What this case covers
Ye Xiandi and the team used Codex to develop the "No Wasted Learning" teaching agent, turning learner sticking points into concrete actions and responsibility assignments. The article reviews prototype design, evidence constraints, and team collaboration; the team spent 9 hours building during the competition, and the public static deployment was completed after the competition.
Questions it answers
- Why is a teaching dashboard insufficient for resolving sticking points?
- How do you give a learner a single executable next step?
- How should data without evidence be displayed?
- How should AI drafting and teacher confirmation divide the work?
- Why did the product have complete features but fail to articulate its value?
Our observations
- The 9 hours relied on industry accumulation and team coordination; it is not a general delivery commitment.
- Whether the learner continues to act is closer to the delivery outcome than whether the ticket was replied to.
From the author
The person in charge saying it is solved does not count; you have to go back and confirm whether that learner actually moved forward.