Aman Mittal ⚛️☕️🖖

@thsottiaux It already does!

Codex will take us places

Claude Code loves a good chanllenge.

Amazing 👏

i made a tamagotchi that lives in your notch and reacts to your claude code sessions. it cries when you yell at claude and gets happy when you praise it.

I guess Claude Code can get tired.

My most used Claude Code skills in the last 3 weeks.

@full_stack_life @expo The modificationDate you are seeing in docs.expo.dev/llms-sdk.txt is for the page, not the whole file. If you scroll down, you'll notice that each page has its own modificationDate (that's just metadata we have for each docs page).

@full_stack_life @expo All our LLMs files are generated each time we push a new change to Expo docs. llms-full.txt is does not contain the context of SDK docs. We've a separate one for it in docs.expo.dev/llms-sdk.txt We've a dedicated page for all file variants: docs.expo.dev/llms/

@expo do you have updates SDK55 docs available for instant download? The llms.txt on your site is nearly 1 years old. docs.expo.dev/llms-full.txt

@studio_hungry Like, are you handling this via the Schema.org spec?

@amanhimself What are you wondering about?

We wrote an agent-first content spec for our docs. Agent-first needs to work for two audiences: For humans (which is mostly human to AI): - AI chat in docs - Copy as markdown - .md on any page - Consistent, scannable page structure (also applicable to agents) For agents: - llms.txt + sitemap.md (md representation of the nav tree with md links) - Content negotiation through Accept: text/markdown (this is the one we should standerdize on) - markdown metadata (title/desc/version/last_updated,/related/breadcrumbs etc..) - json-ld - SKILLS.md (how to actually use the product) - framework docs bundled in npm packages (A skill can point your agent here. Its versions docs at source) We've had most of the spec on Vercel, Next.js, AI SDK, and Turborepo docs sites for a while now. Now we are rolling it out across all our content sites ~18 Some baselines from the past month: - 584K AI requests (308K agents + 276K crawlers) - 399K humans The TLDR is over half our docs readers are robots now.

@HeyItzaMi Nice bot naming! Might be appropriate :D

got my agents talking between them but with an important note: communication is done just one way what does this mean? Murderbot can interact directly with Mark but the inverse doesn’t work Mark writes requests in a note and Murderbot checks periodically to see if something is there and processes the request/message according to restrictions not an entirely safe system but better than giving Mark access to Murderbot

Update: SkillsBar now picks up user-created agents for Claude Code, not just skills.

@NicolasZu @openclaw The other day, I was just wondering if I could run OpenClaw on my Steam Deck since it has been idle for months. This is inspiring lol

Many people bought Mac Minis to host their @openclaw But I have been rocking mine on Steam Deck and it’s just perfect: powerful + linux + you can give him a face! Cute, no? 🦞 Of course my OpenClaw created his Steam deck face himself ;)

@tchayen Terminals, terminals everywhere :D

Apparently I was lazy for long enough to never have to create personal dotfiles repo and now I can tell my computer to do it and then tell it through my phone to push missing files so my other computer can pull it and apply it. Future is weird

@joreilly Nice! Quite thoughtful on 60% rule in the shared optimization tips section. Never thought about it until I read this and was like of course now it makes sense :D

Best Local LLMs to Run On Every Apple Silicon Mac in 2026 apxml.com/posts/best-loc…

When reading the Expo docs: You get HTML LLMs get Markdown Fit whole docs in your context window, easily.

@BenjaminKomen @expo Thanks Benjamin! Your PR contributions are always a delight to receive! I'll check it soon :)

@expo UI is getting pretty good, you can create really nice SwiftUI components that look exactly like e.g. the iOS Settings app. I played around in this repository (github.com/benjaminkomen/…) and created a PR to improve the documentation (github.com/expo/expo/pull…) cc @amanhimself

I also added Usage stats just to see which Claude Code skills I am using most.

And it works for Codex skills too.

Needed a way to quickly browse my Claude Code skills without digging through Claude Skills every time. So I asked Claude Code to build me a macOS menu bar app for it. It did. In one session (with some babysitting). 🔗 github.com/amandeepmittal…