04. Week 9 — Daily Recall¶
Use this file after reading the explainer.
Answer aloud.
Then verify with 02_explainer.md and 03_study_material.md.
Monday¶
- Why is one correct calculator call not enough for a multi-step task? (Exp §1.1)
- LLM call vs single tool use vs agent — define each in one sentence. (Study §1)
- What are the five named placeholders from the handyman analogy? (Exp ELI5)
- [W8] When did RAG fail because retrieval was not enough?
Tuesday¶
- ReAct — what are the three core steps? (Exp §2.1)
- Why does the ReAct loop work better than chain-of-thought alone for external-state tasks? (Exp §2.2, §2.4)
- Write the smallest useful agent loop from memory. What are its stop conditions? (Exp §2.3, §4.4)
- [W7] In one sentence, why is exact tool use similar to external memory?
Wednesday¶
- Name 5 tool design rules from memory. (Exp §3.1-§3.6)
- Give one example of a bad tool name and a better replacement. (Exp §3.1)
- What is idempotency, and why does it matter more for write tools? (Exp §3.5)
- Why are schemas necessary even when the model is strong? (Study §3)
Thursday¶
- Infinite loop, wrong tool selection, hallucinated arguments — one guardrail each. (Exp §4.1-§4.3)
- Stop rule vs give-up rule — difference? (Exp §4.4)
- When should a human-in-the-loop gate appear? Give 3 examples. (Exp §4.5)
- [W6] Why do we prefer bounded systems in production?
Friday¶
- Parallel tool calls — when do they reduce latency safely? (Exp §5.1)
- Tool chaining — what is a good boundary between tools? (Exp §5.2)
- State across turns — messages vs structured state vs memory. (Exp §5.4)
- What five basics does module 10 assume from this week? (Exp §5.7)
Saturday¶
- Recreate the failure-fix table headings from memory. (Exp §6.1)
- Name 3 metrics you would track for a production agent. (Exp §6.4)
- Write one retrieval prompt you would actually use in your hands_on_lab. (Exp §5.5)
- Where are agents brittle, and where are they still worth using? (Exp §5.6)
Sunday¶
- Answer all self-check questions from
01_weekly_plan.mdwithout notes. - Explain your hands_on_lab architecture using only the words: think, act, observe, state, stop.
- What would break first if module 10 added MCP tomorrow? (Exp §5.7-§5.8)
- [Bridge] Why does multi-agent orchestration fail if single-agent loops are weak?