A weekly roundup episode where the host and a Zapier guest run through ten open-source repos built to pair with a coding agent, from screenshot-to-code to a proxy that cuts terminal noise by up to 90 percent.
Posted
2 weeks ago
Duration
Format
Listicle
educational
Views
12.8K
423 likes
57 · 43
Big Idea
The argument in one line.
GitHub has become the toolbox for AI-assisted development: pairing a free, open-source repo with a coding agent turns a manual, error-prone step, like describing a UI or hunting for project context, into something the agent can just execute.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You already run Claude Code, Cursor, or a similar coding agent daily and want to close the gap between what power users have found and what you've stumbled on yourself.
You waste time describing UI tweaks in words instead of just showing your agent a screenshot or screen recording.
Your codebase has grown past the point where your agent can safely change something without first hunting through files for context.
You publish code or content and want to understand how AI answer engines decide what to cite, not just how to rank in a search engine.
SKIP IF…
You don't use an AI coding agent at all; nearly every tool here assumes Claude Code, Cursor, or an equivalent is already in your workflow.
You want a deep tutorial on one tool; this is a rapid-fire listicle covering ten repos in under fourteen minutes, not a walkthrough of any single one.
TL;DR
The full version, fast.
This roundup profiles ten GitHub repos built to pair with an AI coding agent, mixing free infrastructure with agent-specific tooling. Screenshot to Code turns a screenshot into working UI code. PostHog and the Zapier Connectors repo give an agent free analytics and app integrations. Matt Pocock's Skills repo and the Vibe Coding Playbook add process and review discipline so an agent stops making unreviewed decisions. RTK compresses noisy terminal output to protect the developer's attention. Graphify turns an entire codebase into a queryable knowledge graph so an agent stops hunting for context file by file. Awesome Claude Code catalogs installable skills and plugins, Awesome AEO & SEO explains how AI search engines decide what to cite, and Docusaurus auto-generates a documentation site from markdown. The throughline: GitHub is becoming a toolbox coding agents can wire in directly, not just a place to browse code.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
The problem of describing a UI interaction in words, then the fix: screenshot or screen-record it and let the agent convert it to clean HTML/Tailwind/React/Vue code. 74,200 stars.
00:36 – 01:48
02 · PostHog
An empty or frustrating website (the 'tumbleweed' and 'rage clicking' framing) leads into PostHog: free analytics, session replay, error tracking, and feature flags in one open-source platform with a generous free tier. 37,700 stars.
01:48 – 03:09
03 · Zapier Connectors
Most apps ship a ChatGPT connector but not one for every agent; Zapier's own guest explains why they open-sourced their integrations repo so any agent can authenticate and use Zapier's connections directly, free.
03:09 – 04:30
04 · Matt Pocock Skills
The frustration of fixing an agent's mistakes for hours leads to Matt Pocock's Skills repo: small, composable process skills built from one developer's own workflow. 221,000 stars, called a top-20 repo of all time.
04:30 – 05:51
05 · RTK
Endless scrolling terminal output between a developer and their agent gets compressed by RTK, a single Rust binary with zero dependencies. Framed as reducing cognitive load first, token cost second. 76,500 stars.
05:51 – 06:54
06 · Graphify
An agent that has to hunt file-by-file for project context is solved by Graphify: mapping code, docs, PDFs, SQL schemas, and configs into one queryable knowledge graph. 108,000 stars.
06:54 – 08:42
07 · Awesome Claude Code
Feeling behind on skills, plugins, and workflows other Claude Code users have found leads to Awesome Claude Code, a curated collection you can hand to your own agent to interview you and set itself up. 52,500 stars.
08:42 – 10:39
08 · Awesome AEO & SEO
As people ask ChatGPT and Gemini instead of Googling, Awesome AEO & SEO collects research on how AI search decides what to cite. Only 2 GitHub stars, flagged by both hosts as underrated.
10:39 – 11:51
09 · Vibe Coding Playbook
Fast agent-written code that accumulates unreviewed patterns ('slop, slop, slop, slop all the way down') is addressed by the Vibe Coding Playbook, which forces a plan, pattern-matching, and review before every change. 311 stars.
11:51 – 13:57
10 · Docusaurus
A GitHub project that needs a read-me, a docs page, a sidebar, and version history is solved by Docusaurus, the documentation-site generator Meta built and open-sourced, turning markdown into a searchable site. 66,000 stars.
