Your AI Patch Is Cheap to Write and Expensive to Review: A 90-Minute Review-Cost Ledger

Description and use

Problem: AI patches reduce coding effort but raise review time. Solution: a 90‑minute spike using review_ledger.py to measure review minutes and compare AI vs. human patches.

Problem

AI-generated patches reduce coding effort but increase human review time, making software still expensive.

Outcome

The spike produced a lightweight, stdlib‑only tool that measures whether AI patches increase review time beyond a 50% median increase compared to hand‑written patches.

Workflow

• Write hypothesis and kill rule (0–10 min)
• Build ledger script review_ledger.py (10–25 min)
• Run five fixed tasks and generate patches (25–60 min)
• Time each review and record minutes (60–80 min)
• Compute median review minutes, compare ratio, and produce verdict (80–90 min)

Evidence

Agent patches that pass CI cost no more human review minutes than hand‑written patches touching the same files.

Source

Three numbers, not a dashboard

Source

The spike is not about which model is best. It is about whether your review loop absorbs the output at a sane price.

Source