Robotics & World Model Reading Club 02: JEPA Zoo, WAMs & Unified Latent Representations – San Francisco
- Date
- 2026-04-04
- Location
- San Francisco, CA, USA
- Host
- Saturday Robotics
About this event
Robotics is moving fast, but the most interesting progress is not just in better policies or bigger datasets, it is in how machines build internal representations of the world. This reading club is for people who want to go beyond surface-level summaries and actually dig into the ideas shaping world models, JEPA-style learning, and unified latent representations. If you care about how robots can understand, predict, and act with more structure and less supervision, this is the room you want to be in. What Is This? This is the second session of the Robotics & World Model Reading Club, an in-person gathering in San Francisco for people who want to read, discuss, and pressure-test core ideas in modern robot learning. The focus for this meetup is JEPA Zoo, WAMs, and unified latent representations: a cluster of topics that sit at the intersection of self-supervised learning, embodied intelligence, and scalable robot generalization. Rather than treating papers as isolated artifacts, the group will look at how these ideas connect. What kinds of representations are actually useful for downstream control? What do predictive latent spaces buy us in robotics? Where do world action models fit in? And which assumptions seem promising versus fragile when moved from benchmarks into real embodied systems? The format is built for serious but accessible discussion. You do not need to arrive with a polished opinion, but you should come ready to engage with technical ideas, ask questions, and think in public with other people who care about robotics research. What to Expect Expect a discussion-first meetup with enough structure to keep the conversation focused and enough openness to let strong questions lead somewhere interesting. This is not a lecture series and not a generic networking mixer. The value comes from reading-related discussion with people who are actively trying to understand where the field is going. A typical flow will likely include: Arrival and informal introductions so people can quickly understand each other's interests across robotics, representation learning, and world models Framing the session theme around JEPA variants, WAM-related concepts, and the case for unified latent spaces Group discussion of key readings and ideas, including what feels technically elegant, what seems underexplored, and what may not transfer cleanly to embodied settings Open conversation and networking with other attendees working on or curious about robot learning The conversation may touch on questions like: What makes a latent representation useful for planning or control rather than just compression? How should we think about prediction in latent space versus pixel space? What parts of JEPA-style training seem especially relevant for robotics? Where do world models break down when data is messy, multimodal, or physically grounded? Can a unified representation meaningfully support perception, forecasting, and action together? Because this is an in-person reading club, the pace can be more interactive than online paper discussions. People can challenge assumptions, compare interpretations, and connect ideas across papers, labs, and practical robotics constraints. Why Attend If you are trying to keep up with the shift toward structured representation learning in robotics, this meetup gives you a better way to do it than reading alone. You will hear how other technically engaged people interpret the same ideas, which often makes the important distinctions clearer: what is novel, what is repackaged, and what might actually matter for building better robot systems. This is also a good setting for sharpening your own thinking. Reading clubs are valuable not only because of the material, but because explaining an argument out loud forces you to see what you understand and where your assumptions are still fuzzy. You may leave with new questions, but they will be better questions. You can also expect value from the room itself. San Francisco brings together researchers, engineers, founders, students, and independent builders who are all circling related problems from different angles. If you care about embodied AI, it helps to be in conversation with people who think at the level of systems, learning objectives, and deployment realities rather than just headlines. You should attend if you want to: Build a clearer mental model of JEPA-related ideas and their relevance to robotics Understand how world models and action-aware latent representations are being discussed right now Compare research intuitions with others who are reading closely Meet people in the local robotics and embodied AI community Spend time on technical discussion that is focused, current, and genuinely useful Practical Details This event is in person in San Francisco, USA. The in-person format is important: it creates more natural technical discussion, easier back-and-forth, and better opportunities to continue conversations after the core session wraps. The meetup takes place on Saturday, April 4 at 2:00 PM PDT. That timing makes this a strong weekend option for people who want space for a thoughtful, longer-form discussion without trying to squeeze it into a workday schedule. If you are thinking about coming, the best preparation is simple: Review the session theme in advance if you can Come ready to discuss ideas, not just consume them Bring your questions, disagreements, and half-formed hypotheses Expect a room that values curiosity, rigor, and conversation over performance Whether you are deep in robot learning already or actively trying to get more grounded in world-model thinking, this reading club offers a concrete way to learn with others who are paying attention to the same shift in the field.
Who should attend
This is for you if you want a more thoughtful, technical way to engage with robotics and world-model research than scrolling papers alone. - You work in **robotics, embodied AI, or machine learning** and want to discuss JEPA-style methods, world models, or latent representations with other people who can meet you at that level. - You are a **researcher, engineer, student, founder, or independent builder** who learns best by talking through ideas, assumptions, and tradeoffs in a live room. - You are trying to understand how **self-supervised and predictive representation learning** may change the way robots perceive, plan, and act. - You care about the gap between **clean research abstractions and real robotic systems**, and you want discussion that does not stop at paper summaries. - You want to meet people in the **San Francisco robotics community** who are serious about current research directions and open to exchanging perspectives. - You do not need to be an expert on every reading, but you should be excited to **ask good questions, contribute your interpretation, and engage thoughtfully** with technical material.