Your capstone review is a 20-minute window. The reviewer is a working practitioner who's shipped agents into production. They are not here to find every bug. They are here to make sure you understand the system you built.
The arc that works
- Problem (2 min) — what's the task, who owns it today, what's broken about the way it works.
- System (5 min) — show the agent doing its thing on a real input. Don't slide-ware this; run it live.
- Evidence (5 min) — your golden set, recent logs, any wins or failures from real use. The honest version.
- What you'd change next (3 min) — three concrete next steps you'd take if you owned this for another month.
- Q&A (5 min) — let the reviewer push.
What gets you the badge
Practitioners aren't grading polish. They're checking that you can explain why your system works when it works, why it breaks when it breaks, and what the next version would look like. Bring the failures, not just the wins. A learner who shows a 70% solution and knows the 30% gap honestly is more impressive than one who shows a 95% solution and can't say why.
Knowledge check
0/1 answered1. Which of these makes the strongest capstone review?
Discussion
0 commentsBe the first to start the conversation.