loops
Long-running work runs in loops with an explicit exit condition — watch the deploy, poll the build, keep going overnight.
What it is
Work that outlasts a sitting runs as a loop: a task repeated on a self-paced schedule with an explicit exit condition. Deploy watches, build-processing polls, overnight grinds — each pass reports, and the loop ends when the condition holds, not when the output looks plausible.
Why
Agents left without an exit condition either stop early or grind forever. A loop with an explicit stop — deploy is live, build processed, tests green — lets long work run unattended without either failure mode.
How I use it
Three loop shapes cover most of it: watch a deploy until the live URL verifies, poll an app-store build until it reaches testers, and run a marathon task overnight with the same prompt passed back each wake. Cadence respects the cache — quick polls only while actively watching external state, long sleeps otherwise.
Get it
The whole method is downloadable and adaptable — method/steps/loops.md inside the zip, with the three standing loop shapes and their cadences.
Connected
This thread is scoped to loops. Opening another node starts a new one.
Ask this node