LAM playground is one of our new experiments to continue building our vision of developing a cross-platform general agent system that can help users perform actions.

we call it a "playground" for a reason: this is a place for you to see for yourself and play with what we’re working on. AI products by nature are always iterating rapidly. we always believe our community should be a part of this exciting growth — that’s why we are sharing our latest technical progress in a product format so your interactions with it help us better understand the strengths and limitations of our current technology. we want to know from all r1 users what they use the agent for and what they wish the agent can do so we can build out the product from there together.

r1 is the portal for you to talk with these agents that we call rabbits, who hop around the web to help you navigate websites, check information and take actions. try it today on the LAM playground tab in your rabbithole account and find more details in the LAM playground user guide.

since day one, our vision has been to develop a cross-platform general agent system that can help users get things done. We introduced our approach to this vision at our launch as large action model (LAM).

with the first generation of LAM, our approach was to build an agent system that could tackle a specific set of services — including Uber, Spotify, and DoorDash. but we soon discovered a number of technical limitations that impacted our ability to scale the core technology — both in terms of user experience and the number of connected services we can support. thanks to our community, fast feedback from real users made this rapid iteration possible.

LAM playground represents a radical rethinking of our approach and the first among peers to productionize this new technical architecture. the "LAM" in LAM playground was redesigned from the ground up to see and interpret the web differently than first generation LAM. the technology that powers LAM playground is no longer limited to a specific set of web applications — it's an agent system that learns to interact with any website you can think of.

rabbit is growing everyday and so are the rabbits. the hares in the LAM playground today might not do everything perfectly, but they are taking a solid hop from 0 to 1 on websites. next, we’ll work to unlock them across more platforms including mobile apps and desktop with your help.

we are so proud to have gathered an enthusiastic and likeminded community through the launch of r1. the most important thing is to improve as quickly as possible — which we’ve done with the support of our amazing community. now let’s hop around and find out.