Watch · Arun | Engineering Judgment
Spec-Driven AI Coding Has a Blind Spot
Every unit test passes. The feature still breaks. Here's the one extra task that catches it.
YouTubeOpen the channel · Open on YouTube
Related on reasonfrom.com
- New videoOKF explained: the Open Knowledge Format for AI agents
- New videoMaking OKF Production Ready: Tools, Linting & Sync | Part 2
- New videoWhy AI now thinks before it answers?
- New videoRAG from first principles: chunking to evals
- New lessonDay 21 — Fewer Complaints, More Mistakes
- New lessonDay 20 — Nine Fixes, Still 40 a Week
- New lessonDay 19 — 86% Resolved, Nothing Fixed
- New lessonDay 18 — Three Changes, One Test
- New lessonDay 17 — The Message That Opened the Door
- New lessonDay 16 — A Perfect Score for the Broken Version
- New lessonDay 15 — Up 3 Points, 6× Worse
- New lessonDay 14 — 48,212 Apologies
- New lessonDay 13 — Approved in Four Seconds
- New lessonDay 12 — Same Source Twice
YouTube