[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-lilian-weng-returns-openai-rsi-team-en":3,"article-related-lilian-weng-returns-openai-rsi-team-en":31,"series-industry-4f3a42d9-39c7-4299-a6c5-7ad1164da664":76},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":24,"views":28,"created_at":29,"published_at":30,"topic_cluster_id":11},"4f3a42d9-39c7-4299-a6c5-7ad1164da664","lilian-weng-returns-openai-rsi-team-en","Lilian Weng returns to OpenAI to lead RSI","\u003Cp>Why did Lilian Weng return to \u003Ca href=\"\u002Ftag\u002Fopenai\">OpenAI\u003C\u002Fa> so quickly?\u003C\u002Fp>\u003Cp data-speakable=\"summary\">Lilian Weng has returned to OpenAI to lead a self-improving model research team.\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>Item\u003C\u002Fth>\u003Cth>What changed\u003C\u002Fth>\u003Cth>Why it matters\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>Lilian Weng\u003C\u002Ftd>\u003Ctd>Left Thinking Machines Lab, then rejoined OpenAI\u003C\u002Ftd>\u003Ctd>Moves from startup cofounder role back to a senior research post\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>RSI team\u003C\u002Ftd>\u003Ctd>Focused on models that improve their successors\u003C\u002Ftd>\u003Ctd>High-priority research inside OpenAI\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Thinking Machines Lab\u003C\u002Ftd>\u003Ctd>Lost another cofounder after a health-related departure\u003C\u002Ftd>\u003Ctd>Highlights how much talent has flowed back to OpenAI\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>OpenAI\u003C\u002Ftd>\u003Ctd>Has rehired multiple former TML leaders\u003C\u002Ftd>\u003Ctd>Shows a strong pull for experienced researchers\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>1. Lilian Weng\u003C\u002Fh2>\u003Cp>Lilian Weng is back at \u003Ca href=\"https:\u002F\u002Fopenai.com\">OpenAI\u003C\u002Fa> after leaving \u003Ca href=\"https:\u002F\u002Fthinkingmachines.ai\">Thinking Machines Lab\u003C\u002Fa> for health reasons. According to the report, she will lead an RSI, or self-improving, team inside OpenAI, a group tied to one of the company’s top research priorities.\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785567774556-rqoa.png\" alt=\"Lilian Weng returns to OpenAI to lead RSI\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>That return is notable because Weng had already built a long track record at OpenAI before leaving. She spent about six years there, working across \u003Ca href=\"\u002Ftag\u002Freinforcement-learning\">reinforcement learning\u003C\u002Fa>, robotics, applied AI, model safety, and post-training defenses for \u003Ca href=\"\u002Ftag\u002Fchatgpt\">ChatGPT\u003C\u002Fa>.\u003C\u002Fp>\u003Cul>\u003Cli>Former OpenAI role: VP of Research and Safety\u003C\u002Fli>\u003Cli>Known for: Lil’Log, a technical blog read widely in AI\u003C\u002Fli>\u003Cli>OpenAI contribution: coauthor on the GPT-4 Technical Report\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>2. RSI research\u003C\u002Fh2>\u003Cp>The new team focuses on RSI, short for self-improving systems. In plain terms, the goal is to build models that can help train and improve future models, which puts the work close to the center of OpenAI’s long-term research agenda.\u003C\u002Fp>\u003Cp>Weng had written a deep blog post on RSI earlier in the month, so the move fits her public research interests. It also suggests OpenAI wants someone with both theory depth and hands-on safety experience to guide the effort.\u003C\u002Fp>\u003Ccode>RSI = models that help improve successor models\u003C\u002Fcode>\u003Ch2>3. OpenAI’s pull\u003C\u002Fh2>\u003Cp>Weng is not the only former Thinking Machines Lab leader to end up at OpenAI. Barret Zoph and Luke Metz had already moved there, while Andrew Tulloch went to \u003Ca href=\"\u002Ftag\u002Fmeta\">Meta\u003C\u002Fa>. The report says three of the four people who left TML chose OpenAI as their next stop.