Decisions and Patterns - How Your AI Learns
Two Types of Learning
Decisions: Explicit choices you make. Patterns: Implicit styles your AI detects.
Both compound your intelligence.
Decisions - What You Choose
When you make a choice, your AI asks:
AI: "You chose Zod over Yup for validation. Should I remember this
for future projects?"
You: "Yes"
That's it. Next time validation comes up, your AI already knows.
Decisions are explicit. You choose. Your AI remembers. You can change your mind anytime.
More decision examples:
- "Use FathymAPI for APIs" → AI applies to future API projects
- "Prefer functional components" → AI writes functional, not class
- "Test with Deno.test, not Jest" → AI uses native testing
- "Name files in kebab-case" → AI follows your convention
Patterns - What Your AI Detects
After working with you across sessions, your AI notices consistency:
AI: "I've noticed you always add error boundaries around async operations.
Should I apply this pattern automatically?"
You: "Yes"
Patterns are implicit. Your AI observes your work, detects what you do consistently, and proposes making it automatic. You approve what sticks.
Pattern detection examples:
- Code style: "You always destructure props in function signatures"
- Architecture: "Your API routes always have input validation, then auth check, then handler"
- Workflow: "After creating a new file, you always add it to the barrel export"
- Quality: "You always add error handling before shipping"
The Difference
| Decisions | Patterns | |
|---|---|---|
| How | You choose explicitly | AI detects from your behavior |
| When | At the moment of choice | After observing consistency |
| Confirmation | "Remember this?" - Yes/No | "Apply automatically?" - Yes/No |
| Override | Change anytime | Always overridable |
| Speed | Immediate | Builds over sessions |
What You Don't Do
No config files to write. No style guides to maintain. No rules to teach your AI manually.
You just build. Your AI learns from how you work.
- Commands → - All commands