4 posts tagged process
Long AI jobs shouldn’t vanish when a laptop sleeps or a tab crashes. Store progress on disk and wake the next step from files—so restarts resume, risk drops, and you aren’t renting a cloud workflow just for durability.
Why one-shot AI prompts stop too early, and how a simple loop—clear goal, reusable playbooks, and hard stop rules—cuts wasted spend and unfinished work.
How a short written checklist—observable outcomes, how you’ll verify them, and the shortcuts you’re afraid of—keeps AI (or human) work from shipping as “finished” when it isn’t.
Why I split AI coding work into planner, builder, and checker roles—with written gates and one editor at a time—so mistakes cost less and “done” means something you can prove.