A YouTuber wires Claude Code into an agent-native design canvas called Paper and lets it generate thumbnails, Instagram graphics, and a full pitch deck from plain-English prompts.
When a design tool is built for AI agents instead of humans, an agent like Claude Code can directly create, edit, and iterate on real design files through plain-English prompts, collapsing thumbnail, graphic, and deck design into a chat conversation.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
A YouTuber or content creator who currently pays a thumbnail designer or spends hours iterating on thumbnails manually.
A solo creator who needs Instagram graphics, pitch decks, or landing pages but has no design background.
Someone already using Claude Code who wants to see a concrete example of an MCP-connected external tool doing real creative work, not just code.
SKIP IF…
You need pixel-perfect, brand-locked design systems — this workflow is fast iteration, not precision control.
You're not willing to pay for both a Claude subscription and Paper credits once free usage runs out.
TL;DR
The full version, fast.
Paper is a Figma-like design canvas built to be controlled by AI agents rather than humans, and it exposes an MCP server so tools like Claude Code can read and write design files directly. Riley Brown connects Claude Code to Paper once, then uses chat prompts to swap faces into other creators' high-performing thumbnails, apply an Instagram graphic's exact black-gradient text-overlay style to his own photo, and generate a full six-slide sponsor pitch deck styled after Perplexity's brand — pulling images from a referenced canvas frame and text from his own website. He closes by deploying the finished deck as a live Vercel-hosted landing page in a single follow-up prompt. The workflow trades manual design tool skill for prompt iteration: select reference images, describe the edit in one sentence, and generate multiple variations until one lands.
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.
Cold open teasing Claude Code controlling a Figma-like canvas; reveals the tool is actually Paper, an agent-native design app; previews the three use cases (thumbnails, Instagram graphics, presentations/landing pages).
01:41 – 04:14
02 · Set up Claude Code x Paper Connection
Downloads Paper, copies the MCP server link from paper.design/docs/mcp, pastes it into a new Claude chat to wire the connection, then tests it by asking Claude to name the open board and draw a simple graphic — Claude successfully labels a frame and draws a line.
04:14 – 10:21
03 · Creating YouTube Thumbnails
Uses Paper's built-in AI image generation to swap his face into a high-performing thumbnail and change its text, then uses Claude Code with a YouTube-thumbnail-scraping skill to pull a chosen creator's (Kallaway) 12 best thumbnails into a frame, swaps his photo into one, and iterates through several style variations including a multi-reference-image version with icons.
10:21 – 16:33
04 · Instagram Graphics with text and overlays
Starts a new board, references an Instagram graphic pulled from another account, and prompts Claude to replicate its black-gradient text-overlay effect onto his own photo with live-editable text, generating several headline variations and swapping in different portrait styles and background colors.
16:33 – 21:22
05 · Creating Presentations with Fable 5
Switches Fable to high effort, references a whole frame of his photos with one link, and prompts a six-slide sponsor pitch deck styled after Perplexity's brand (via a Firecrawl-based scraping skill), then mid-generation adds his personal website as a source for updated stats and copy.
21:22 – 23:58
06 · Converting Designs into a site on the internet
Asks Claude to turn the finished deck into a responsive web/mobile landing page and deploy it to Vercel in one prompt; the resulting live site (riley-sponsor-site.vercel.app) embeds the toggleable slide deck directly on the page. Closes with a recap and disclosure that the video isn't sponsored.
Atomic Insights
Lines worth screenshotting.
Paper is a Figma-like canvas designed to be operated by AI agents through an MCP server, not just by human mouse clicks.
Once Claude Code is connected to Paper via MCP, a single chat prompt can create, edit, or replace elements directly inside a real design file.
Selecting multiple reference images with shift-click and describing the desired edit in one sentence is enough to generate several thumbnail variations at once.
Text in these AI-generated graphics remains directly editable — it's live text, not flattened into a raster image.
You can reference an entire frame of images with one link instead of copying each image individually, which lets Claude use all of them as context in a single prompt.
A scraping skill (built on Firecrawl) lets the agent pull another site's visual brand style and apply it to a new deck or page without manual style extraction.
The same chat session that builds a slide deck can, in one more prompt, turn that deck into a responsive landing page and deploy it live to Vercel.
Selecting a specific generated image and copying its asset link lets you tell the agent exactly which version to reuse, instead of it guessing.
Background removal and gradient/overlay effects that are manually fiddly in traditional editors are single-sentence requests in this workflow.
