Video Generation and Physical World Models
Tracks the evolution of video generation and embodied robotics toward physics-grounded world simulators.
1 sourcesVoices in this topic
All paragraphs
Oct 24, 2026
- 王淏楠 Video generation still depends more on pretraining, data, and capital, and 淏楠’s team never saw the “R1 moment” it expected. He began the year thinking RL could deliver an efficient leap, but changed his view after six months. His caveat: post-training scaling has …
- Real-time video is not mature, and embodied AI may not need to wait for a complete world simulation. Speeding generation up can break consistency, while improving quality often requires larger, slower models; 淏楠 cannot predict when genuinely …
Oct 13, 2026
- Vibhu The technical bet is Dyna-2, which Jason calls the first “world action model” trained on more than 1 million hours of data. It predicts future video or a representation of the future world and uses that to guide actions: “analogous to chain of thought—the models …
- Vibhu identifies a possible weakness in vision-only world models: missing modalities. He says a vision-plus-tactile model that is 100× more data-efficient could be an alternative, since tactile information may be denser than …
Oct 2, 2026
Oct 1, 2026
Sep 27, 2026
Sep 25, 2026
- Anastasis Germanidis Anastasis Germanidis argues there is no evidence that a JEPA-style architectural detour is needed for world models. His rebuttal to Yann LeCun's claim that "understanding the dynamics of the world is very different than generating cute videos" is blunt: …
- Real-time generation is framed as inevitable, not optional: "I'll be very surprised if two years from now we don't primarily use real-time models." Runway's characters model — a step-distilled autoregressive avatar model running at 24 FPS — is, he says, "the largest deployment of …
- Latent Space The most expansionary claim of the episode: interface world models render software as generated pixels — "there is no HTML, CSS, React... why generate the code that generates the pixels? Just generate the pixels directly." The end game is a "fully neural operating system," and …
- Runway "accidentally created a state-of-the-art model for robotics by just scaling video models," and the data thesis is substantial: third-person video is three orders of magnitude more plentiful than egocentric data, and "the most plentiful source of data ultimately wins." GWM-1, built on Gen 4.5, uses hundreds of hours of robot fine-tuning versus the hundreds of thousands to millions of hours in robotics …
- The Sora moment of February 2024 produced an "existential crisis for a few hours" amid "Runway's done" chatter, answered by a three-month sprint that scaled model size and compute by 10×, built model parallelism from zero expertise, and shipped Gen-3 that summer. That resilience story, plus the original "slightly irrational" Series B decision to build a 1,000-A100 cluster, is the pattern: bet ahead of …
- Anastasis Germanidis On competitive dynamics, Germanidis concedes the West is behind in video: "the majority of models in the top ten, top twenty are Chinese models" on leaderboards. The Cosmos Coalition, co-founded with NVIDIA, aims to bring world-model research into the open through open weights, physics benchmarks and …
