Link Worthy

Shows a practical pattern where separate planning, building, and evaluation agents argue with each other to produce more reliable outputs on complex coding tasks.

Breaks down practical harness design decisions (context management, checkpointing, error recovery) that keep agents running reliably over long tasks.

First-person account of building an AI-powered robot butler (voice → LLM → arm actuation) that won a robotics hackathon — concrete end-to-end demo.

Straight-talk commentary on AI displacing knowledge workers — not if, but when and how fast.

Trail of Bits Claude Code skills for security research, vulnerability detection, and audit workflows.

From the creator of ReasonML — a minimal, fast text format exploring what structured prose could look like beyond Markdown.

Conversation with Pixar/DreamWorks animation legend Richard Chuang on AI acceleration in creative work — how iteration speed changes what's possible.

Claude Code can now wrap any CLI tool — Zoom, OBS, Blender — as a pseudo-MCP at a fraction of the cost. Practical demo of the CLI-Anything pattern.

This is a fun one. See your public git commits go exponential.

Recently released by Anthropic, a guide to creating agent skills for Claude Code.

Why git worktrees are the right primitive for running parallel AI agents — each agent gets its own isolated branch without the checkout overhead.

Simple CLI for spinning up and managing git worktrees — pairs well with running multiple AI agents in parallel.

Agent Browser is really nice. I use it for all the screenshots in this app.

At first glance, I gotta say designers, you better get on board with this stuff!

The nicest guy on the web also has skills.

Getting a ton of buzz and it's pretty solid. I don't actually install it -- I use the repo as a reference to build my own skills.

This guy is absolutely killing it, in my opinion, on AI commentary.

I use this for everything. I rarely type anything in a prompt -- always voice. Highly recommend.

Great commentary on the reality of where we're heading -- conceding to the fact that AI is definitely the more capable one within a pair coding environment.