Learning roadmap
Prompt vs RAG vs Fine-tune
The decision matrix. When each one wins, the cost-quality-latency triangle, the no-regret defaults.
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:
simple default, feasibility test, 80 20 rule, cost of complexity
Practice quizzesLab taskTake 5 features your team wants. For each, write the smallest prompt that ships 80% of the value.
Build it in the LabCheckpointYou have a written rule for "when prompting is good enough". You enforce it for a week.
What you'll be able to do
- Pick the right approach for any new AI feature in <60 seconds.
- Defend the decision with a rough cost + quality estimate.
- Migrate from prompt → RAG → fine-tune as the data and the budget allow.
Or try a related path