Reusable Knowledge
Patterns
Build patterns that keep working across projects. This page grows as patterns emerge and prove themselves.
🌱
Early Days
This page is intentionally sparse. Patterns are added only after they've been used in 2+ projects and proven to work. The goal is signal, not volume.
Emerging Patterns
Patterns that have shown promise but need more validation.
Target Schema Extraction
AIemerging
Use Claude function calling to extract structured data from natural language. Define a target schema, let AI populate it from conversation.
From: guildry
// Define what you want to extract
const targetSchema = {
clientName: "string",
industry: "string",
painPoints: "string[]",
budget: "number | null"
};
// AI extracts from conversation
const extracted = await claude.extract(conversation, targetSchema);Discovery-First Schema
Architectureemerging
Don't lock data models upfront. Start with flexible JSONB columns, let real usage reveal what structure you need. Migrate to typed columns once patterns stabilize.
From: guildry
// Start flexible CREATE TABLE clients ( id UUID PRIMARY KEY, data JSONB NOT NULL -- Store anything ); // Later, extract columns that matter ALTER TABLE clients ADD COLUMN industry TEXT; UPDATE clients SET industry = data->>'industry';
Categories to Fill
Areas where patterns will accumulate as projects progress.
AI Patterns
- Prompt templates
- Function calling schemas
- Context management
- Error handling
Architecture Patterns
- Data modeling
- Auth flows
- API design
- State management
Distribution Patterns
- Launch sequences
- Content templates
- Feedback collection
- Pricing experiments
Development Patterns
- AI collaboration
- Code review
- Testing strategies
- Documentation
Pattern Lifecycle
experimentalUsed once, promising
→emergingUsed 2+ times, still evolving
→provenWorks reliably, stable