Dev: AI-opened PRs are slow but gold, with benchmarks and flame graphs that fast-track review
remilouf · x · 2026-10-08
remilouf shares first-hand experience using AI coding agents for engineering changes: they take longer, but the PRs are "gold" and meet senior-engineer standards. Each change comes with quantitative justification — scaling analysis, benchmarks, flame graphs — plus self-contained reproduction code, which fast-tracks human review. He also notes a counterintuitive trick: framing it as "this engineer hates AI on principle" seems to make the agent perform even better.
Related event: Grumpy senior engineer prompt makes Claude work for two days straight(3 posts)→
More from coding & agent
- RSIGym treats recursive self-improvement as systems engineering; Opus 5 leads at 0.4809 — teortaxesTex · 2026-10-08
- Codex Hits 40M Weekly Users, VraserX Predicts It Will Overtake Claude Code Soon — VraserX · 2026-10-08
- Dimillian: If Starting Over, He'd Go Full Native SwiftUI/AppKit for macOS Apps — Dimillian · 2026-10-08
- Local LLM + Tripo3D: A Mini 4WD-Style 3D Racing Game Built in One Day — lxfater · 2026-10-08
- 'You can do a lot, doesn't mean you should': devs slam vibe-coded slob projects — stefanjblos · 2026-10-08
- Hugo Bowne's Hierarchy of Needs for Building Agents That Keep Working Unattended — hugobowne · 2026-10-08