Definition
Plain language
A well-known 2023 experiment that dropped AI characters into a small simulated town and let them socialize.
As stated in the literature
Stanford's generative-agents simulation introducing the memory-and-reflection architecture for LLM agents; ran one to seven simulated days with a single model, cited as the lineage behind later long-horizon multi-agent platforms.
Why it matters: It demonstrated how giving agents memory and reflection makes their behavior more lifelike, setting a template that later multi-agent simulations build on.
For example, simulated townspeople with their own memories planned a Valentine's party and spread word of it to each other over several simulated days.
Heard on the show
“Stanford's Smallville did it back in twenty-twenty-three — that's where the memory-and-reflection architecture comes from — but it ran for one to seven simulated days.”Episode 123 — Five Identical Worlds, One Swapped Model: What Happens When AI Agents Run for Fifteen Days