$200M · AI Research + Simile

What they're building

Simile builds a foundation model for human behavior — think 'synthetic users' you can run experiments on before touching real customers. Companies use 'agentic twins,' generative AI agents grounded in real human interview and behavioral data, to test products, marketing, pricing, or policy decisions on a simulated population before real-world rollout. It ships with a first-of-its-kind confidence model that predicts how accurate each simulation is. CVS built 100,000 AI patient twins from 2.9M records to test pharmacy experiences, replicating in hours what took months. Simile grounds its agents in consented real human data rather than generic LLM personas, which is the fidelity claim it competes on.

Why this matters

Greenoaks (Stripe, Figma, Nubank, Brex) leading a $200M round five months after a $100M Series A is a conviction signal you rarely see. CEO Joon Sung Park is the Stanford researcher behind the landmark 'Generative Agents'/'Smallville' work. The stated mission: simulate all eight billion people on Earth, accurately. Know the key benchmark and its caveat: per Park et al. (arXiv:2411.10109), the generative agents 'replicate participants' responses on the General Social Survey 85% as accurately as participants replicate their own answers two weeks later,' based on agents built from qualitative interviews with 1,052 real individuals — impressive, and exactly the number skeptics will probe.