HN Showcase
Back to browse
GitHub Repository

What % of your AI engineering effort went to fixing the AI's own work? One-page rework report from git history. Read-only, local-first.

1 starsPython

I audited 162 agent-written PRs – 27% were the AI fixing itself

by aimattb·Jun 13, 2026·3 points·1 comment
Visit ProjectView on HN

AI Analysis

●●SolidBig BrainSolve My Problem

Measures AI rework tax from git history when nobody else tracks self-correction rates.

Strengths
  • •Transparent classification cascade with explicit signals and thresholds in one config block
  • •Local-first architecture with read-only API access and no telemetry leaving your machine
  • •Real data from 162 PRs showing 27% were AI fixing its own work
Weaknesses
  • •Heuristic classification means squash merges blur attribution accuracy
  • •Narrow audience limited to teams already heavily using AI coding agents
Target Audience

Engineering teams using AI coding agents heavily

Similar To

Cursor Analytics · Sourcegraph Cody · GitHub Copilot metrics

Similar Projects