Architectural Assets

This section is where “daily learning” turns into reusable building blocks.

Not everything needs to be fancy. The point is to capture what’s reusable.

What will live here

1) Templates (copy/paste)

• PR review checklist for AI changes
• RAG evaluation checklist (offline + online)
• Agent risk review (tool access, escalation, rollback)
• “Before we ship” reliability checklist

2) Reference architectures

• RAG baseline (simple, measurable)
• Agentic workflow (planner/executor/tooling)
• Observability + eval loop (what to log, what to score)

3) Decision shortcuts

• When to use RAG vs fine-tuning
• When to use agents vs workflows
• Where to enforce structured outputs
• How to choose evaluation metrics

How I’ll keep it honest

Anything that lands here must pass one test:

“Would I reuse this in a real system six months from now?”

If yes, it belongs here.