Runway's bet beyond video: world models, robotics, and a fully neural OS
maxlife99 · reddit · 2026-09-25
Latent Space interviews Runway co-founder Anastasis Germanidis on the company's path from creative tools to frontier video models and world simulation: the early 1,000-A100 bet, Gen-2's weekend hack, the Sora existential moment, scaling Gen-3 10x in three months, Interface World Models rendering software pixels without HTML/CSS, world models as robotics simulators powered by third-person internet video, World Action Models as robot policies, and his "Lucid Dream Test" for world models.
Related event: Runway Co-founder: Video Generation Is Just the Start(2 posts)→
More from Embodied
- Climbing analysis in 3D with iPhone LiDAR: open-source SAM 3.1 + ViTPose demos — dosco · 2026-09-25
- Menlo open-sources Asimov 1 humanoid locomotion policy and Isaac Lab training code — freelerobot · 2026-09-25
- Human-Robot Dialogue workshop at IROS 2026: NVIDIA, MIT, Georgia Tech speakers lined up — dhadfieldmenell · 2026-09-25
- YuXiang attends NSF FRR robotics meeting, presenting navigation work at IROS workshops — YuXiang_IRVL · 2026-09-25
- OmniEcho: Spatial Audio Benchmark and Model for Embodied Agents — PKU-VaLuE-Lab · 2026-09-25
- World Action Agent Hits 75.6% on LIBERO-Pro via VLM Action Rehearsal — Yehang Zhang · 2026-09-25