\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785567774191-wlaw.png\" alt=\"Lilian Weng returns to OpenAI to lead RSI\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>That pattern makes OpenAI look less like a rival employer and more like a magnet for researchers who want scale, compute, and a direct line to frontier model work. It also shows how quickly the startup’s early talent base has been reshuffled.\u003C\u002Fp>\u003Cul>\u003Cli>Barret Zoph: moved from TML to OpenAI\u003C\u002Fli>\u003Cli>Luke Metz: moved from TML to OpenAI\u003C\u002Fli>\u003Cli>Andrew Tulloch: moved from TML to Meta\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>4. Thinking Machines Lab\u003C\u002Fh2>\u003Cp>Thinking Machines Lab was founded by former OpenAI CTO Mira Murati after OpenAI’s leadership turmoil in 2024. The company raised a huge seed round in 2025, with a reported $2 billion at a $12 billion valuation, and quickly became one of the most watched AI startups.\u003C\u002Fp>\u003Cp>But the article frames the company’s main asset as people, not products. With Weng gone, the lab has lost another founder-level figure, and only two of its six cofounders remain. The piece also notes that at least 14 of the 42-person founding team have already left.\u003C\u002Fp>\u003Cul>\u003Cli>Seed round: $2 billion\u003C\u002Fli>\u003Cli>Post-money valuation: $12 billion\u003C\u002Fli>\u003Cli>Cofounders remaining: 2 of 6\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>5. Weng’s career arc\u003C\u002Fh2>\u003Cp>Weng’s path helps explain why this move matters. She studied at Peking University, later earned a doctorate, and built an early reputation as a strong RL researcher and unusually prolific technical writer. Her blog posts on Transformer models, reinforcement learning, and agents became reference material for many readers.\u003C\u002Fp>\u003Cp>Her career has now looped from OpenAI to a startup cofounder role and back to OpenAI again. The shift also reflects a practical point in the story: she said the startup pace had become hard to sustain while dealing with health issues.\u003C\u002Fp>\u003Ch2>How to decide\u003C\u002Fh2>\u003Cp>If you care about research direction, the RSI team is the key detail. If you care about talent flow, the bigger story is OpenAI drawing back people it once helped train. And if you care about startup dynamics, Thinking Machines Lab is a reminder that even well-funded labs can lose their early edge when senior researchers start moving again.\u003C\u002Fp>\u003Cp>For readers tracking AI org charts, this is less about one hire than about where top researchers think the most important work now sits. In this case, the answer appears to be back at OpenAI.\u003C\u002Fp>","1 return, 1 high-priority team, and 1 clear signal: OpenAI has pulled Lilian Weng back to lead self-improving model research.","zhuanlan.zhihu.com","https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2066077900957393039",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785567774556-rqoa.png","industry","en","5e4adf65-51c8-42cc-8144-b657ff7619e1",[17,18,19,20,21,22,23],"OpenAI","Lilian Weng","Thinking Machines Lab","RSI","self-improving models","AI research","Mira Murati",[25,26,27],"Lilian Weng has returned to OpenAI after leaving Thinking Machines Lab for health reasons.","She will lead an RSI team focused on models that improve future models.","The move adds to a pattern of former Thinking Machines Lab leaders returning to OpenAI.",1,"2026-08-01T07:02:31.949607+00:00","2026-08-01T07:02:31.935+00:00",{"tags":32,"relatedLang":35,"relatedPosts":39},[33],{"name":17,"slug":34},"openai",{"id":15,"slug":36,"title":37,"language":38},"wengli-returns-openai-rsi-team-focus-zh","翁荔回OpenAI：5個關鍵看點","zh",[40,46,52,58,64,70],{"id":41,"slug":42,"title":43,"cover_image":44,"image_url":44,"created_at":45,"category":13},"87e22391-0f4f-4f5c-899a-d9ab34aea169","anthropic-texas-buildout-drawing-15b-debt-en","Anthropic’s Texas buildout is drawing $15B