ConeheadAI · leftover lesson
SNK3-078 · Curriculum How-to
When a tool loops: remember what failed, fix, re-run once
L1–L3 · Hands On–Verifier (bootstrap — system rebalances) · ~4m
Look for this insight
A useful AI work loop is not more prompt flair — remember what failed, apply one fix, re-run with a human stop. That is tool use with memory, not a platform you must build.
Lesson hook
The hype talks about 'systems that prompt themselves.' At work, you usually just need a smaller habit: note what the tool got wrong, change one instruction or setting, run it again, then stop for a human review. Memory plus a fix plus one re-run beats endless re-prompting. And you still own send and merge.
Do this now
Open your notes. For one AI task that didn't go well, write: (1) what failed, in 8 words or less (2) the one fix you'll try (3) where you'll stop for a human check. You're done once you've re-run it with the fix and noted pass or fail in 12 words or less.
Completion check: writes_8_word_failure_memory
Staff review Sign in