Mirror Particle Builds 'World Model' to Predict Human Behavior — SkimNews
Get the Culture newsletter
Daily culture — film, music, books, the trends and ideas worth your attention. Free.
- Mirror Particle is building a foundation 'world model' from scratch that simulates why humans do what they do and how behavior changes over time, rather than relying on LLMs to role-play target demographics
- CEO Abhivyakti Ahuja argues LLMs are 'fundamentally broken' for predicting human behavior, comparing fine-tuning them to 'bringing a super soaker to Niagara Falls' because the models are trained on far more data than any fine-tune can override
- Mirror Particle has closed an angel round, says it is close to closing its first venture round, and will compete next week in TechCrunch's Startup Battlefield 200 at Disrupt 2026 in San Francisco on October 13–15
- The startup combines clients' customer data with current events, pop culture, and social media to model demographic segments that evolve over time, focusing on 'revealed behavior' — what people actually do — rather than self-reported survey answers
- In a pet food brand pilot, Mirror's engine found the client was asking the wrong question: packaging imagery (chicken, beef, vegetables) didn't matter because the brand's mass-market, cheap perception was the real ceiling on sales
- Ahuja, originally from India, studied neuroscience and computer science at the University of Toronto, where she was inspired by Geoffrey Hinton, and later worked at Amazon Robotics, where she met co-founders Will Song and Thomson Yen
- Rival startups in the human-behavior-prediction space have raised large sums this year: Simile at $200 million on a $2 billion valuation, Aaru at $88 million on a $1 billion valuation, and Humans& at $480 million on a $4.48 billion valuation for its Persimmon product
Why it matters: Mirror Particle is entering a category that has already attracted roughly $768 million across three well-funded rivals (Simile, Aaru, Humans&) in the past year, so its bet is that a custom world model — not a fine-tuned LLM — is the only way to capture the 'changing person.' The pet food pilot shows what that buys: instead of optimizing the question a client asked, the model surfaced a perception problem the brand hadn't even considered, reframing the brief entirely.
Ask SkimNews
