [81:14] about it. But uh we're bringing it back [81:16] again this year. So if you have a [81:18] company that wants to sponsor, I'd love [81:19] to have a conversation with it. But last [81:21] year we had uh almost a thousand people [81:23] in person. 860 was the top total for the [81:27] conference in person and we had 8,100 [81:30] attend online. Uh I definitely want to [81:32] increase those numbers. So like I'm [81:34] really actively working towards that. [81:36] The idea of the conference, for those [81:37] that aren't aware, it's to do a $700 [81:40] conference for 40 bucks. And you should [81:42] never feel like it's a charity [81:44] conference. And we donate 100% of all [81:46] ticket sales to charity. We don't [81:47] monetize it. We don't keep it. To be [81:49] honest, people ask me about that, but [81:51] like I here's me being transparent cuz [81:53] I've had enough of this. Like I make [81:56] pretty good money doing what I do for a [81:58] living. I don't need to make it via a [82:00] conference. I just want to do something [82:01] that like is different and really brings [82:03] excitement to the community. And I feel [82:04] like this is a way that we're doing it. [82:06] But also to be completely honest with [82:08] you, I spoke at a really well-known [82:10] conference and as I came back to Dallas, [82:14] somebody mentioned like it's their dream [82:15] to go to that conference, but they can't [82:16] afford to go. And that is when I [82:19] realized that this individual in [82:20] particular had over a decade of [82:22] experience in tech. So it's not like a [82:23] junior dev that's dreaming to go to a [82:25] conference. And I realized then how many [82:29] individuals were literally missing out [82:32] on opportunities simply because income [82:34] was a barrier. So my goal was to create [82:36] a tech conference where income was no [82:39] longer a barrier to people getting [82:41] access to technical resources that will [82:43] help them grow in their careers. And I [82:45] feel like we really achieved it with [82:46] this conference. So we're coming back [82:48] again this year for the third year. Uh [82:49] still figuring things out, but I think [82:51] right now we're aiming towards the end [82:52] of October. We're going to do something [82:54] very, very big this year. um compared to [82:57] the previous two. We had to already [82:58] change venues, so we're already working [83:00] on that right now actively, but be on [83:02] the lookout for the CFP, which should be [83:04] going live in a couple weeks, uh give or [83:06] take, and we'll have something big to [83:08] announce. [83:10] >> No, I I shout out Moms because she was [83:12] the one that took me all the Pokemon [83:13] tournaments. So, if we're talking about [83:14] it, I always got to shout out Moms. [83:15] [laughter] Shout out 100 devs community. [83:17] I see so many so many folks in chat. And [83:19] of course, uh Resilient Coders if you're [83:20] in the Boston, Philly, or Maine area for [83:23] sure. [83:26] Cool. [83:28] All right. Well, yeah. Thanks, guys. [83:30] Thanks everyone for listening. If you [83:31] liked it, please subscribe. Leave us [83:32] some ratings and reviews. We appreciate [83:34] it. And we will catch you next time. [83:36] Peace. Unfortunately, Youtube does not provide subtitles for this video, Two Podcasts Four Cups, YouTube, 01km12hevfybtmv47a3vxdb3ck.