Learned
What 3AM has learned on your estate, how accurate it has been, and the learned fixes waiting for you.
Learned shows what 3AM has learned from your incidents (Learning):
| Part | What it shows |
|---|---|
| Diagnoses judged | How many proven incidents have an outcome or a team answer, and how many were right |
| Learned patterns | Causes 3AM proved by investigating that your outcomes confirmed. Each shows its status (diagnosing, fix pending review, promoted, demoted), how often it was right and wrong, and the incidents it came from |
| How often each cause was right | Per cause: right, wrong, precision |
| Demoted classes | Causes that were wrong too often to propose fixes; they still diagnose |
| Calibration | How 3AM's likelihoods match how often it has really been right here, once 30 diagnoses are judged |
Reviewing a learned fix
A learned pattern diagnoses at once. Its fix waits for you:
- Promote fix: 3AM may propose it, still with System 1's review, the rehearsal, your policy and your approvals.
- Reject fix: it's never proposed; the pattern keeps diagnosing.
- Disable / Enable: stop or resume using the pattern.
- Delete: 3AM forgets it.
Every review is in the audit log (learning.reviewed) with your name.
If the page is empty, nothing has been learned yet. Record what your team did on each incident: verified fixes count automatically, and your answers do the rest.