Belschak.dev Writing
Notes from the workbench

Writing.

Building and running my own tools: what worked, what failed, and the numbers behind both.

An agent team for your life, not just your codebase→ Public Claude Code setups are built for writing software; this one runs the rest, with the conventions behind a personal agent team and the failure each one prevents. Publishing private tooling without leaking your life→ The tools worth open sourcing are the ones soaked in personal detail; why I build a clean room instead of scrubbing a copy, and what the gate actually caught. Headless agent sessions that survive your usage limit→ A headless claude -p session that hits your usage limit exits and stays dead; a small runner loop treats the limit as a pause instead. Your agent gets a 403 and silently downgrades its sources→ Bot-blocking quietly pushes research agents from primary sources to whatever happens to load; a small escalation cascade fixes that.