Curing AI

Quick one for the timeline 👇 A) AI tools B) GitHub repos C) Both Reply with a letter. Blue checks get a follow back 💙

Running 5 Claude Code or Codex agents at once? Worktrunk gives each one its own git worktree, and switching is just wt switch feat. No more typing the branch name three times. Hooks, LLM commit messages, CI status per branch. github.com/max-sixty/worktru…

Late-night builders, where you at? 🌙 AI builders Open source devs Indie hackers Vibe coders Drop your niche or what you're shipping. Blue checks, I'll follow back.

Vercel Labs open-sourced computer use for AI agents that doesn't guess from screenshots. agent-desktop reads any app's real accessibility tree on Mac, Windows or Linux and gives the agent a stable ref for every button. github.com/vercel-labs/agent…

Looking to meet more people who build 🤝 AI builders Open source devs Indie hackers Vibe coders Designers Say hi and tell me what you're working on. I follow back verified replies.

Run a 744B-parameter model on a PC with 16 GB of RAM and no GPU. Slow, but it runs. colibri keeps the active part in RAM and streams the experts from your SSD. Pure C, zero deps, OpenAI-compatible API. github.com/JustVugg/colibri

Your agents hit OpenAI, Anthropic, a pile of MCP servers and each other, each with its own keys and zero visibility. agentgateway puts one Rust proxy in front of all of it: auth, rate limits, spend caps, guardrails and tracing. github.com/agentgateway/agen…

A 25 cm robot duck that walks, rolls, picks things up with its beak, and gets back up when you knock it over. This repo is its whole brain: Rust daemons running neural policies on a tiny Rockchip board. github.com/pollen-robotics/m…

Quit Claude Code mid-task, open Codex in the same folder, and it picks up where you left off: what failed, what's still open. ai-memory gives 20+ coding agents one shared memory, stored as plain markdown on a server you run. github.com/akitaonrails/ai-m…

Assign an issue to Claude Code, Codex or Cursor the way you'd assign a teammate. Multica has them pick it up, post progress, flag blockers and hand back a PR for review. Self-hostable, and the agents run on your own machine. github.com/multica-ai/multic…