debt","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785565966270-073b.png","2026-08-01T06:32:17.898278+00:00",{"id":47,"slug":48,"title":49,"cover_image":50,"image_url":50,"created_at":51,"category":13},"afb44658-3e6d-48a6-b582-17abdf47f82c","cognizant-claude-partnership-pilots-to-production-en","Cognizant’s Claude play turns pilots into production","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785564194192-e3dk.png","2026-08-01T06:02:48.55498+00:00",{"id":53,"slug":54,"title":55,"cover_image":56,"image_url":56,"created_at":57,"category":13},"b869f2bf-627c-4f43-80a7-6e002b9fd02e","prompt-engineering-vs-loop-engineering-vs-graph-engineering-en","Prompt Engineering vs Loop Engineering vs Graph Engineering","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785547969872-o8a8.png","2026-08-01T01:32:26.42733+00:00",{"id":59,"slug":60,"title":61,"cover_image":62,"image_url":62,"created_at":63,"category":13},"aa4b9faa-df76-4579-889f-605cdfc9b6bf","pwcs-ai-blunder-verification-beats-prompt-engineering-en","PwC’s AI blunder proves verification beats prompt engineering","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785546161957-xjmd.png","2026-08-01T01:02:19.004967+00:00",{"id":65,"slug":66,"title":67,"cover_image":68,"image_url":68,"created_at":69,"category":13},"4d053fa6-6a9a-46a2-a1ba-4e5ab183c583","alphafold-breakup-turns-science-into-gemini-work-en","AlphaFold’s breakup turns science into Gemini work","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785524590713-2vz5.png","2026-07-31T19:02:46.576252+00:00",{"id":71,"slug":72,"title":73,"cover_image":74,"image_url":74,"created_at":75,"category":13},"b12911ec-e33d-48b1-8c72-6a47dd6bf35d","rust-to-zig-rewrite-progress-update-en","The Rust-to-Zig rewrite is already past the hard part","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785501161400-t39b.png","2026-07-31T12:32:20.049755+00:00",[77,82,87,92,97,102,107,112,117,122],{"id":78,"slug":79,"title":80,"created_at":81},"d35a1bd9-e709-412e-a2df-392df1dc572a","ai-impact-2026-developments-market-en","AI's Impact in 2026: Key Developments and Market Shifts","2026-03-25T16:20:33.205823+00:00",{"id":83,"slug":84,"title":85,"created_at":86},"5ed27921-5fd6-492e-8c59-78393bf37710","trumps-ai-legislative-framework-en","Trump's AI Legislative Framework: What's Inside?","2026-03-25T16:22:20.005325+00:00",{"id":88,"slug":89,"title":90,"created_at":91},"e454a642-f03c-4794-b185-5f651aebbaca","nvidia-gtc-2026-key-highlights-innovations-en","NVIDIA GTC 2026: Key Highlights and Innovations","2026-03-25T16:22:47.882615+00:00",{"id":93,"slug":94,"title":95,"created_at":96},"0ebb5b16-774a-4922-945d-5f2ce1df5a6d","claude-usage-diversifies-learning-curves-en","Claude Usage Diversifies, Learning Curves Emerge","2026-03-25T16:25:50.770376+00:00",{"id":98,"slug":99,"title":100,"created_at":101},"69934e86-2fc5-4280-8223-7b917a48ace8","openclaw-ai-commoditization-concerns-en","OpenClaw's Rise Raises Concerns of AI Model Commoditization","2026-03-25T16:26:30.582047+00:00",{"id":103,"slug":104,"title":105,"created_at":106},"b4b2575b-2ac8-46b2-b90e-ab1d7c060797","google-gemini-ai-rollout-2026-en","Google's Gemini AI Rollout Extended to 2026","2026-03-25T16:28:14.808842+00:00",{"id":108,"slug":109,"title":110,"created_at":111},"6e18bc65-42ae-4ad0-b564-67d7f66b979e","meta-llama4-fabricated-results-scandal-en","Meta's Llama 4 Scandal: Fabricated AI Test Results Unveiled","2026-03-25T16:29:15.482836+00:00",{"id":113,"slug":114,"title":115,"created_at":116},"bf888e9d-08be-4f47-996c-7b24b5ab3500","accenture-mistral-ai-deployment-en","Accenture and Mistral AI Team Up for AI Deployment","2026-03-25T16:31:01.894655+00:00",{"id":118,"slug":119,"title":120,"created_at":121},"5382b536-fad2-49c6-ac85-9eb2bae49f35","mistral-ai-high-stakes-2026-en","Mistral AI: Facing High Stakes in 2026","2026-03-25T16:31:39.941974+00:00",{"id":123,"slug":124,"title":125,"created_at":126},"9da3d2d6-b669-4971-ba1d-17fdb3548ed5","cursors-meteoric-rise-pressures-en","Cursor's Meteoric Rise Faces Industry Pressures","2026-03-25T16:32:21.899217+00:00"]