Revision 16 — Recall & Drill (S31+S32)
Revision session for Chat System Design and LLM Serving — KV Cache, Batching, Speculative Decoding — flashcards, blank-page recall, hands-on redo, drills.
What this revision covers
Two sessions:
Structure (90 min)
1. Flashcards (10 min)
Pull out the flashcards you wrote after each session. Do 2 passes: quick recall (front → back), then reverse recall (back → front). Mark the ones you got wrong.
2. Blank-page recall (25 min)
Close all tabs. Open a blank doc or paper. For each session, write down — from memory — the 5-7 most important concepts, in your own words. Include one diagram if the topic is visual.
Then compare to the original session notes. What did you miss? Those gaps are the highest-value learning.
3. Hands-on redo (25 min)
Redo the hands-on drill from each session (in the "Hands-on drill" block of each). This time, aim to do it faster + cleaner, ideally without looking at the original.
4. Drill / mixed problems (20 min)
- Explain both topics to an imaginary interviewer in 3 minutes each. Time yourself.
- One "cross-topic" question that combines both: how would concepts from Chat System Design and LLM Serving — KV Cache, Batching, Speculative Decoding interact in a real system?
5. Gap analysis + notes update (10 min)
- What are the top 3 things you got wrong or forgot?
- Update your flashcards with those items.
- Add anything to your long-term notes / cheat-sheet document.
What's next
Session 33 continues the series. See the hub page for the full sequence.