Learning paths
Answer one question: which sentence is you right now?
- “I am slow, and I do not know how to drive the assistant.” → Path 1
- “I need the model inside my own product.” → Path 2
- “The demo works; I cannot ship it or price it.” → Path 3
The three paths stack. Default order is 1 → 2 → 3. Skip to 2 only if you already use a coding agent every day.
This is not a product manual (Products) and not model internals (Learn LLM).
In reading order
1. Use AI tools
Problem: boilerplate by hand, docs by hand, half-day debugging.
After: you can drive Cursor, Copilot, Claude, or Gemini on a real repo.
Next: put a model in your app.
2. Add AI features
Problem: you can use the assistant, but your React / Vue app has no chat, search, or fill.
After: API calls, streaming UI, RAG search.
Need: path 1 habits; TypeScript.
3. Build AI products
Problem: the demo is fine; production times out and the bill explodes.
After: deploy, rate-limit, observe, cost, eval.
Need: at least one working feature from path 2.
How to choose
| You are here | Take |
|---|---|
| Almost never used a coding agent | Path 1 |
| Use one daily; need it in the product | Path 2 |
| Already wired an API / RAG; need to ship | Path 3 |