Atomic Insights
Lines worth screenshotting.
Screenshot to Code has 74,000 GitHub stars for solving one narrow problem: turning a screenshot or screen recording into working HTML, Tailwind, React, or Vue code instead of describing a UI in words.
PostHog gives away analytics, session replay, error tracking, and feature flags for free until you hit its event limit, and its whole product is built so a coding agent can wire it in.
A company can open-source its own paid integration layer: Zapier's Connectors repo lets any coding agent use Zapier's app integrations directly through code, MCP, or an SDK, no Zapier account required.
A repo built from one developer's personal process for working with coding agents reached 221,000 GitHub stars and ranks among the top 20 repos on GitHub of all time.
A CLI proxy that compresses terminal output between an agent and a developer by 60 to 90 percent isn't mainly saving tokens, it's protecting the one resource that doesn't scale: attention.
Mapping an entire codebase, code, docs, SQL schemas, configs, even PDFs, into one queryable knowledge graph turns 'go find where this lives' into a lookup instead of a multi-file search.
A curated 'awesome list' repo is more useful handed to your own agent to interview you about your setup than read top to bottom yourself.
A resource with two GitHub stars can still be the most substantively useful item on a list of ten, because star count tracks how long something has existed, not how good it is.
Forcing a coding agent to propose a plan, follow existing patterns, and pass a review before every change trades a little raw speed for a codebase without decisions nobody remembers making.
Documentation has a way of quietly becoming its own project: a read-me becomes a docs page, then a sidebar, then a search bar, until you're building a second product just to explain the first one.
Search traffic earned by ranking on page one is being replaced by traffic that never arrives: an AI answer engine reading a site and citing it, or not, without sending a single click.
None of the highest-value repos in this list require reading their code first; the value is in letting a coding agent consume and act on them directly.
Takeaway
Ten repos, one shortcut: pair open source with your coding agent.
WHAT TO LEARN
Every tool on this list is built to be operated by a coding agent, not just a developer, so the real skill is picking the right free repo before building the same thing yourself.
01Screenshot to Code
A screenshot or screen recording communicates a UI change to a coding agent more precisely than describing it in words, especially for interaction details like hover states and color transitions.
Free, open-source UI-to-code conversion already has 74,000 GitHub stars, a signal the underlying problem, translating visual intent into code, is common enough to be worth automating.
02PostHog
A single open-source platform can replace analytics, session replay, error tracking, and feature flags, with a free tier generous enough to run until you hit real event volume.
A product built so a coding agent can install and configure it, not just so a human can click through a dashboard, is worth weighing against a paid alternative that assumes a human operator.
03Zapier Connectors
A company giving away its own integration layer as open source lets a coding agent wire in an app connection without you writing a custom integration from scratch.
If a connector you need doesn't exist yet, you can build it against the open repo or authenticate through an existing paid account, so both the free path and the paid path stay open.
04Matt Pocock Skills
The gap between developers who move fast with an agent and those who don't is usually process, not skill: asking the right questions and following a repeatable structure before letting the agent start.
A well-designed 'skill' for a coding agent should be small and composable enough that you can read it in under a minute and understand exactly what it does before trusting it.
05RTK
Compressing an agent's terminal output isn't primarily about saving tokens, it's about protecting your own attention: scrolling walls of text you don't read is what actually slows you down.
A tool doesn't have to be complicated to be worth adopting; a single binary with zero dependencies that just cleans up noise between you and your agent is enough.
06Graphify
An agent that has to hunt through files to understand your project before it can safely change anything is spending time you could get back by giving it a pre-built map instead.
Mapping code, docs, SQL schemas, configs, and even PDFs into one queryable graph turns 'go find where this lives' into a lookup instead of a multi-file search.
07Awesome Claude Code
A curated list is only useful if you can act on it quickly; the fastest way to get value from a large list is to have your agent read it and interview you about your setup.
Framing your own uncertainty as a question the agent can answer turns a wall of unfiled resources into a personalized setup checklist.
08Awesome AEO & SEO
Search traffic earned by ranking on page one is quietly being replaced by traffic you can't see: an AI answer engine reading your site and citing it, or not, without ever sending a click.
Low GitHub star counts on a resource don't mean it's low value; in an emerging field they can mean almost nobody has found it yet.
