Osmosis (YC W25) reposted this
On Saturday, we brought together 200+ founders, researchers, and engineers for RL IRL at Y Combinator HQ! At Osmosis (YC W25), our belief is that RL is the missing piece for AI agents that truly learn and work - but despite all the recent attention on RL, there weren't any major in-person events focused on how to actually *apply* RL. We wanted to fix this, so we brought together some of our friends to talk about how they're using RL today! The response was overwhelming! We had over 800 registrations and were literally at max capacity at YC. The talks covered everything from how to set up products for RL, to how to design RL training loops, to how companies build & scale RL environments today: -Osmosis (YC W25) and Greptile - Designing RL-Ready Products -ZeroEntropy (YC W25) - Scaling Reranker Models with RLAIF -Reducto - RL for Ill-Defined Objectives -hud (YC W25) and The LLM Data Company (YC X25) - Building & Scaling RL Environments -Encord - Data Scaffolding for Multimodal RL -Turing - RL Environments for the Enterprise *Link to thread of summaries + session recordings is in the comments.* Thank you so much to our co-organizers that helped bring RL IRL together, and a special thank you to all the attendees that came! We'll be hosting more RL-focused events - in the meantime, if you're interested in learning about how to use RL for your own use cases, don't hesitate to reach out!