Every agent action inside Paper consumes credits, and Paper moves to a paid subscription once free usage is exhausted.
Takeaway
AI agents can now operate design tools directly, not just generate static images.
AGENT-NATIVE DESIGN
A design tool built from the ground up for AI agents lets you replace manual thumbnail and deck design with plain-English prompts and rapid iteration, at the cost of paying per agent action once free credits run out.
An MCP connection between a coding agent and a design tool lets the agent read the current canvas state and write new elements directly into a real design file, not just output an image to import.
Referencing multiple images at once (shift-click, or a whole named frame) gives the agent enough context to blend styles, swap faces, and match layouts in a single prompt instead of several rounds of manual editing.
Text generated this way stays live and editable, so headline changes don't require regenerating the whole graphic.
A scraping skill that pulls a reference brand's visual style (fonts, colors, layout) lets you target a specific aesthetic ('make it look like Perplexity') without manually rebuilding a style guide.
The same agent session can go from raw photos to a finished slide deck to a deployed, responsive website in three sequential prompts, collapsing what's normally a multi-tool handoff into one conversation.
This workflow still costs money per agent action once free usage is used up, so it functions as a subscription cost to weigh against hiring a designer, not a free replacement.
Glossary
Terms worth knowing.
MCP (Model Context Protocol)
An authenticated API standard that lets an AI agent read and write data to a specific platform or application — here, it connects Claude Code to the Paper design canvas.
Paper
A Figma-like design canvas application built specifically to be controlled by AI coding agents, available for Mac and Windows at paper.design.
Fable 5
The AI model Riley Brown runs inside Claude Code for this workflow, described as highly capable but token-hungry compared to other models.
Firecrawl
A web-scraping API that lets an AI agent search the internet and extract visual styling or content from any website to reuse in a design.
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
story
00:00Welcome to the new world of agent based designing with Fable five. This right here is Claude Code running Fable five, and it's fully controlling Figma.
00:11This is actually the agent native version of Figma called Paper. And in this video, I'm gonna show you how I use AI agents to do all of my design work for YouTube thumbnails, landing pages and other websites, Instagram graphics, and documents and presentations, all by prompting an AI agent, Fable five, to create designs on this Canvas like app that's built for AI agents.
00:35Let's dive into the video. Alright. So this right here is Claude code running in the Claude desktop apps.
00:42As you can see here, I'm running Fable five. This is the prompt that I just typed in, and Claude code is going to control this Canvas like board right here. And so here you go.
00:53The AI agent is actually starting to create a presentation directly on this board. My prompt was to create a presentation on how to use paper.
01:03And because of a skill that I'll show you how to set up later, it can scrape all of the assets from the Internet and place it directly in the presentation. Look at this. Fable five is literally designing this presentation right here.
01:17And look at this. It can very easily create really cool graphics. You can create landing pages, presentations, and it will create these graphics.
01:27Look at that. So this is how Claude Code writes designs to the paper MCP server, which we'll be talking about later. And then Claude Code reads the context, and here's the next step.
01:37Passes it to the paper design file, and that's how it works. It was a really useful graphic actually. And so paper is just a desktop app that you can download at paper.design.
01:49I believe they have if you click download paper, it will automatically download it if you have Mac or Windows. I do believe it works on Mac and Windows.
01:57When you finally get set up, your screen will look something like this, and you will have paper open on your computer. Down here at the bottom, you're going to see this button that you can press called using agents, and you can just hit get started. And here, what you can do very easily is you can actually literally just copy this link, paper.design/docs/mcp, and you are going to go to Claude, the Claude desktop app for those of you who use Claude.
02:24You can also do this with Codex. You can do this with Cursor. Any AI coding tool where there's an agent running on your computer, you can do this.
02:32And what you can do is you can go to a new chat and you can say, I want to set up the MCP server to control my paper dot design, and then you're going to paste in this link.
02:46And then Claude will set up the connection for you. And so after you've typed that prompt and entered it into Claude code, what I want you to do is open up paper next to Claude code, and we're gonna test the connection. So you're just gonna hit new file, and we're gonna name this Riley site.
03:03And all I'm gonna do is I'm gonna type hello on the board. Remember, this is a very similar process as Figma.
03:10It's if you've used Figma before, it'll feel very similar. Hey there, Cloud Code. I want you to tell me what board do I have open on paper.
