--- schema: 1 type: video title: 'Robot Demos Are Easy. Reliability Is Hard — Jason Ma, Dyna Robotics' description: >- A generalist robot that succeeds 80 to 90% of the time makes a great video but a poor product. Jason Ma, co-founder and CTO of Dyna Robotics, shows how Dyna got its model, Dyna-1, to a 99.4% success rate folding restaurant napkins for 24 hours straight, including recovering after it pulled over the whole stack. The key is a reward model that watches the robot and scores its progress. When the score dips, the robot is making a mistake, so the team can collect targeted recovery data and fine-tune again in a human-in-the-loop active-learning cycle. Ma also covers Dyna's pre-training data pyramid of more than 200,000 hours, and its architecture that pairs a reasoning model with a world action model. He shows deployments in restaurants and a Sacramento laundromat, and a robot folding T-shirts for three days straight at CoRL in Korea with no site-specific data. Speaker info: X/Twitter: @JasonMa2020 (https://x.com/JasonMa2020) Website: https://jasonma2016.github.io/ Related links: Dyna Robotics: https://www.dyna.co Timestamps: 0:00 Intro 0:32 About Dyna Robotics 1:12 The research and deployment flywheel 2:16 How robot foundation models work 3:11 The pre-training data pyramid 4:21 Reasoning model plus world action model 5:01 New tasks with under an hour of data 6:10 Why 80–90% success isn't enough 7:40 Case study: restaurant napkin folding 9:00 Dyna-1: 99.4% success over 24 hours 9:59 What makes napkin folding hard 11:04 Why standard post-training stalls at 80% 11:39 Reward models that score robot progress 13:04 Scalable supervision and active learning 14:08 Error recovery highlights 15:43 Real deployments: restaurants and a laundromat 16:28 Working at new sites with no new data 17:13 Three days of T-shirt folding at CoRL 18:13 Opening Red Bull cans at live events 18:58 Summary 19:52 Q&A --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/Sjfz1TqxzEs-youtube-9096b441d4cb2f01ced2 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/Sjfz1TqxzEs' markdown: 'https://liquidgold.cc/watch/Sjfz1TqxzEs.md' llms: 'https://liquidgold.cc/watch/Sjfz1TqxzEs/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-23T22:20:52.307Z' language: null children: [] videoId: Sjfz1TqxzEs durationSeconds: 1602 publishedAt: '2026-09-24T14:30:12.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 46 endSeconds: 78 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=46' heldReason: rights - ordinal: 1 text: null startSeconds: 89 endSeconds: 144 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=89' heldReason: rights - ordinal: 2 text: null startSeconds: 155 endSeconds: 200 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=155' heldReason: rights - ordinal: 3 text: null startSeconds: 211 endSeconds: 265 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=211' heldReason: rights - ordinal: 4 text: null startSeconds: 265 endSeconds: 276 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=265' heldReason: rights - ordinal: 5 text: null startSeconds: 320 endSeconds: 374 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=320' heldReason: rights - ordinal: 6 text: null startSeconds: 385 endSeconds: 463 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=385' heldReason: rights - ordinal: 7 text: null startSeconds: 542 endSeconds: 552 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=542' heldReason: rights - ordinal: 8 text: null startSeconds: 620 endSeconds: 674 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=620' heldReason: rights - ordinal: 9 text: null startSeconds: 685 endSeconds: 711 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=685' heldReason: rights - ordinal: 10 text: null startSeconds: 797 endSeconds: 852 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=797' heldReason: rights - ordinal: 11 text: null startSeconds: 864 endSeconds: 951 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=864' heldReason: rights - ordinal: 12 text: null startSeconds: 1009 endSeconds: 1043 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=1009' heldReason: rights - ordinal: 13 text: null startSeconds: 1054 endSeconds: 1098 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=1054' heldReason: rights - ordinal: 14 text: null startSeconds: 1108 endSeconds: 1143 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=1108' heldReason: rights - ordinal: 15 text: null startSeconds: 1369 endSeconds: 1449 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=1369' heldReason: rights - ordinal: 16 text: null startSeconds: 1459 endSeconds: 1602 watch: 'https://liquidgold.cc/watch/Sjfz1TqxzEs?t=1459' heldReason: rights --- # Robot Demos Are Easy. Reliability Is Hard — Jason Ma, Dyna Robotics A generalist robot that succeeds 80 to 90% of the time makes a great video but a poor product. Jason Ma, co-founder and CTO of Dyna Robotics, shows how Dyna got its model, Dyna-1, to a 99.4% success rate folding restaurant napkins for 24 hours straight, including recovering after it pulled over the whole stack. The key is a reward model that watches the robot and scores its progress. When the score dips, the robot is making a mistake, so the team can collect targeted recovery data and fine-tune again in a human-in-the-loop active-learning cycle. Ma also covers Dyna's pre-training data pyramid of more than 200,000 hours, and its architecture that pairs a reasoning model with a world action model. He shows deployments in restaurants and a Sacramento laundromat, and a robot folding T-shirts for three days straight at CoRL in Korea with no site-specific data. Speaker info: X/Twitter: @JasonMa2020 (https://x.com/JasonMa2020) Website: https://jasonma2016.github.io/ Related links: Dyna Robotics: https://www.dyna.co Timestamps: 0:00 Intro 0:32 About Dyna Robotics 1:12 The research and deployment flywheel 2:16 How robot foundation models work 3:11 The pre-training data pyramid 4:21 Reasoning model plus world action model 5:01 New tasks with under an hour of data 6:10 Why 80–90% success isn't enough 7:40 Case study: restaurant napkin folding 9:00 Dyna-1: 99.4% success over 24 hours 9:59 What makes napkin folding hard 11:04 Why standard post-training stalls at 80% 11:39 Reward models that score robot progress 13:04 Scalable supervision and active learning 14:08 Error recovery highlights 15:43 Real deployments: restaurants and a laundromat 16:28 Working at new sites with no new data 17:13 Three days of T-shirt folding at CoRL 18:13 Opening Red Bull cans at live events 18:58 Summary 19:52 Q\&A --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/Sjfz1TqxzEs-youtube-9096b441d4cb2f01ced2 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/Sjfz1TqxzEs · 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.