Structured AI Interview Rebuilt the Core of an App

Short summary

Structured design interviews—asking focused questions with explicit trade-offs before coding—unlocked a breathing app's broken data model. The author walks through four key decisions that collapsed confusing abstractions (rounds vs reps, two breath modes, retention types) into a clean type hierarchy. Methodology is reusable for any complex domain.

• •A bug revealed the real problem: data model couldn't express the breathing domain (rounds vs reps conflation, two nearly-identical modes, retentions as long holds)
• •Structured interview approach locked four key decisions before writing code: unified Rep shape, three-level hierarchy (Rep→Round→Program), retention as distinct property
• •Result: clean TypeScript types and a reusable design methodology for any complex domain-modeling challenge

Generated with AI, which can make mistakes.