Final Exam

1. What are the three most common failure modes in agent architectures?

2. What is the hardest design decision in multi-agent architecture?

3. What is Cost Per Successful Completion (CPSC) and why is it better than raw cost?

4. What structural mitigation defends against indirect prompt injection?

5. In Google DeepMind's WebAgent (2023) evaluation, what gap did stratified analysis reveal?

6. What distinguishes a research agent from a RAG-based search system?

7. On SWE-bench Verified (2024), what did the subgoal data reveal that the headline number missed?

8. When should the peer-to-peer (debate/critique) pattern be used?

9. In the e-commerce production incident documented by LangChain, what happened when an agent entered a reasoning loop?

10. How should teams test for objective misalignment before deployment?

11. What end-to-end reliability does a 5-step agent pipeline achieve if each step is 95% reliable?

12. Why might adding agents make a system worse rather than better?

13. What is the "consistency problem" with shared memory in multi-agent systems?

14. How did Insilico Medicine use agent pipelines in drug discovery?

15. What problem did Adept AI's ACT-1 encounter with its dual perception channels?

16. What is the primary failure mode where ReAct agents get stuck?

17. What is the key design principle for sub-agents?

18. For complex multi-step workflows using Claude 3.7 Sonnet's extended thinking, what percentage of total token cost can reasoning overhead represent?

19. How does the "Can you draw a complete flowchart before it runs?" question help determine if an agent is needed?

20. When does Plan-and-Execute fail as an architecture?