Same prompt, four spec-driven coding tools: 4 minutes to 2 hours

A developer ran a stricter real-world benchmark—same drag-and-drop feature on an identical TypeScript/Express/SQLite/React codebase—across OpenSpec, Spec Kit, BMAD and Kiro, finding completion times ranged from 4 minutes to 2 hours and task counts from 13 to 29.

2026-09-19 ~ 2026-09-20 · 2 related posts