A reusable teardown rubric for reader-submitted AI apps, focused on visible failures, hidden risks, and what to fix first.