About

Mentat Logs is a working notebook on agentic AI — Claude Code, MCP, sub-agents, prompt engineering, and the tooling built around them. Posts here are technical write-ups: what was tried, what the actual output was, and what it means, not polished takes written after the fact.

Most posts start as research or experiments run for real projects, then get adapted into something worth sharing more broadly. Expect methods, code, and real results — including the ones where the hypothesis didn't hold.