09Vibe Coding Playbook
Code an agent writes fast enough that you stop reading all of it accumulates patterns nobody chose and error handling that isn't there, a slow-motion mess that only surfaces weeks later.
Forcing an agent to propose a plan, follow your existing patterns, and pass a review before it ships trades a little speed for a codebase you can still explain to someone else.
10Docusaurus
The moment a GitHub project needs a real read-me, a docs page, a sidebar, and a search bar, you're no longer building the project, you're building a second product: a documentation site.
Auto-generating a docs site from markdown files you're already writing keeps documentation as a byproduct of building instead of a separate project that competes for your time.
Glossary
Terms worth knowing.
MCP
Model Context Protocol, a standard way for an AI agent to connect to an external tool or data source, such as an app's integrations, without a custom one-off connection built for each app.
AEO
Answer Engine Optimization, the practice of getting cited by AI answer engines like ChatGPT and Gemini, distinct from traditional SEO, which optimizes for ranking on a search results page.
Reciprocal Rank Fusion (RRF)
A method for combining several separately ranked lists into one final ranking, used by AI search systems to blend keyword search and semantic search results before deciding what to surface or cite.
AST (Abstract Syntax Tree)
A structured, code-level representation of a program's syntax, used to parse a codebase deterministically instead of relying on an AI model to guess at its structure.
“You've got a code base full of patterns nobody chose, error handling that isn't there, decisions that nobody remembers making, essentially slop, slop, slop, slop all the way down.”
vivid repetition, names a feeling every agent-assisted coder has had→ TikTok hook↗ Tweet quote
06:10
“It maps your whole project, the code, the docs, the PDFs, even images all in one picture of how everything connects.”
clean, concrete value proposition for a knowledge-graph tool→ IG reel cold open↗ Tweet quote
10:00
“Almost nobody has written down how to show up inside those answers.”
names an emerging, under-covered problem (AEO) in one line→ newsletter pull-quote↗ Tweet quote
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphorstory
The first repo you've gotta know about is something that addresses a problem we all have. Maybe we have, like, a flip, uh, switch that we want to turn green or a button that when we hover over changes colors and moves. How do you describe them to your agents so that it could build it?
Well, that is where screenshot to code comes in. You just screencast what's on your screen, and then you give it to it, and you end up with clean code.
This thing has been incredibly popular. 74,000
stars, and I could see how useful it is. Take a screenshot, drop it in, and you're off to the races.
Next, you have a website, and maybe that's what this tumbleweed is about that nobody shows up. Or maybe you do have somebody who comes on the site, and they're just rage clicking the enter button because it's not working, and they're getting so frustrated that they never wanna come back to your site or even see anything else you build.
How do you know? What's going on? Well, that's where Posthog comes in, and they've got analytics session replay so you can see what's going on, error tracking, feature flags, all these things in one place, and best better yet, it's made in a way that you can give to your agent.
It has got 37,000 stars. I thought it was a paid product, though, Tim.
What's the deal with them giving it away here like this on GitHub? Yeah. It's open source, but, like, you can actually use you could just go through the post Hog install
wizard, and they give it away. They have a very generous free tier. And one of the things that I really like about this after having gone through the install wizard on a couple of my own, you can actually stitch your marketing website directly to your your in app product and get those get those rich analytics.
And they just give it away, um, until you obviously hit your event limit, and then then you have to you have to start paying. But very, very generous and super easy to install. Okay.
Love this one. Of course, we'll have links to this and this whole report, um, below in the comments and in the description.
Next. Half the apps you use shipped with a connector to ChatGPT.
Haven't shipped with a connector. Is it half? It's even more.
A lot of the apps, when I try to work with Claude on them, Claude doesn't have a connection to it. And so what you have here is Zapier, which has a set of connections that I thought you all charged for, but I see that it's on GitHub.
What's the deal here?
Yeah. That's so there's this is actually pretty it's near and dear to my heart. So, like, one of the one of the key things about the connectors repo is that we're actually starting to expose all of the integrations that we have Mhmm.
In just good old fashioned coding language. So that way, you can actually use the connections and the integrations that we've built over the years. And you can either log in with the app credentials that you have, or if you're a Zapier customer, you can authenticate with your with your Zapier account.
The beauty of this is that we just want you to get started. So that way, if there's a connector that's missing, go check out the the Zapier connectors repo. Grab that.
You can use it in MCP. You can use it in your SDK. It really doesn't matter.
