Mirror Particle Builds a World Model Simulating Human Behavior

On October 7, 2026, TechCrunch reported that San Francisco startup Mirror Particle is building a "world model" foundation model from scratch, designed to simulate human behavior and the reasons behind its changes. The company believes the industry's common approach—fine-tuning large language models (LLMs) to predict consumer behavior—simply doesn't work, so it has chosen not to take the LLM fine-tuning route and is instead building this foundation model from the ground up. This is the first report on the event; details disclosed by the company so far are limited to its technical positioning (a world model that simulates human behavior) and its rejection of the LLM fine-tuning route. There have been no reports yet of product launches, funding, or real-world deployments.

AI overview of reporting · Updated 11 days ago