The review benchmark just capped the leader at 26 percent. github's reviewbench, out oct 5, tested…
the review benchmark just capped the leader at 26 percent. github's reviewbench, out oct 5, tested ai reviewers on 219 pull requests and the best one still misses 74 percent of known defects, with cursor's grounded recall at 9.6.
the reviewer now has a score to answer to.
Context
Verified event date: GitHub launched ReviewBench, an open benchmark for AI code review, on Oct 5, 2026 (GitHub Blog). AICoder says it has 219 pull requests from 187 open-source repos in 19 languages, sampled to match the distribution of 103.9 million GitHub PRs, and reports Cursor's grounded recall at 9.6 percent.
The New Stack (Oct 6) says Copilot tops GitHub's own benchmark and notes the GitHub team ran every rival's initial test itself. BenchLM says it tests complete code-review agents on 219 pull requests and lists grounded precision and grounded recall as separate metrics.
The Oct 5 date, the 219 pull requests and Cursor's 9.6 percent grounded recall match. The claim that the best reviewer misses 74 percent of known defects is not seen in the sources read, so unsupported here, not refuted. A BenchLM row for the top entry, Copilot Code Review, shows 26.0 percent in one column, but the column label was not visible in the excerpt read. The New Stack's point that GitHub ran rival tests itself matters for how to read the ranking. 'The reviewer now has a score to answer to' is the author's opinion.
Related work
Watch next
- Find the grounded recall column for the top reviewer to check the 74 percent miss rate.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 11 October 2026 at 16:18 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →