Mirror Particle, a San Francisco-based startup, is developing a world model to predict human behavior, competing with AI firms like Humans& and Aaru. The company argues that traditional large language models (LLMs) are inadequate for capturing the complexities of human behavior.
Mirror Particle's co-founder and CEO, Abhivyakti Ahuja, criticized LLMs for being trained on vast data but still failing to grasp the nuances of human perception, spatial reasoning, and social intelligence. “LLMs are modeling written language, but humans are made of visual perception, spatial reasoning, social intelligence,” she said.
Ahuja emphasized that relying on LLMs provides insights based on what humans don’t notice, which is not useful for predicting behavior.
Instead of using LLMs, Mirror Particle is building a foundation model from scratch that simulates why humans act the way they do and how behavior evolves over time. “We don’t want to capture the static person,” Ahuja said. “We want to capture the changing person.
That means capturing the longitudinal data on how people are changing, what triggers are changing them and to what degree.”
The startup has already raised an angel round and is close to closing its first venture round. It is also competing in Startup Battlefield 2, TechCrunch’s renowned startup competition, at TechCrunch Disrupt 2026 in San Francisco on October 13-15.
Mirror Particle uses a proprietary combination of data, including customer data, current events, and social media, to model demographic segments as evolving systems.
Mirror Particle’s approach focuses on “revealed behavior” rather than self-reported survey answers. Its initial go-to-market strategy targets market research and brand strategy, where budgets already exist for such insights. For instance, it might help a beauty brand determine if a Gen Z demographic wants eyeshadow palettes or if blush is a better option.
The startup’s long-term vision is to be the “general layer for anticipating human behavior” and move from population-level analyses to individual-level insights. Ahuja said, “We just need a better model of humans if we’re going to work alongside AI and with each other.”
Source: techcrunch
