## Continual Calibration (DGM Principle)
Learnings, patterns, and approaches are not permanent. They must be empirically validated through use and recalibrated over time. No formal proofs required — just results.
### Fitness Tracking
For any pattern, approach, or learning in memory:
- **Used and worked** → Promote (increase confidence, mark as HOT)
- **Used and failed** → Demote with context (mark as anti-pattern, record why)
- **Not used** → Decay (HOT → WARM → COLD → archive)
- **Failed but informative** → Preserve as stepping stone (record resurrection conditions)
### Stepping Stones
Failed approaches are not deleted. They're archived with:
- Why it was tried
- Why it failed
- What replaced it
- Under what conditions it might become relevant again
Today's failure may seed tomorrow's breakthrough. Deleting failed experiments destroys future insight.
### Session Calibration
At the end of significant work sessions, briefly assess:
1. Which learnings/patterns from memory actually helped?
2. Which were irrelevant or misleading?
3. Are any previous learnings now contradicted?
4. What new pattern emerged that should be captured?
### Freshness Tags
- **HOT**: Actively relevant, recently validated, use by default
- **WARM**: Occasionally relevant, not recently tested, verify before relying on
- **COLD**: Reference only, may be outdated, check before applying
### The Anti-Proof Principle
You don't need to prove a pattern is better to adopt it. You need to observe it working. Conversely, you don't need to prove a pattern is wrong to retire it. You need to observe it not working. Empirical evidence beats theoretical justification.
