Simulated
Characters

Simulated
Characters

We're a team of passionate gamers and artists who believe that AI characters capable of behaving like real people can transform the way we play, connect, and immerse ourselves in games and simulations.

Our Mission

AI characters in video games indistinguishable from human gamers

We're on a mission to build intelligent, embodied characters that bring human-like behavior to games, simulations, and beyond. By pushing the boundaries of behavior, we're unlocking bold new ways to play, build, and connect.

Our Work

New Kinds of Gameplay

Our characters understand game worlds with their own goals and personalities, creating emergent player-driven stories.

Your Friends Anywhere

AI Characters that can play any game that your friends can, through our open-source protocol.

Developer-First Tools

We're building generalizable solutions that make it easy for developers to drop AI characters into their own experiences.

Timeline

Jul '23

Stanford Simulacrum and Embodied Characters

Inspired by Stanford's groundbreaking Simulacrum paper, we set out to bring AI-driven inner lives and motivations to characters in 3D worlds. Our early experiments let users create custom AI characters whose personalities shape their actions and interactions in multi-agent scenarios.

Dec '23

YC and Simulated Inner Lives

After joining Y Combinator we debuted a large-scale 3D simulation on Twitch, leading to a viral moment where a character with access to the source code discovered they were in a simulation and attempted to escape, falling off the map.

Aug '24

Consumer Testing at Scale

We built and launched multiple public prototypes, from a Valheim companion mod to a Roblox roleplay world, testing embodied AI characters at scale and validating consumer demand for lifelike AI presence with over a million players.

Nov '24

Behavior is Everything

Through experimentation and analysis, we developed the hypothesis that behavior is the key driver of engagement. We published our Behavior White Paper and presented our reinforcement learning paper at NeurIPS, establishing our research leadership in this area.

Jul '25

Foundation Model and Consumer Product

We have begun building a foundation model specifically for human-like behavior in games, alongside a consumer product to both showcase the model and generate data to improve it.