The "Loop of Loops": A Better Mental Model for AI Agents

Key Takeaways

• Three levels of automation. A prompt is one request; a loop is a recurring job with memory (the school-trip packing list, every month); a loop of loops is several loops that notice each other and hand off context — the packing, weather, schedule, and calendar loops all waking together.
• The value lives between apps. Email holds the confirmation, the school portal holds the pickup change, the grocery app holds the list — but the loop lives in the wiring between them, which has been on you by hand for 20 years.
• Good loops organize attention. The point isn't AI that prompts itself; it's small remembered workflows that run cleanly in the background and only wake you when your judgment actually matters.
• Design each loop around four questions. What can it do safely? What should it ask you? What record does it leave behind? How does it get smarter next time? — plus "what other loop needs to know about this?"
• Start low-stakes. Pick something tedious but forgiving for your first loop of loops (explicitly not banking) — e.g. drafting product use-cases → Linear tickets → PRDs — so that if it goes off the rails you can chuckle about it.