What this case covers
The author used the SPEC method to sort out the MVP of a coupon aggregation mini program, had Codex complete the PRD and core development, and reviewed filing, naming, real-device testing, and feature trade-offs.
Questions it answers
- Why start from the coupon scenario?
- How does SPEC help shrink the MVP?
- Why does AI tend to overbuild the first version?
- What ordering issues are there with filing and naming?
- What did real-device testing reveal?
Our observations
- After development speed improves, product trade-offs become the new bottleneck.
- No matter how complete the first version's backend is, it cannot replace validation of the core need.
- When system capabilities such as copy-and-jump are involved, real-device results are the truth.
From the author
The most important thing about the first version is not feature completeness, but validating the MVP.