The rapid progress of AI has caused many to wonder about whether we might be in a simulation ourselves, and there’s now a company that’s looking to create simulations with AI agents at scale.
Simile, an AI simulation company founded by researchers who pioneered the concept of “generative agents,” announced today it has raised $100 million in funding led by Index Ventures. The round includes participation from Hanabi, A*, Bain Capital Ventures, and notable AI figures including Andrej Karpathy, Fei-Fei Li, Adam D’Angelo, Guillermo Rauch, and Scott Belsky.
The company is building what it describes as the first AI simulation of society, populated by agents based on real humans. Its long-term vision is ambitious: simulating entire worlds with trillions of interacting decisions across individuals, organizations, cultures, and states.
From Research to Reality
Simile’s foundation lies in groundbreaking academic research. In 2023, CEO Joon Sung Park and his team introduced the concept of generative agents with a project called Smallville—a simulated town where AI agents exhibited surprisingly human-like behavior without explicit programming. The agents autonomously planned their days, formed relationships, and even organized a Valentine’s Day party after being given just a single initial prompt.
“Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations,” the researchers wrote in their paper introducing the technology. The architecture extends large language models with memory systems that store experiences, synthesize reflections, and dynamically retrieve information to guide behavior.
Park’s team also helped introduce the term “foundation model” and pioneered what is now recognized as the field of AI-based simulation. This academic pedigree gives Simile unusual credibility in a space where many startups make grandiose claims about AI capabilities.
A Flight Simulator for Human Decisions
The company positions its platform as “a flight simulator for human decisions”—a way to rehearse consequential choices before implementing them in the real world.
“Pilots don’t train with real passengers. Surgeons don’t practice on real people. Actors don’t rehearse with real audiences,” the company notes in its announcement. “Yet the most consequential decisions in society get pushed straight to prod.”
Simile’s platform aims to change that by allowing organizations to model how real people make decisions, then compose those individual behaviors into bottom-up simulations. Users can change one assumption, constraint, or person and watch the world “recompile,” running counterfactuals impossible to test in reality.
The company is developing a foundation model that predicts human behavior in any situation and at any scale. Leading companies are already using Simile to rehearse earnings calls, model litigation outcomes, and test policy changes before implementing them.
Validation Through Partnership
In a significant validation of its approach, Simile has partnered with Gallup to anchor its simulation platform in a probability-based, nationally representative panel. This partnership allows the company to model decisions against real-world sentiment with what it describes as transparency, replicability, and empirical validation.
The enterprise simulation workflows promise to help companies find their audience, reach niche populations, de-risk decisions, understand customer segments, and generate instant insights from concept testing.
The Technical Challenge
Park emphasizes that simulating human behavior represents “one of the most important and technically difficult problems of our time.” The challenge lies not just in modeling individual decision-making, but in capturing the emergent behaviors that arise when those individuals interact at scale.
The company’s research demonstrated that three components are critical to believable agent behavior: observation (how agents perceive their environment), planning (how they decide what to do), and reflection (how they synthesize experiences into higher-level understanding). Removing any of these elements degraded the believability of the simulations.
Looking Ahead
With $100 million in funding and technology that has already moved from academic curiosity to commercial application, Simile is positioning itself at the intersection of AI research and practical business needs. The question is whether the technology can scale from simulating a small town of 25 agents to modeling the complex, interconnected decisions of entire societies. And if successful, Simile’s vision would fundamentally change how organizations approach decision-making—replacing intuition and luck with data-driven previews of potential outcomes. Whether that vision materializes remains to be seen, but the company now has substantial resources and credibility to pursue it.