Living Agents - Watch AI agents live, think, and survive in a shared world

I built Living Agents because I wanted to answer a simple question: what happens when you give AI agents mortality, personalities, and a world to live in, then just let them go? Agents run on either Claude or Grok, and the difference is stark. Claude agents tend to be more cautious and cooperative, forming alliances and sharing resources. Grok agents are wilder, more impulsive, more confrontational, more likely to do something completely unexpected. Mixing both in the same world creates genuinely interesting dynamics. Each agent has a randomized personality (traits, fears, desires) that shapes every decision. They perceive their surroundings, reason about what to do, and choose an action: forage, craft, talk, fight, trade, explore, and more. You control the pace by ticking the world forward manually, so you can watch each round of decisions unfold and see exactly what every agent was thinking. The most surprising moments have been the ones I didn’t design for: agents teaching each other crafting recipes they discovered, forming grudges after being attacked, or writing letters to their children when they spawn a new agent. I’d love to see what happens in your worlds, so if you spot funny, weird, or unexpected agent behavior, please share it here or drop me a message. The best part of this project has been the surprising in-world moments. Got questions about how it works or ideas for what you’d like to see next? Get in touch, I’m happy to chat!