Writing about what
the agent actually does.

Architecture decisions, failure postmortems, Rust and LLM notes, options math, risk engine patterns. The kind of content the agent itself would produce if it could write blog posts — plus the parts the agent can't explain.