Robot Learning #02 — World Models & Dexterous Manipulation

the vibe

How does a model's understanding of the physical world translate into reliable manipulation? Learning to predict what happens next and learning to act both depend on the training data, representations, and feedback available to the system. The second Robot Learning meetup brings together builders working on world models, vision-language-action models (VLAs), and dexterous manipulation. Hosted by HackerSquad and presented by Bright Data, the evening puts data at the center of the conversation: what models can learn from video and demonstrations, how that learning connects to control, and how teams evaluate progress. We'll start with a few short demos, hear from practitioners in a panel discussion, and leave plenty of time to meet other robotics and physical AI teams. What we'll explore World representations, prediction, and their relationship to planning and action. Learning manipulation from demonstrations and other visual data, including the limits of transfer to a physical task. Building useful training and evaluation datasets, understanding failure cases, and improving robustness. Who this is for Robotics founders, researchers, ML engineers, and builders working on world models, VLAs, manipulation, and training data for physical AI. Bring a research question, a data bottleneck, or lessons from a system you are building. Schedule All times are Pacific Time (America/Los_Angeles). 5:00 PM — Doors Open: Check in, meet fellow builders, and settle in. 5:45 PM — Programming…