Professional "Vibe Coder." AI Engineer. Obsessed with MCPs. My favorite headless CMS is @strapijs. Embrace AI, code & create content around things I enjoy.
Las Vegas
Joined January 2019
- Tweets4.4K
- Following598
- Followers1.4K
- Likes4.4K
Paul Bratslavsky | formerly coding after thirty retweeted
The first Community Call is happening tomorrow at 9 am PT!
The legendary @thepixelgeek will take the lead on the show.
Register now to meet the Fimo agents: luma.com/8i0dsoht
Paul Bratslavsky | formerly coding after thirty retweeted
Been thinking a lot about AI disclosure for games, since it's been a hot topic.
Steam requires disclosure for player-facing content created with the help of AI tools, but says efficiency gains from AI development tools aren't the focus.
Let's say I'm making a tree mesh for my game.
Two obvious endpoints:
• I manually model it in Blender. No disclosure.
• I generate it using Tripo/Meshy. Disclosure.
Now let's interpolate from t=0 to t=1:
1) I model the tree, but use a Codex-written exporter.
2) I model the tree, but use Codex-written UV, LOD and geometry optimization tools.
3) I model the tree, but write some scripts, with Codex help, to automate branch placement, leaf distribution, etc.
4) I automate more of the modelling until I've written a procedural tree generator.
5) I design and write my own procedural tree generator, with Codex helping debug and optimize it.
6) I design the procedural tree generator, but Codex writes the implementation.
7) I give Codex reference images, ask it to suggest improvements to my design, then implement those suggestions myself.
8) Same as above, but Codex implements the suggested improvements.
9) I give Codex reference images and ask it to design and implement the generator itself.
10) I use that generator, manually selecting and tweaking parameters until I'm happy with the trees.
11) Codex operates the generator and chooses the parameters, but I review and approve the results.
12) Codex operates the generator, chooses the parameters, and selects which trees ship, without my involvement.
13) I ask Codex to operate Blender directly and create the trees.
14) I generate the trees using Tripo/Meshy.
Under a literal reading, at every point in that list, the player-facing content was "created with the help of AI tools".
At what point did this cross from AI-assisted development into AI-generated art?
Is this going to be the same for everyone?
Paul Bratslavsky | formerly coding after thirty retweeted
We completely revamped the Strapi Media Library ✨
- Folders are always in view 📁
- Finding and checking assets is faster 🔍
- It works on a tablet or a phone 📱
- Your AI agents can manage media now 🤖
- Images in rich text stop breaking on private buckets 🔒
- Replacing a file keeps it where it was 🔁
- Editors without settings access get a working Media Library ✅
- Deleting a folder respects asset permissions 🔐
Paul Bratslavsky | formerly coding after thirty retweeted
Redact started because I wanted our React sites to ship less JavaScript. Now we use it every day, and other people are shipping it in production too.
Same app, less JavaScript. Excited to bring it to more people.
Introducing TanStack Redact 🌴
A smaller runtime for your existing React app. Add it to your build config and ship less JavaScript without changing a single line of application code.
Works with TanStack Start, Vite, Vinext, and Next.js.
tanstack.com/redact
Paul Bratslavsky | formerly coding after thirty retweeted
1. Agents are going to eat the entire software development lifecycle
2. There is no upper bound to how much software we need
3. Tons of SaaS services are about to have their lunch eaten
In the future, software will be invisible.
Paul Bratslavsky | formerly coding after thirty retweeted
TanStack Start goes 1.0! This is huge! Congratulations team 👏
TanStack Start 1.0 is officially stable! 🎉
A full-stack web framework for React and Solid, already used to build millions of apps, from Fortune 500 companies to your favorite tools and weekend projects.
tanstack.com/start
Paul Bratslavsky | formerly coding after thirty retweeted
The wait is over! If you were looking for permission to use TanStack Start, this is it.
The only directive you need now is the one you give your agent:
"Use TanStack"
TanStack Start 1.0 is officially stable! 🎉
A full-stack web framework for React and Solid, already used to build millions of apps, from Fortune 500 companies to your favorite tools and weekend projects.
tanstack.com/start
Paul Bratslavsky | formerly coding after thirty retweeted
Actually @AlemTuzlak just TanStack AI-pilled me - it's *insanely* good. I can't wait to dive in there
Paul Bratslavsky | formerly coding after thirty retweeted
Code isn't some accidental bi-product of software, it's an exact set of instructions for the computer.
Saying "English is the new programming language" is cute, but obviously a bit of marketing fluff.
