Thu Feb 19 14:44 EST
While scrolling on LinkedIn today I found about this really interesting startup and they raised 100M$! Series A. I got this very same interesting idea last week -- to understand how Agents living in an apartment/small town would live while I orchestrate all the real-world problems (corruption, politics, social heirarchy, professions, etc.) and ofcourse these agents will have a emotional persona injected into them either by me or an Agent itself. This is kind of like an aquarium, while we observe how the Agents react, interact with each other.
I'm particularly curious how they would react if the simulation were run in oppressive empire? maybe they'll overthrow the empire? in a non-violent way??
The more I think about it, the more I realize we're basically just glorified fish keepers, except our fish have PhDs and existential crises. But here's the thing -- what would we actually learn? Would these simulations tell us something profound about human nature, or would they just reflect our own biases in how we program them? Like, I could code an agent to be "resilient" but my definition of resilience is probably different from someone who grew up in a completely different environment than me.
Still, there's something compelling about the idea. Maybe it's not about predicting human behavior but understanding the edge cases — the weird, emergent behaviors that nobody expected. The moments where the agents surprise you. That's probably where the real insights are hiding.
Anyway, I might actually try building a small prototype. Nothing as ambitious as 100M worth of infrastructure, but maybe a small sandbox with like 10-15 agents in a neighborhood. See what happens. Worst case scenario, I waste a weekend and have a cool story. Best case? I stumble onto something interesting that's worth exploring further.
I'll probably write a follow-up if I end up building anything. Or maybe I'll just forget about this idea like the other 50 side project ideas I had this month. We'll see.