Google tests two screening policies to cut noise from agentic automated bug repair
burkov · x · 2026-11-01
Burkov highlights an article on agentic automated program repair: such systems can diagnose and patch code across large repos, but patches still need human review — and attempts on unsolvable bugs or plausible-but-wrong fixes waste reviewer time and erode trust.
The authors built and tested two screening policies with an internal repair agent and three bug sets from Google's codebase:
- Pre-repair screening: an LLM estimates, from the human-written bug report, whether the repair agent is likely to succeed; bugs below a threshold are skipped.
- Post-repair screening: build and test checks combined with LLM review decide which patches reach human review.
A practical study on filtering low-value agent patches before they hit developers.
More from coding & agent
- Fully vibe-coded browser CS 1.6 clone adds crates and skins, supports 16 players — TAbrodi · 2026-11-01
- MyCaisse ships a remote MCP server: agents read/write freely, money moves only after human approval — Glittering-Cup9298 · 2026-11-01
- Simo Guard: a security layer that vet AI agent tool calls before execution — CIRRUS_IPFS · 2026-11-01
- uivoid lands in Claude's directory: spin up hosted MCP servers by chatting with Claude — Corridl · 2026-11-01
- Gemini 4 Argon codes a Three.js fight scene in a standalone HTML file after 3 hours — Ryoiki-Tokuiten · 2026-11-01
- Vibe-coded order flow trading terminal runs entirely in your browser — mdancho84 · 2026-11-01