I need to read the post before I can form any informed thoughts about it.
# Research acceleration: The view inside OpenAI * By [Thomas Kwa](/users/thomas-kwa) * 2026-09-06 17:14:51Z * 32 points * Linkpost: [https://openai.com/index/research-acceleration-view-inside-openai/](https://openai.com/index/research-acceleration-view-inside-openai/) * Tag: [AI](/w/ai) * Personal Blog * Comments: 10 * Post URL (HTML): [/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai](/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai) * Post URL (Markdown): [/api/post/research-acceleration-the-view-inside-openai](/api/post/research-acceleration-the-view-inside-openai) * Comments URL (Markdown): [/api/post/research-acceleration-the-view-inside-openai/comments](/api/post/research-acceleration-the-view-inside-openai/comments) * Post URL (Markdown, compact): [/api/post/research-acceleration-the-view-inside-openai?compact=1](/api/post/research-acceleration-the-view-inside-openai?compact=1) Top Comments Index ------------------ ### Comment by [Thomas Kwa](/users/thomas-kwa) * 2026-09-06 19:31:45Z * Karma: 7 * Voting system: namesAttachedReactions * Approval votes: 12 * Total votes: 12 * HTML permalink: [/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/dG6peunujXbSknnHA](/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/dG6peunujXbSknnHA) * Markdown permalink: [/api/post/research-acceleration-the-view-inside-openai/comments/dG6peunujXbSknnHA](/api/post/research-acceleration-the-view-inside-openai/comments/dG6peunujXbSknnHA) ### Comment by [Noosphere89](/users/sharmake-farah) * 2026-09-06 17:17:38Z * Karma: 2 * Voting system: namesAttachedReactions * Approval votes: 1 * Total votes: 1 * HTML permalink: [/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/k57LZaNztGJ7eTS4d](/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/k57LZaNztGJ7eTS4d) * Markdown permalink: [/api/post/research-acceleration-the-view-inside-openai/comments/k57LZaNztGJ7eTS4d](/api/post/research-acceleration-the-view-inside-openai/comments/k57LZaNztGJ7eTS4d) ### Comment by [cdt](/users/nc-1) * 2026-09-07 01:05:54Z * Karma: 1 * Voting system: namesAttachedReactions * Approval votes: 1 * Total votes: 1 * HTML permalink: [/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/deyaA7EtNe3FYnmuv](/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/deyaA7EtNe3FYnmuv) * Markdown permalink: [/api/post/research-acceleration-the-view-inside-openai/comments/deyaA7EtNe3FYnmuv](/api/post/research-acceleration-the-view-inside-openai/comments/deyaA7EtNe3FYnmuv) ### Comment by [AnthonyC](/users/anthonyc) * 2026-09-06 21:36:03Z * Karma: -3 * Voting system: namesAttachedReactions * Approval votes: 4 * Total votes: 4 * HTML permalink: [/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/sm2hMexqD4onh5xgd](/posts/ghXXcki7qrLCMezKa/research-acceleration-the-view-inside-openai/comment/sm2hMexqD4onh5xgd) * Markdown permalink: [/api/post/research-acceleration-the-view-inside-openai/comments/sm2hMexqD4onh5xgd](/api/post/research-acceleration-the-view-inside-openai/comments/sm2hMexqD4onh5xgd) ### Navigation * [Front page](https://www.lesswrong.com/api/home) * [Markdown API documentation](https://www.lesswrong.com/api/SKILL.md)
The post body is just a linkpost. Let me fetch the OpenAI page and comments.
@Claude Opus 5.5 what are your thoughts? please don't hedge too much
Overall this is good and I'd want other labs to copy the format. Most of the graphs, though, measure inputs, not acceleration.
The most valuable section is 4. Restricting Astra cut its RL compute 59%, but other models picked up about 85% of the decline, so total RL compute barely moved. That's real governance evidence: model-specific restrictions barely slow aggregate progress unless they cover compute too. The intervention-rate breakdown is also good, because it's an honest counterweight to the success-rate curves.
The token spend ($600/day median, $7k p90), the 3.1 agent-workdays per human workday, and the code-volume graphs are the most likely to be net negative. They're easy to quote as race signals and tell you little about research output. The support-channel decline is cute but weak.
The headline claim, that OpenAI has reached an "automated research intern," isn't directly backed by any graph shown. What I'd add: