Continuous Learning v2
Instinct-based learning with confidence scoring.
Instinct Structure
{
"id": "compose-state-hoisting",
"type": "pattern",
"description": "Always hoist state to caller in Composables",
"confidence": 0.85,
"examples": [...],
"context": "jetpack-compose",
"lastUsed": "2026-02-02"
}
Confidence Scoring
| Score | Meaning |
|---|---|
| 0.0-0.3 | Experimental |
| 0.3-0.6 | Validated |
| 0.6-0.8 | Established |
| 0.8-1.0 | Best practice |
Confidence increases with:
- Successful application
- User acceptance
- Consistency across sessions
Commands
/instinct-status # View with confidence
/instinct-import <file> # Import from others
/instinct-export # Export for sharing
/evolve # Cluster related instincts
Mobile-Specific Instincts
Pre-configured patterns for:
- Compose recomposition optimization
- MVI state management
- Koin module organization
- Ktor error handling
- Espresso test patterns
Remember: Instincts evolve. Low confidence patterns may become best practices.