failure-loops

name	failure-loops
description	Use this to classify what happens right after a first attempt fails, and which second moves pay off versus spiral.

Failure Loops

The Big Idea

The first attempt failing is noise. The response to it is signal. Most of a person's wasted hours live inside two or three recurring loops they cannot see from inside.

The Second-Move Taxonomy

Classify every post-failure move:

• gather new evidence before retrying
• repeat with small variation (same theory, new dice roll)
• increase prompt specificity at the model
• change architecture mid-flight
• change tool
• change model
• widen scope
• narrow scope
• research externally
• demand a complete rewrite
• revert to last known good
• abandon the task
• press on despite weak evidence

Count each move per task type. Then grade outcomes:

Response	Effect on this corpus
gather evidence	usually shortened or lengthened failure?
switch model	real progress or same loop restarted
rewrite demand	fresh start or lost partial learning

Grade with outcome data, not intuition.

Loop Cards

For every recurring loop write one card:

TRIGGER -> RESPONSE -> SHORT-TERM BENEFIT -> LONG-TERM COST -> INTERRUPTION STRATEGY

The interruption strategy must be observable and cheap: a question to ask, a check to run, a rule to enforce. Rank cards by expected hours recovered.

When It Backfires

• Treating persistence as sunk-cost. Sometimes the third attempt is exactly right. Judge by whether evidence changed between attempts.
• Missing the hidden benefit. Every loop survives because it pays something short-term. Name the payment honestly.

One-Line Memory

The second move decides the session.