Simile raised $200 million at a $2 billion valuation over the last year. Aaru raised $88 million at a $1 billion valuation. Humans&; announced a $480 million seed round in January at a $4.48 billion valuation. These figures show that startups claiming to predict human behaviour are attracting serious capital.
Mirror Particle rejects the standard approach
Most current methods rely on large language models prompted to role-play as a specific group. Mirror Particle, a two-year-old San Francisco company, considers this method broken. Abhivyakti Ahuja, the co-founder and CEO, compares the approach to bringing a super soaker to Niagara Falls. She argues that fine-tuning a model trained on hundreds of billions of data points with a small dataset does not change its behaviour. The system remains stuck in the past.
Ahuja believes large language models do not see the world as humans do. They model written language, whereas humans rely on visual perception, spatial reasoning, and social intelligence. Using these models produces insights based on what humans do not notice, which fails when trying to predict actual behaviour.
A foundation model built from scratch
Mirror Particle is building a foundation model described as a world model. This system simulates why humans act and how their behaviour changes over time. The company does not want to capture a static person. It focuses on longitudinal data showing how people change, what triggers those changes, and the degree of that change. If people are not changing, the company treats that as a signal in itself.
The startup has raised an angel round and is close to closing its first venture round. It is competing in Startup Battlefield 200 at TechCrunch Disrupt 2026 in San Francisco on October 13-15.
Data sources and market focus
Mirror Particle uses a proprietary combination of data. This includes client customer data, current events, pop culture, and social media. The company models a demographic segment as a system that evolves. It tracks how motivations shift as the group moves through experiences. Much of the focus is on revealed behaviour. This means looking at what people actually do rather than self-reported survey answers.
Like competitors, the initial go-to-market strategy targets existing budgets for market research and brand strategy. For example, a beauty brand might use the tool to determine if a demographic wants a product before writing ad copy. Ahuja notes that a brand might find the target audience does not want eyeshadow palettes. Blush might be a better option for selling to that market.
The prediction engine provides the reasons behind current or future behaviour. It lists motivations, constraints, and additional context to justify recommendations. This helps brands make smarter decisions.
A pilot case study
In one early pilot, a well-known pet food brand asked what imagery to put on packaging to boost sales. The options included chicken, beef, and vegetables. Mirror’s technology found the brand was asking the wrong question. The imagery did not matter. The problem was that the brand was so recognisable it was seen as mass market and cheap. Sales would plateau until the brand addressed that perception issue.
Long-term vision and background
Ahuja compares the evolution of the model to how a baby learns about the world. She notes that babies move from vision to language to body awareness to social intelligence. This interest in modelling the human brain comes from her background in neuroscience and computer science. Originally from India, she studied at the University of Toronto. She became inspired by AI pioneer Geoffrey Hinton’s contributions to neural networks.
After school, Ahuja worked at Amazon Robotics building robots that build other robots. There she met co-founders Will Song and Thomson Yen. Song has spent his career building sales personalization engines. Yen focused on using deep learning to understand how AI agents perceive human behaviour.
The startup’s long-term vision is to become the general layer for anticipating human behaviour. The goal is moving from broader population-level analyses to individual-level insights.
“We just need a better model of humans if we’re going to work alongside AI and with each other,” Ahuja said.
