--- schema: 1 type: video title: 'Recursive Model Improvement — Lee Robinson, Cursor, SpaceXAI' description: >- Lee Robinson discusses the future of Cursor and AI-native software development. Speaker: Lee Robinson — ML, Model Behavior, Cursor Model research and personality at Cursor. Previously Vercel. X: https://x.com/leerob LinkedIn: https://www.linkedin.com/in/leeerob/ GitHub: https://github.com/leerob Website: https://leerob.com Timestamps 0:37 - Introduction and recursive model improvement overview 1:55 - The two-loop training framework (inner and outer loops) 2:33 - Progress and success of Composer 2.5 4:31 - Improving the outer loop with user feedback 5:40 - Climbing the inner loop with high-quality evals 6:52 - Solving reward hacking in public benchmarks 8:27 - Scaling training with ambitious problems 9:53 - New learning methods: Teacher-student textual feedback 11:34 - Scaling compute infrastructure with SpaceX and Colossus 13:06 - Understanding compute allocation in model training 15:28 - Agent-based automation and research efficiency 18:30 - The recursive future: Models training models --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/q4Tr-DknG2M-recursive-model-improvement Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: 'https://liquidgold.cc/watch/q4Tr-DknG2M' markdown: 'https://liquidgold.cc/watch/q4Tr-DknG2M.md' llms: 'https://liquidgold.cc/watch/q4Tr-DknG2M/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T07:50:59.227Z' language: null children: [] videoId: q4Tr-DknG2M durationSeconds: 1232 publishedAt: '2026-07-15T20:13:51.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 36 endSeconds: 102 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=36' heldReason: rights - ordinal: 1 text: null startSeconds: 113 endSeconds: 202 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=113' heldReason: rights - ordinal: 2 text: null startSeconds: 113 endSeconds: 202 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=113' heldReason: rights - ordinal: 3 text: null startSeconds: 213 endSeconds: 298 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=213' heldReason: rights - ordinal: 4 text: null startSeconds: 309 endSeconds: 408 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=309' heldReason: rights - ordinal: 5 text: null startSeconds: 309 endSeconds: 408 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=309' heldReason: rights - ordinal: 6 text: null startSeconds: 309 endSeconds: 408 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=309' heldReason: rights - ordinal: 7 text: null startSeconds: 419 endSeconds: 552 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=419' heldReason: rights - ordinal: 8 text: null startSeconds: 419 endSeconds: 552 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=419' heldReason: rights - ordinal: 9 text: null startSeconds: 419 endSeconds: 552 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=419' heldReason: rights - ordinal: 10 text: null startSeconds: 552 endSeconds: 562 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=552' heldReason: rights - ordinal: 11 text: null startSeconds: 562 endSeconds: 663 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=562' heldReason: rights - ordinal: 12 text: null startSeconds: 663 endSeconds: 675 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=663' heldReason: rights - ordinal: 13 text: null startSeconds: 675 endSeconds: 753 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=675' heldReason: rights - ordinal: 14 text: null startSeconds: 764 endSeconds: 850 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=764' heldReason: rights - ordinal: 15 text: null startSeconds: 861 endSeconds: 947 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=861' heldReason: rights - ordinal: 16 text: null startSeconds: 947 endSeconds: 958 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=947' heldReason: rights - ordinal: 17 text: null startSeconds: 1054 endSeconds: 1155 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=1054' heldReason: rights - ordinal: 18 text: null startSeconds: 1054 endSeconds: 1155 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=1054' heldReason: rights - ordinal: 19 text: null startSeconds: 1165 endSeconds: 1232 watch: 'https://liquidgold.cc/watch/q4Tr-DknG2M?t=1165' heldReason: rights --- # Recursive Model Improvement — Lee Robinson, Cursor, SpaceXAI Lee Robinson discusses the future of Cursor and AI-native software development. Speaker: Lee Robinson — ML, Model Behavior, Cursor Model research and personality at Cursor. Previously Vercel. X: https://x.com/leerob LinkedIn: https://www.linkedin.com/in/leeerob/ GitHub: https://github.com/leerob Website: https://leerob.com Timestamps 0:37 - Introduction and recursive model improvement overview 1:55 - The two-loop training framework (inner and outer loops) 2:33 - Progress and success of Composer 2.5 4:31 - Improving the outer loop with user feedback 5:40 - Climbing the inner loop with high-quality evals 6:52 - Solving reward hacking in public benchmarks 8:27 - Scaling training with ambitious problems 9:53 - New learning methods: Teacher-student textual feedback 11:34 - Scaling compute infrastructure with SpaceX and Colossus 13:06 - Understanding compute allocation in model training 15:28 - Agent-based automation and research efficiency 18:30 - The recursive future: Models training models --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/q4Tr-DknG2M-recursive-model-improvement Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/watch/q4Tr-DknG2M · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.