I run Claude Code, Cursor, and Codex most of the day, and I keep flip-flopping between two setups that both feel wrong.
Setup 1, babysit: I watch every command and approve everything myself. Nothing slips past me, but I'm stuck at the terminal and the whole point of running an agent is gone. I might as well be typing it myself.
Project Telos is scheduled to launch here on June 26. I am building it around one rule: if AI work matters, the person and the system should be looking at the same checkable state, not trusting the model's self-report.
The public line is five flagships:
- gather: witnessed intake and provenance receipts
- index: rerunnable workspace maps and MATCH / DRIFT / UNVERIFIABLE certificates
