autonomous-looplisted

# Autonomous Loop A harness for running long, unattended, multi-goal work — the kind where a human isn't approving each step. It is a direct application of Addy Osmani's **loop engineering** (`https://addyosmani.com/blog/loop-engineering/`). The four ideas that make it work: > "You design the system that does it instead." — don't do the task by hand; build the > machine that does the task, then let it run. > > "The model forgets, the repo doesn't." — externalize all state to files on disk, so any > fresh session (or a cron run) can resume with zero lost context. > > "Split the one who writes from the one who checks." — the agent that implements a goal > never grades it. A separate agent verifies. Maker ≠ checker. > > "Verification is still on you." — autonomy multiplies output, including mistakes. The > guardrails below are what keep it from confidently amplifying drift. ## When this applies Use it when the work is **large, decomposable, and verifiable**: a backlog of features, a migration, a hardening sweep, a spec that expands into dozens of goals, or a standing "keep improving this" mandate. The payoff is that you keep working across many iterations without the user re-prompting, and the work survives your own context being reset. Do **not** reach for it on a single quick edit, a one-shot question, or anything a human wants to steer turn-by-turn. The loop's overhead (spine files, reviewer panels) only pays off across many goals. ## The shape of the system Everything