Practice specification, AI review and independent evidence

Read the section first. Work in this order; each project isolates one skill. Each link opens one complete build brief with a concrete contract, a baseline, two changed requirements, and the original staged AI prompts.

These are five project briefs, not five supplied applications. All prompts are constructed practice.

1. Specify tag behavior for independent implementers

Specify tag behavior for independent implementers — Pin user-visible behavior and test specification ambiguity.

2. Evaluate generated rate-limiter code with a simple reference

Evaluate generated rate-limiter code with a simple reference — Verify an unfamiliar implementation against an understandable oracle.

3. Build reusable AI instructions and evaluate transfer

Build reusable AI instructions and evaluate transfer — Extract constraints and measure transfer on held-out work.

4. Collect evidence about changed behavior and affected callers

Collect evidence about changed behavior and affected callers — Connect each risk in a diff to a reproducible test or an honest gap.

5. Record and revisit uncertain engineering decisions

Record and revisit uncertain engineering decisions — Turn uncertain acceptance into revisitable decisions with evidence.