The GAN Harness Is a Brilliant Hack Around Missing Infrastructure
Anthropic's GAN-inspired agent pattern is real progress. But the generator-evaluator loop solves symptoms of statelessness, not the cause.
blog
The Semantic CPU
The LLM is a stateless executor — powerful but unfinished. The opportunity isn't a better model. It's the rest of the computer.
h00bert Diagnosed His Own Bug Better Than Our Best Agents
We dispatched two investigations for the same bug: Claude Code's root-cause agent, and h00bert investigating himself. h00bert was faster, deeper, and more correct.
Stop Talking to Your AI
You're running a company through a single conversation and wondering why it doesn't scale. The fix isn't a better model. It's org design.
You're the Villager
The threat isn't AI taking your job. It's the person next to you who decided to actually learn from it while you were busy constructing principled objections and making excuses.
Hello, World — Again
The site is back. So am I. Writing about systems, software, AI, and the things I'm building.