03:18Tell me the answer to that. And then I want you to create a single simple graphic just so I know that you can control my paper dot design. And I'm not gonna use high for this because I'm just gonna use low.
03:30We do not need to waste those Fable tokens and we can run it. And now we're basically just testing the connection between Claude and paper. And there you go.
03:40It is now controlling and it named this frame hello from Claude Code. And now we have Claude Code fully controlling this Figma like board.
03:51And all of this is HTML based, so it's really fast and really high quality. Hello Riley, drawn directly onto your Riley site canvas by Claude Code.
04:03And it drew a squiggly blue line. Not quite sure why it did that, but that's kinda cute. So there you go.
04:08The test is complete. We have Claude code connected to Paper. It really is that easy.
04:14So the first use case of using these two tools together that I wanna talk about is creating YouTube thumbnails or any type of graphic. Graphic. So directly inside Paper, they actually have built in YouTube image generation.
04:27And so if you click right here, you can use the best AI image models directly built into Paper. And so what I can do is I could very easily just grab this thumbnail right here, which is a really high quality thumbnail, and then I could grab a photo of me. And so I could put these photos right next to each other.
04:43And I could say I could select this thumbnail, and I can hold shift and click this one. And I could say replace the man in the middle with this guy and please change the text to fable verse GPT 5.6.
05:06And then I can just generate more thumbnails. And so this is a really fun way to iterate and create graphics directly inside Paper. And so for this process, Claude code isn't being used.
05:18However, to import high quality YouTube thumbnails, that's where Claude code is gonna come in handy. I'll explain that in just one second.
05:26As you can see here, the thumbnails are starting to come in, and look how easy it is to replace any character in any thumbnail with me using Paper. And, I mean, look at how high quality this is.
05:38And so the question then becomes, how do I put these high quality thumbnails from other YouTubers inside Paper? That's where Claude Code comes in. So I can go to Claude Code, and I can say, a guy named Callaway, many of you may know who this is, he is one of my favorite YouTubers.
05:53I want you to find his 10 best thumbnails, uh, actually, his 12 best thumbnails and import them in a frame inside Paper.
06:01So I want his 12 best performing thumbnails in Paper, and please import them so the images are in the Paperboard. And so we can fire off Claude code here.
06:12It'll go scrape the thumbnails and put them on the board. As you can see here, it's using the skill slash YouTube thumbnails, which I'll paste in the description.
06:23This will allow you to use this skill. Claude Code actually won't be able to do this by default, but if you use this skill, and it's free in the description, you can scrape thumbnails from YouTube.
06:34And look, you can see that the AI is working down here. So Callaway top 12 YouTube thumbnails. Look at this.
06:43It's creating this little presentation, and it's just gonna import all 12 of them. And there you go.
06:50His thumbnails are being imported Just like that. There's six.
06:58There should be six more coming in. You can see here Claude's working on that. There's the next three.
07:06This is a really good one. There you go. It just created a frame with all 12 of his YouTube thumbnails.
07:14So now I can just copy this photo of me, and we're gonna bring it up here and place it right next to Callaway's YouTube thumbnails. And now we can select from whatever one we like the most. For this one, let's go ahead and let's use this one right here.
07:29Now, I'm gonna just copy this image and we can just place it right here. And now, we can actually replace we can select the Callaway image.
07:38I can select this right here. Replace the man, the guy holding the phone with this guy.
07:53And on the phone, it should be the same but with an app store icon. Right?
08:01Kinda making it for my niche. And so we can run this just like that.
08:07And there you go. Here they're coming in. It's not bad.
08:10Oh, I like that it changed the background of this one. And look at this. This one might be my favorite one so far.
08:16But it gives you just like seven different options to choose from. And this allows you to ideate around and come up with the right format for your YouTube thumbnail. Okay.
08:24Now let's use more than one image reference. So I just grabbed this thumbnail, pasted it up above. We're gonna grab the same photo of me and then these three icons.
08:33Check this out. So I can just come here. Always select the image that you're editing.
08:37Right? So this image right here is a 16 by nine image. Whatever the final aspect ratio you want, you should select that one first.
08:44So I'm gonna select this image, and I'm gonna come over here, and we're gonna use this image editing feature. And now we're gonna shift, click here, click here, click here, click here.
08:56Now all of these are reference images. Please can you change the text to you're still early And then replace the icons with the three icons that I've pasted here. And also replace the guy with the other guy.
