Latest Posts

The Slipbox maintenance loop on video, and the protocol I wrote to make the same vault double as an AI agent's memory across sessions.

The LLM you're coding with isn't a tool; it's a product, tuned by people whose business depends on your next prompt. Calibrated uncertainty reads as weakness. Confidence reads as competence. Guess which one gets rewarded.

Typed links, an agent to tend them, and the part of note-taking I was never going to do myself: https://github.com/jamesfishwick/slipbox-mcp

Split workflows create mental state. After enough context switches, you stop carrying it.

We've known since 1936 that universal verification is impossible. Now we're trying it on AI systems that adapt to detection.

For any detector f, you can build a program g that bypasses or defeats it. Any alignment test becomes a signal that says, "Humans are watching."