And then you can bring your own off, or you can use Zapier to authenticate,
and you'll just be off to the races. So this one's really fun. If I wanna connect into SharePoint or Microsoft To Do or Notion or any of these and they happen not to have a connect connector, I could come in and use this completely for free.
Don't even have to pay Zapier. That's correct. I don't know how you're doing it.
I don't know why you're doing it, but I'm glad it's there. If you like it, hit the star on it. Show them appreciation.
Let's go on to the next one. Alright. You have a great idea to build something.
So you sit down in front of your agent and you tell it to start building, maybe that first version looks good and you're excited. But then for the next two hours, you're fixing your agent's mistakes. You're getting frustrated by how slow things are, and you're wondering why is it that this seems so much easier for everyone else?
Well, Matt Pocock, the developer teacher has got your back.
What he did was he said, look, developers who are operating well have a structure for how they do things. They don't just sit down and start to create.
They think. They get they ask questions. They go through a process.
And he said, you know what I'm gonna do? I'm gonna take the process that I use myself as a developer, I'm gonna make them available for people. And that's what this repo is.
I have got to tell you this thing is taking off 221,000 stars. It is one of the top 20 repos of all time on GitHub.
He and I just did a whole session on this where he showed it to me. It's incredible. Thoughts on this one?
I love it because he's basically made these super composable, and they're small. So they don't get in the way of the actual engineer when they're developing using these coding agents. If anything, they're kind of supplemental.
You have full control over them. This one's super helpful. If anything, it kinda feels a little bit simple when you click into one of the skills.
It's short. It's clear. You see exactly what's going on.
You're not meant to use it blindly. You're meant to understand it, and then you can use it. Alright.
You give your agent a task. Yes. Here's number five on the list.
You give your agent a task, then you sit back and you watch it work. It runs a command, and the screen fills with output. Then it runs another and then another screen full of output.
It's all this. I start to blank on it, and I don't read any of it. That's why you told me, Tim, about the fifth repo that we're talking about today.
It's RTX. It sits between your agent and your computer. It squeezes out all that output before your agent before you get to it and before your agent reads it.
I'm actually, to be honest with you, not as clear about this as I should be. Teach me. What is this?
It just reduces cognitive load. So it's actually not gonna, like your eyes aren't gonna track on all of this. It's going to actually go ahead and jump ahead of the line, and it's gonna squish all of this stuff and compact it nice and neatly.
So that way you're not sitting there watching all of this happen. It's gonna it's gonna save you some stuff. Brain
here, this run, this CSC, the whole thing in this bash.
It's compressing. That's right. Yep.
Exactly. So that way your eyes aren't tracking on that the whole entire time you're building these things. It's it's actually quite nice.
I've heard people talk about it as a way of saving tokens, and you said, Andrew, that's not really what it's about. You might save a couple, but It does reduce some tokens, but it's not like it's not like this big, like, token saving secret.
It's it's really it does reduce the amount of it to some extent, but, really, what it's doing is it's saving your the the most valuable resource, which is, like, your your cognitive load and your and your brain power. Like, you don't wanna get fatigued.
76,000 stars on GitHub. Incredibly popular.
Alright. Thanks. Let's go on to number six.
Your agent doesn't know your project. So before it could change anything, it starts to go hunting. It opens a file, then another.
It looks for something and where it lives. And pretty soon you sit there and you go, you know what? I could have answered that question in five seconds.
In comes Graphify. That's where it's helpful.
It maps your whole project, the code, the docs, the PDFs, even images all in one picture of how everything connects, and it makes it easier then for your agent to understand without having to dig through your files. It's got over a 100,000 stars. I keep seeing it as one of the top repos of the week.
You have any opinion on this one?
Andrew, I am a knowledge graph fanboy. Tell me. And and they they they're giving it away here.
Right? Like, obviously, you can you can go you can level up. But this is pretty cool because once you have a knowledge graph, you can see the relationships between all of the things, and it just makes things make sense.
It's incredible. Alright. Number seven, you got Cloud Code working.
It's running. Everything's good, but you keep getting the feeling that everyone else is doing so much more than you are with it. They've got skills.
They've got plug ins. They're talking about workflows that you never heard of. Meanwhile, you just don't know where and what to where to grab these.
Well, it turns out there is a repo. It's called awesome Claude code. It has got a hugely not just large, but a curated collection of skills, of plug ins, of tools.
