filesz supplies specialized multimodal video training data to AI labs developing world models and, increasingly, AGI.Our work rests on three core beliefs:(1) AGI will unlock the greatest value creation of our time.
(2) AGI will be trained on video that reflects real-world physics, context, and natural laws.
(3) A platform that can reliably deliver this kind of training data is inevitable.
Our focus within video data spans domains that matter most for world models and embodied intelligence, including:• Egocentric real-world video: wearable camera footage of daily tasks
• Physical interaction & manipulation footage: sequences of tools use
• Cause-effect and learning trajectories: walking, flipping, pushing, and observing outcomes.
We are happy to talk about partnerships and questions at [email protected]