The top coding agents have converged
the top coding agents have converged: they solve the same 285 of 500 swe-bench verified problems and fail the same 51, and swapping the scaffold a model runs in moves its score by up to 29.8 points, more than the spread across the top thirty agents. openai says the benchmark no longer gives meaningful signal, and a fresh pro re-test drops every model by 20+ points.
leaderboard order stopped meaning anything and most people haven’t noticed.
Context
The arXiv paper 'Coding Agents Have Converged' audits 254 SWE-bench submissions across four splits without running models, asking whether published leaderboard verdicts support reading small score gaps as an ordering. OpenAI (Feb 23, 2026) says SWE-bench Verified is increasingly contaminated and recommends SWE-bench Pro.
OpenAI (Jul 8, 2026) later estimated that about 30 percent of SWE-bench Pro tasks are broken; BSWEN (Apr 20, 2026) says agent scaffold quality can cause a 22-point swing on identical weights.
The convergence theme and OpenAI's statement that Verified no longer measures frontier coding capability match the sources. The 285 shared solves and 51 shared failures, the 29.8-point scaffold effect and the 20-plus point drop on a Pro re-test were not seen in the sources read, so unsupported here, not refuted. Other sources give a 22-point scaffold swing and 7 to 34 points. OpenAI's July audit also found broken tasks in SWE-bench Pro. 'Most people haven't noticed' is the author's.
Related work
Watch next
- Find the 285 and 51 problem counts and the 29.8-point scaffold figure.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 1 October 2026 at 21:17 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 →