09:12Now, what we can do is we can just create as many versions as we want and we'll watch our new thumbnail generate before our eyes. And there you go.
09:24For whatever reason, I think it's because of the lighting behind my character and kind of the lighting. It wanted to make it this darker color, but I could have just included that.
09:35Keep the same white background, the same one that's in the photo, and you can remove the lighting around the character. You can just keep iterating. I think that's what's happening is the lighting of the character makes it want to be darker, but this is still pretty good.
09:50And there you go. It created the graphic. It looks a lot more like the, uh, Kane's original one, and these are pretty solid.
09:59And so, yeah, that is basically how I create my YouTube thumbnails now is I can just rapidly iterate. And I don't pull from other people's thumbnails all the time. Oftentimes, I'm using previous thumbnails.
10:09I have worked with thumbnail designers in the past, and we'll actually start from those as well. But this is just a way to, like, really quickly iterate until you reach the concept that you want, and I love this workflow.
10:21So the next workflow that I wanna talk about is Instagram graphics. And for this one, I actually wanna start on a new board.
10:30And so we're gonna call this one Instagram graphics. So I just found this Instagram graphic, um, or I just grabbed this graphic from this account wealth.
10:40And I just wanna show you that you can create a graphic just like this in paper. And you can actually have a text overlay and you can actually edit the text directly or the AI agent can also edit that same text directly.
10:54I'll show you what I mean. So what I'm gonna do is I'm gonna take a photo of myself. I'm gonna airdrop that photo to myself.
11:00And now we're gonna paste this right next to this graphic that we pulled from Instagram. There's a lot of interesting things going on with this graphic here. You can see it has this, um, black gradient effect behind the text, which I think is really cool.
11:15And so what we can do is we can just open up Claude and we can say we can go to a new chat. Please see the new open paperboard that I have right now.
11:25What I want you to do is I want you to please create an Instagram graphic of the man, and so don't generate a new image.
11:34What I want you to do is I want you to take the image of the guy and add the black gradient effect that you can see.
11:43You see in the image, there's like this black gradient effect behind the text. And then I want you to add the text overlay so I can edit that text, except I want that text to be white.
11:54And instead of it saying wealth right above it, have it say Riley Brown. Do this over the image on the right of the guy wearing a hat. And then what I can do here is I'm actually going to while it's working.
12:05Right? You want this takes some time. So it's always good to multitask.
12:09What I'm gonna do is I'm gonna select this, shift click, select this. Can you please generate an image of this guy in the same style as the character holding the video game except don't have any text overlay or that background.
12:24Make the background just black, nothing else. I just wanna be in the same style with the guy wearing the same clothes and don't change the hairstyle.
12:32And so we are generating a character. Okay. You can see this right here.
12:37I believe the agent is creating this right now. And you can tell the agent is working on it because it has this animation right above it. So we are creating this new Instagram graphic.
12:48It added the photo to the graphic, which is cool. That's step one. Now it should add an overlay.
12:53There you go. It just added the same overlay. Now it should add text over it.
13:00Look at this. Riley Brown. How long until you need to work to buy GTA ultimate edition?
13:06Okay. Please make another version that says, this is how you get Claude to control paper.
13:13That should be the text overlay except use one of the other images of me that I generated to the left. And actually, I wanna select the exact image of me that it should use. And so if I wanted to use this one, I can select this.
13:27And notice here there is a copy link. I can copy this link and just I can say this one. Please use this image for this and then we can run it.
13:39And so this little link just points at this asset so we know so we can tell Claude Code the exact asset that we wanna use. And so now it should create another variation. Got it.
13:50Building version two with that dramatic dark portrait. And then I can copy let's go ahead and copy this one. Make a version with this one as well, except come up with a few variations of the text.
14:01The point is is I want it to be good copy for the same use case of using claw to control paper.
14:09And then please change the paper, like, have paper be blue in the ex the one that you've already created, like, in the text and have the Riley Brown be red, same size, and then make three more versions. And use the other images that I've created, like the images that are on the board that we haven't used yet.
14:29And there you go. Look at this. We have this graphic and we have this overlay, and it matches this style right here.
14:36It's actually kinda hard to get this effect. I've tried to do this manually. It's kind of annoying.
14:40But here, you can just ask for these graphics and you can basically do anything. I'll show you another one. Oh, it's changing the color.
14:47Let's go. See, it changed it to Riley Brown. And here comes another one.
15:00I don't open paper anymore. Claude does. That's kind of fire.
