哈佛研究称顶级模型停用 19 天未影响开发者产出

原标题:New Harvard paper finds that when Claude Fable 5 was suspended for 19 days, its users' pull requests...
72AI 研判分

哈佛论文利用 2026 年 6 月 Claude Fable 5 因美国出口管制被停用 19 天的自然实验,对比 297 名受影响用户与 8,908 名持续使用 Opus 4.8 的开发者,发现 7 天 PR 合并率仅提升 1.4 个百分点,处于噪声区间。在另外 5 次模型升级中合并率同样未见上升,且此前已有 94.5% 的 PR 被合并,说明前沿模型能力提升对成熟开发流程的边际产出可能远低于预期。

报道正文

Rohan Paul (@rohanpaul_ai) · 2026-11-22 16

New Harvard paper finds that when Claude Fable 5 was suspended for 19 days, its users' pull requests got merged about as often on the weaker Opus 4.8.

A U.S. export control directive cut off Claude Fable 5 for 19 days in June 2026, and most users switched to Opus 4.8. The study tracked GitHub pull requests from 297 Fable 5 users against 8,908 developers who stayed on Opus 4.8.

Access to Fable 5 raised the 7-day merge rate by only 1.4 points, a change within noise. Across 5 other model upgrades, merge rates did not rise either, though 94.5% of pull requests were already merged beforehand.

推文配图