Daily AI CatchupRunway releases Praxis-1: open-weight world action model that controls real robots
Praxis-1 is an open-weight world action model that converts Runway's video pretraining into real-world robotic control. The model is being tested with early partners on their own hardware and will be released publicly in the coming months. This bridges simulation and reality by training on video to directly control physical systems, sidesteping the classic sim-to-real transfer problem.
Why it matters
💻 Developer · Praxis-1 open-weight means you can finetune and deploy robot controllers without closed-garden dependency. Video pretraining to control is a big step toward generalizable robot learning. When it drops, this will be worth integrating into robotics stacks immediately.
📦 Product · Open-weight robot control is disruptive to commercial robotics: it lowers the barrier to entry for companies wanting to build customized robot systems. This accelerates adoption of robot automation outside the Tesla/Boston Dynamics ecosystem.
🎨 Design · Robot UX is still nascent. As control models improve, designers will have more agency in how robots behave in physical space. Praxis-1 sets a new baseline for what behaviors are achievable, which will reshape what humans expect from robots.
📈 Business · Open-weight models undercut proprietary robotics vendors. Runway is positioning itself as infrastructure, not the robot company. This is smart: control the underlying model, let others build hardware and applications on top.
🤔 Just Curious · Video-to-control models are a genuinely novel approach to robot learning—instead of hand-crafted reward signals, learn from watching. It's still early, but if it scales, it could reshape how robots learn skills.
Sources: Praxis-1