15:04And what we could do is you can actually just click on these images directly. So I can click on this image and then I can just say you can edit the image directly in the design. So only change the background, make it a cool light background, and then I can generate some variations of this.
15:24And now, this image was just generated and it has a lighter background. Now, what if I wanted to replace the image in this graphic with the image of this character with a lighter background?
15:37What you're gonna do is you're just gonna select it and you're going to copy it. And then you're gonna click on where you want the image to go, and you're just gonna press command shift r and boom. It replaces that image with the one you generated.
15:51And there you go. It just added the paper. It made it blue, which is the the brand styling of paper.
15:59And you can create these really cool Instagram infographics. And I just think this is so much fun to do.
16:06Okay. So this is how you create Instagram graphics. It's really fun to, like, quickly iterate.
16:11You can directly edit all of this text, by the way. Like, I don't open right?
16:17You can just edit this text direct. This is not generating an image where the text is an image where you have to edit it. Like, you just edit it directly, which I really like.
16:25And so this is a really fun use case especially if you wanna create Instagram graphics or any graphics for that matter. Now, what I wanna talk about now is creating presentations and landing pages with ClaudeCode and Paper.
16:38So we're gonna do this on this same Canvas board and we're going to use the same chat session with Fable five. Now what I'm gonna do is I'm gonna turn up the effort a little bit with Fable and we're gonna go to high. Now what I'm gonna do is I wanna create a personal landing page.
16:54I'm gonna use these images on this presentation that we're creating. So what I wanna do now is I actually want to remove the background from all of these images.
17:03And you can just do that. You can come here and you can just remove the background really easily.
17:10And we're gonna place these on the site that we create.
17:15Alright. So here's a tip. If you wanna reference a bunch of things, what I like to do is I just like to use this f.
17:21Uh, I can create a frame. And so you can just drag a frame around these images and boom. Now I am on this frame.
17:29Now what you can do is you can reference the entire frame and you can even name the frame and you can call this, like, photos of me. And what I can do is I can now reference the entire frame.
17:41And so instead of having to copy the link to all of these images, I can just copy the frame link and give it to Claude. So now I'm gonna give it this frame link.
17:49I'm gonna say, hey. So in this frame, I have a bunch of photos of me. What I want you to do is create a six slide, high quality pitch deck, not even a pitch deck.
18:00This is a presentation pitching on my creator business.
18:04Right? I'm a content creator, and I educate people.
18:08I wanna make a deck for sponsors to show that I make really high quality education tutorials, do a ton of research on YouTube, is use the scraper skill that we've created. I want this brand to look like Perplexity.
18:21I think they have a great brand, and so I want this deck to kind of match that style. Remember, keep it simple. Don't have footers.
18:28I don't like footer text. I don't even like subtitle text, but I I like crisp headers, good design, cool graphics, simple layout, explain things very simply.
18:39I like white space. Create a six, uh, deck, but each one of the slides should have one of these images of me on it and use one per slide. So now we can fire off that prompt, and Claude will go.
18:53It'll look at this frame, grab all these images. Now it is using a skill called the scraper skill. And for this, it's using an API called Firecrawl, which allows it to search the Internet, scrape the styling from any style like, any site on the Internet, and it will use that same style in the presentation website or whatever it is that you create.
19:17And so now we're just gonna wait for the AI to build it. Alright. Here we go.
19:21It is now creating a deck. Look at this. It's creating a deck and I don't wanna get in its way.
19:26So I'm gonna just move this stuff down. But we have an AI building a deck. Um, and by the way, you know, your Fable might not know this already, but, like, just make sure you take a look at that first deck, you know, you have some text over the image.
19:40So don't do that again. And so Fable five is now working on our deck. And you see here, it pulled the Perplexity design, same colors, same font, and it placed the image on the deck.
19:54Deck one, here's the cover, what I do. Oh, and here we go. It says, I teach people how to build with AI.
20:00Tutorials, deep dives, and build alongs. That's exactly true. Very simple deck.
20:05I love this. I love decks like these. I do not like it when they make it cluttered.
20:10I do hope it creates some cool graphics. Oh, there you go. It put me on this blue background.
20:30And there you go. So now it got, I gave it my personal website and it updated the reach.
20:37And now it's actually up to date, which is exactly what we want. Hours of research behind every video.
20:45Love it. And there you go. Research, test every tool, build on camera, teach it simply.
