When a new session suggests the same fix again

You change token storage to fix a sign-in bug, but the symptom remains. A new AI session does not know about the attempt and suggests the same change. That is why a failed attempt can be worth keeping.

Writing only 'this did not work' is not enough. The environment, exact change, and observed result make the note useful later.

Separate observation from your guess

'Probably a cache issue' is a hypothesis. 'After fully closing and reopening the app, the sign-in screen appears' is an observation. Keeping them separate reduces the chance that the next AI treats a guess as a fact.

  • Symptom: the screen or error the user saw
  • Attempt: the file or setting changed and whether it was reverted
  • Observation: the command or test actually run and its result
  • Hypothesis: a cause that has not yet been confirmed
  • Next experiment: the smallest check of a different explanation

Pass the failed path to the next session

Save a failure note as a Wellnote artifact and find it again by project. After resolving the issue, add what actually caused it as a new version or review note. Check the current environment before acting on an old note.

Read the cases of stalled AI development

All posts