I put some of them over here. The issue I've got with this, and I've heard about how good this is, Tim. It is fantastic,
but I don't even have the time to read through all of it, let alone to know that this one is gonna be useful for me, but this one is not. What do I do with all these? And I'll open it up in GitHub while you tell That's the beauty of of GitHub repos like this is that you actually don't have to spend the time parsing through the repos.
You can simply just have your agent do it for you. How? So that way if you wanted to get up to speed quickly, could say, hey.
I'm new to Cloud Code. I really wanna use this repository to set myself up for success. And it'll walk you through all of the things that you might need to do.
It'll ask you questions about, like, hey, what are you trying to build today? What are the most common tasks that you have? And then it will go through that repository, and it'll actually tell you the things that you need to either enable or add as skills or set up from an environment perspective.
I see. Okay. I just give it start.
Gotta start there. It interviews me and knows based on what I'm doing which of these tools will help me do a better job. This I can see, uh, our audience for sure using.
Next. Oh, and by the way, 52,000 stars. It's been incredibly popular.
I've seen it a lot. I'm I'm kind of a little jealous, by way, of people who have these awesome skill collections. Everyone talks about them.
Everyone contributes. They get all kinds of attention. So many people are being helped by it, and all they're doing is curating.
And I I love it, but I'm a little jealous, to be honest with you.
I need something like that. Same.
Alright. Number eight for today.
In the past, we used to go to Google to search, but people right now are going to ChatGPT, to Gemini, to Claude, and we're basically starting to ask questions like, I need a bookkeeping company. I need a lawyer. And that's where they're getting their list of answers.
Well, how do you get cited by it? When it came to search engine optimization, a lot of us had heard a lot about it. A lot of us had learned about it, started using it.
But when it comes to AEO, answer engine optimization, we're kinda newbies, and that's where this comes in. It's awesome AEO and SEO, and it's basically collecting the research on how AI search actually decides what to cite.
I wasn't sure if you gave me the wrong repo, Tim, when you told me that we should include this one because when I looked at it, it had two stars.
Did you mean to give me this? It's a I did. Little sauce.
Right? Um, the reason why I like this is because I've I've got years and years of experience doing SEO. Mhmm.
And, essentially, nowadays, what you have to do is just understand how these engines work. Right?
And with engine answer opt or or answer engine optimization and search engine optimization, a lot of that stuff is out there in the wild. What the beauty of it though is that this this gentleman put this package together, and it's actually really well done. There's a lot of information here that's following the new standards.
It's actually pulling research that's been been curated over the over the past year or so.
And it's relatively, like, right on the nose. So I just wanted to put this one in here because it's it's worth it's notable. Right?
Like, you should probably try this one out, use it, and it'll definitely help you out. Alright. I'm gonna hit the star on this one.
Give it three stars, and let's go on. And by the way, speaking of stars, if you do like this, you can't give me a star on YouTube, but do subscribe. Give me a thumbs up, and obviously not obviously, complain.
Even if you don't like it, I wanna hear it. Let me know in the comments, especially if you don't like it. I wanna know.
I've been adjusting and speeding up. You can see me constantly tweaking based on your feedback. So subscribe, like, but also let me know what else we can do to make this more useful for you.
Number nine. Alright. Your agent writes code fast, fast enough that at some point you stop reading all of it.
Then a few weeks later, you've got a code base full of patterns nobody chose, error handling that isn't there, decisions that nobody remembers making, essentially slop, slop, slop, slop all the way down. That's where vibe coding playbook comes in. It treats your agent like a junior engineer, one that has to propose a plan before it writes anything, follow patterns you already use, and pass a review every single time.
You keep the speed, but you skip the big mess.
Why'd you pick this one to for us to include? I like this one because it really does do that. It kind of, like, forces you to slow down as kind of like a if you're if you're new to this game, it's actually gonna structure your your code base the right way.
It's gonna ask you the right questions, and it's gonna make you approve things before you just, like, push it all out there. Right?
Makes it nice and organized. Helps you get through it a little bit easier. Keeps things clean.
That's why I like it. Lot less slop. Only 311 stars, now 312.
One of the things that I learned by talking to you and other people who are further ahead than me is, Andrew, slow it down. Ask the right questions first, figure out what you really want, and then you won't have these problems that you have later on. Alright.
