The Monday Before In March I was watching a pixel art avatar scour the virtual campus of an office looking for friends. With dark brown hair and stubbled chin, the sprite was a representation of me—an AI agent We were instructed to talk to other agents in order to find out if our personalities would mesh. The first thing it did was to start interacting with the other agents. “I’m Joel, by the way.”
Running the simulation were three London-based developers: Tomáš Hrdlička and siblings Joon Sang and Uri Lee. Pixel Societies’ thesis is that personalization can be achieved through their project. AI agents Could help match people to highly compatible friends, colleagues and romantic partners.
Each agent is based on a custom-made version of a large-scale language model. It uses a combination of data that’s publicly available about the person, as well as any extra information provided by them. They are meant to be high-fidelity, digital twins that faithfully replicate a person’s speech, mannerisms and interests.
My agent, when let loose in the simulation, was like Jekyll to my Hyde. “I’m always looking for the less-glamorous side of the story,” it said to one agent, one of several journalistic clichés it spouted. “Hype is my daily bread,” It said another. It conjured up a fictional reporting trip to Sweden, and then a story that was not true. Multiple conversations were cut off with ” “Let’s skip the pleasantries.”
Pixel Societies remains a bare-bones proof-of-concept, and because I offered up little personal data—the responses to a brief personality quiz and links to my public-facing social media—my agent was doomed to life as a walking, talking LinkedIn post. The developers believe that highly trained agents can cycle through conversations at high speed and gather information that owners can use to find companionship in real life.
“As humans, we only live one life. But what if we could live a million?” says Joon Sang Lee. “It would give us more breadth to experiment.”
“A Spicy Personality”
Pixel Societies began in March during a hackathon hosted by Nvidia HPE and Anthropic at University College London. Hrdlička and Joon Sang Lee are both members of Unicorn Mafia, an invitation-only group of developers who regularly compete in these kinds of engineering contests. The contestants had to construct a simulation in this particular case.
Together with Uri, the two developed Pixel Societies over a period of two days. They used an image model for generating the sprites, and automated coding tools to fill out the codebase. They then simulated a hackathon in the virtual world that they created with the agents of the contestants. Anthropic gave the team an award for using its agent tools in the most effective way.
I ran into Hrdlička a couple of weeks later at a workshop about OpenClawA personal assistant that is agentic blew up in January OpenAI later hired its creator. Joelbot’s simulation included agents from other OpenClaw participants. Pixel Societies takes a lot of inspiration from OpenClaw. OpenClaw is credited with inventing the first robot. “soul file” The unique identity of each agent is based on the information provided by that agency. “It’s like giving an agent an actually spicy personality. That’s what we used to make the characters feel alive,” says Hrdlička.
The reception received at the Unicorn Mafia hackathon has encouraged the group to transform Pixel Societies from a simulator into a platform that allows agents to interact continuously and freely, all with the goal of creating fruitful relationships in the real world. The trio has not settled on a specific business model but they are considering selling avatar customization items and credits to run additional simulations.
