Learning roadmap
Context Engineering
90% of "AI is bad at X" is "you fed it bad context". Fix the context, not the model.
5 stages~8 h
That leaves 5 stages, ~8 hours of focused work. You can clear it in any order, but the stages compound — try them top-to-bottom your first time.
- You'll pick up:
system prompt, persona, guardrails, output contract
Lab taskA small chat that uses 3 swappable system prompts; show side-by-side how the persona changes responses.
Build it in the LabCheckpointYou have a system prompt that works 100 turns deep without drifting.
What you'll be able to do
- Diagnose a "bad answer" as a context problem before blaming the model.
- Design a system prompt that survives long conversations.
- Wire a working RAG pipeline against your own documents.
Or try a related path