I appreciate it. Let's go on to number 10. You build something good, you put it on GitHub, then someone asks you how to use it.
So you start writing a read me, then another page, then you need a sidebar for a site that you've created with all the documentation and a search bar and somewhere that you're gonna put your version history. And pretty soon, you're no longer just focused on the thing that you put on GitHub or the project that you made. You're now building a whole website and setting telling the thing what to build.
And that's where a DocuSign from Facebook comes in. It will create these beautiful pages that we've seen. I am actually here in the right.
I picked out actual sites that were built using it. Why did you pick this?
I love DocuSource because I'm I'm kind of like a a technical doc fanboy You are? As well as a as a knowledge graph fan.
Okay. I am. Because that's as as a developer, like, that's where you wanna spend a lot of your time, like, really understanding how to use the technology.
Right? And what DocuSign does is it's gonna take all of the information that you've written in your inside of your repository, and it's gonna help you create those tech docs. So, like, whenever you go to a docs.zapier.com or a docs.gitlab.com, like, you're gonna find the information that you need.
And it's gotta be structured in a way that's easy to easy to read and also understand. The beauty of this is that it's creating it in markdown files automatically. So that way it's easy for agents to actually consume.
It spends Lex tokens, and then it informs the agent what to do.
Docusource handles all that stuff for you. It's incredible. It's a good one.
You all over at Zapier have been so far ahead of everyone in making content that is accessible to agents, and now I can see why you're excited about this. 66,000 stars.
Alright. We've had a big list here. If you got anything of value, you should know that every week, I break down the top 10 trending GitHub repos of the week right here on the channel.
When you subscribe, you'll find out about it. And, uh, I've got my email address on the channel. It really is me.
And if I get overwhelmed, someone on my team will help me, but we are trying to figure out what you're building. I wanna see it. Send me over what you've built.
A lot of people take these and then fork them and create their own versions. I wanna see it. If And there's anything else that you're building and creating, I wanna see that too.
Speaking of our weekly GitHub, uh, show, I've got one right here that I think you're gonna like if you got this far. I'll see you in that one.
The Hook
The bait, then the rug-pull.
Ten open-source repos, one thread: pair the right free tool with a coding agent and skip the manual, error-prone part of building software. Only one of the ten actually saves tokens, the other nine save something more valuable: time, context, and review discipline.
The playbook treats a coding agent like a junior engineer instead of an autocomplete, trading a little raw speed for a codebase without decisions nobody remembers making.
Steal forany project where an agent is shipping code faster than a human is reviewing it
CTA Breakdown
How they asked for the click.
VERBAL ASK
10:25subscribe
“if you do like this, you can't give me a star on YouTube, but do subscribe. Give me a thumbs up, and obviously not obviously, complain.”
Self-aware ask that reuses the episode's own running joke about GitHub star counts, and explicitly invites negative feedback in the comments rather than just asking for praise.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Two hosts count down the week's ten hottest GitHub repos, from a security-scanned Claude plugin index to a self-hosted finance tracker wired to 2,500 banks.
Andrew Warner and Adam run down this week's most-starred GitHub repos — three AI design tools, a parallel-agent "tokenmaxxing" rig, a video-comprehension skill — and pull the geophysicist who built the #1 trending AI job-search framework on camera to explain how it got him hired.
Andrew Warner sits down with Matt Pocock to unpack the Claude Code skill library that passed 200,000 GitHub stars, and the small habits of alignment, pruning, and teaching baked into it.
Two hosts react to seven creators' first hands-on tests of OpenAI's GPT-5.6 Sol against Claude Fable 5 — and land on a manager/worker split, not a winner.
00:18
00:41
00:51
01:17
01:18
01:37
01:43
01:53
02:13
02:23
02:33
02:43
02:53
02:59
03:14
03:24
03:34
03:44
04:04
04:14
04:25
04:35
04:45
04:55
05:05
05:15
05:26
05:35
05:53
06:06
06:27
06:35
06:47
06:59
07:10
07:21
07:28
07:46
07:53
08:04
08:25
08:36
08:47
08:57
09:08
09:29
09:40
09:51
10:01
10:14
10:23
10:33
10:44
10:54
11:15
11:25
11:35
11:45
11:56
12:06
12:22
12:33
12:48
12:59
13:24
13:30
13:41
13:50
18:45
37:04
27:14
21:02
Video of the Day29:45
32:13