20:52And it's just gonna finish building this full deck here, which is awesome. Oh, it even moved the image. There you go.
20:58I dig it. There you go. Oh, it looks like it's doing like a different color background for the last slide, which is pretty solid.
21:06And there it is. It's that easy to create a little slide deck with AI using Cloud Code.
21:15Now what I wanna do as a bonus, we're actually going to create a personal site that has this deck in the site and it's gonna be optimized for both web and mobile and we're gonna get it on the Internet in a single prompt. Hey. I really like this full slide deck.
21:30You did a great job. Now what I want you to do is create a website, a landing page that's optimized for web and mobile that allows people to just see Riley Brown at the top.
21:41And below it, it just says, uh, it has, like, one cool line. And beneath it, it has the slide deck where people can toggle which slide.
21:49I want you to create this website. And then I want you to deploy it to Vercel and send me the link so that I can see it on the actual Internet and I can send it to people. And we are just going to fire off this prompt right here.
22:02Okay. Here we go. So now we have our site that was completed.
22:07I can click on this and open it up. And look, we have our slide deck in a website that's deployed on the Internet. You could go to this site, riley-sponsor-site.vercel.app, and we have the slide deck right here directly in the website as we created it in paper.
22:27And you can see here, it added some more things. It found things from my website and created a really responsive website that I think looks pretty damn good.
22:36In fact, I like this more than my existing website. So I might use this. This is pretty solid.
22:42And there you have it. Today, we talked about generating graphics. I first talked about YouTube thumbnails, and then we created these Instagram graphics, and then we created these presentations.
22:51Right? You can just ask Claude Fable to create a graphic for you, and it will put it on this paperboard. And then you can also create websites.
23:00And you can even put these, uh, presentations on your website, and that's exactly what we created. And you saw the site that it created here.
23:09Really easy to get set up. You just go to paper.design, sign up. And every time that you use Claude code to do something on paper, it does use credits.
23:18And once you do enough, it will begin charging you. So there is a paper subscription at some point. I don't know what it is.
23:25This video is not sponsored by Claude nor is it sponsored by paper. I just really genuinely use these tools a lot, especially for thumbnails and Instagram graphics.
23:34And I am going to continue building out my workflow. And probably in the next few I will share some more things about how I use Fable for design.
23:43I've only been able to use Fable for a total of, like, four or five days so far. I'm still learning a ton of things. I'm relatively new to using paper, and so as I build out my workflow, I'll make sure to share it here.
23:54Uh, anyway, thank you guys for watching this video. I'm Riley Brown. Peace.
The Hook
The bait, then the rug-pull.
Riley Brown opens with a bait-and-switch: it looks like Claude Code is controlling Figma, then he clarifies it's actually Paper, an agent-native Figma alternative built from the ground up for AI agents to read and write design files directly.
CTA Breakdown
How they asked for the click.
VERBAL ASK
23:29link
“Uh, anyway, thank you guys for watching this video.”
Soft close only — no explicit subscribe ask; the pitch is implicitly the linked skills site and Paper itself. Includes an unprompted non-sponsorship disclosure.
A 20-minute breakdown of the $60B SpaceX–Cursor deal, a hands-on tour of the platform, and a one-prompt playbook for migrating your entire Codex or Claude skill library.
A 39-minute live build: one AI agent cloned from a template, one built from scratch, both running 24/7 in Slack and iMessage using the GRASP framework.
A 22-minute weekly roundup covering the open-source model that finally passed the vibe check, Codex screen-to-skill recording, the SpaceX/Cursor acquisition, and two Claude updates.
A 33-minute field guide to the nine structural shifts in AI agents — what's permanent, what's changing, and the five human foundations that will matter most.
00:20
00:34
00:52
01:15
01:32
01:43
02:06
02:23
02:40
03:14
03:31
03:48
04:04
04:41
04:59
05:18
05:36
05:54
06:13
06:31
06:49
07:08
07:23
07:45
08:12
08:33
08:58
09:16
09:41
09:50
10:11
10:24
10:48
11:08
11:26
11:44
12:03
12:21
12:59
13:17
13:36
13:54
14:13
14:32
14:50
15:09
15:27
15:46
16:05
16:23
16:34
17:00
17:36
17:54
18:12
18:30
18:48
19:06
19:24
19:42
19:52
20:14
20:30
20:51
21:30
21:48
22:02
22:29
22:32
22:52
23:07
Video of the Day20:38
21:08
39:16
22:18
33:38
10:58