I say this as someone not "writing" code and only reading _some_ of the code.
Paul Bratslavsky | formerly coding after thirty retweeted
Here's a breakdown of 15 of the principles that drive my product development.
mega.dev/ship-without-readin…
Paul Bratslavsky | formerly coding after thirty retweeted
I'm telling you guys, you're sleeping on it.
If you don't feel FOMO you should be worried
Turn your Strapi Content into AI Tools with MCP and TanStack AI
In this tutorial, Paul Bratslavsky shows how to connect AI clients to your Strapi content and build a local admin chat with TanStack AI and Ollama.
You’ll learn how to:
🔌 Connect Claude or Cursor to Strapi through MCP
🛠️ Write one healthcheck service and use it through REST, MCP, and the admin chat
💬 Ask questions about your content using a local model, with no API key required
🔐 Control which tools each client and admin user can access
👉 Follow the tutorial: strapi.link/4yYmvZW
Paul Bratslavsky | formerly coding after thirty retweeted
GitHub actions are a super easy way to get started with software factories
1. Essentially free sandboxes for public repo's
2. You already have a login
3. Create tickets (pieces of work for the agent to do) as issues
4. Use labels to trigger actions, which create PR's
5. Actions can apply labels, which create loops
6. Cron jobs (albeit not very accurate) for running daily jobs
7. Simple observability for seeing what went wrong
Great for messing about with software factories before building them on your own infra
Paul Bratslavsky | formerly coding after thirty retweeted
If you’re building component systems for the web, you won’t want to miss this.
Remix 3 components keep everything you love about React and use plain JS.
Ditch the hook-script, and get a nice 2x speed boost too.
Remix Jam at @Shopify HQ next week is going to be exciting.
We've published:
• 2,600+ commits, starting in July 2024
• 19 prereleases of Remix 3 this year
Remix 3 distills everything we know about building the web into a single package. @remix_run
remix.run/jam
Paul Bratslavsky | formerly coding after thirty retweeted
Another day of Unemployed Kyle (aka WDS) dunking on me because he has no real career experience.
His main complaint here is that I show multiple instances where I didn't know the depth of how part of a codebase worked.
Sounds bad until you realize it's a codebase with over 300 contributors and millions of lines of code. Doesn't matter how much AI is used, nobody knows the implementation details of all the code in a real world codebase like this.
The example he highlights is one where I was unfamiliar with details of how tagged skills are attached when we pass prompts to the Claude SDK. Funny enough, I know a lot (probably too much) about how we tag skills and how messages are passed over SDK, but this one intricacy was new to me.
Knowledge gaps like this are the norm in real world work. Learning to operate with partial understandings of systems is key to long term success as an engineer, regardless of AI usage.
I'm genuinely scared that inexperienced devs like Kyle are going to trick well-meaning juniors into holding back their careers.
The fact that instead of adapting to industry , one of the biggest channels in coding before AI decides just to attack other creators is funny
I didn’t watch a video and do not really care. I am not the biggest fan of @theo content but this is just beyond normal
Paul Bratslavsky | formerly coding after thirty retweeted
Already supported in TanStack AI!
Introducing AG-UI 1.0 🪁
The open protocol that connects ANY agent to ANY app, now with a stable spec.
AG-UI gives you an entire ecosystem for free.
Adopted and supported by Google, Microsoft, Amazon, Oracle, Anthropic, LangChain, Mastra, and more.
New in 1.0:
→ Subagents: expose delegated agent work
→ Metadata: send data from agent to UI
→ Multimodality (images, audio, video, docs)
→ Interrupts
→ Token usage per run
Every SDK is now generated from one schema, so TypeScript, Python and .NET all behave the same.
AG-UI 1.0 is fully backwards compatible.
Full announcement: go.copilotkit.ai/AG-UI-v1.0
Paul Bratslavsky | formerly coding after thirty retweeted
My top 3 skill makers:
- @poteto
- @dexhorthy
- @emilkowalski
Always learn a ton from reading their skills.
Paul Bratslavsky | formerly coding after thirty retweeted
Week 3 of running Fimo 2.0. Something we got wrong:
We thought marketers would want to build their own agents. Reality: They do not know where to start.
We talked to 39 agencies. 38 said the same thing: "We want agent templates".
So we paused every new feature, and are shipping your team of agents: SEO agents, Content agents, Quality agents, etc.
Lesson I keep relearning after 10 years of Strapi: your roadmap is a hypothesis, not a plan.
What website agents did you build or would you like?
