7 INSANE loops you need to try right now
TL;DR
Loops remove humans from the equation: AI agents work autonomously toward goals defined by triggers (manual, scheduled, or action-based) and success criteria that are either verifiable or LLM-judged.
The Loop Library is free and ready to use: Berman launched a searchable collection of loops you can copy directly into tools like Codex or Claude Code using the /goal command.
Sub-50ms page load loop ran for 50 minutes: The agent continuously optimized every page until all loaded under the threshold, demonstrating how loops handle concrete, verifiable goals best.
Overnight loops keep codebases healthy: Scheduled loops can auto-update documentation, fix production errors, and refactor architecture while you sleep.
Two major caveats apply: Loops struggle with open-ended feature building (the AI's direction becomes unpredictable) and burn through tokens quickly, running anywhere from 10 minutes to days.
The Breakdown
Loops let AI coding agents work autonomously toward specified goals by combining a trigger with a measurable outcome, and Matthew Berman demonstrates seven production-ready examples while launching a free library of copy-paste prompts.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.
Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.