The field of human behavior prediction is currently thriving, with several startups securing significant funding. For instance, Simile recently obtained $200 million, reaching a $2 billion valuation; Aaru secured $88 million at a $1 billion valuation; and Humans&, an AI company, launched Persimmon for human behavior modeling after a substantial $480 million seed round in January, valuing it at $4.48 billion.
Currently, predicting human behavior largely involves using large language models (LLMs), which are either prompted or fine-tuned to emulate specific demographics. However, two-year-old San Francisco-based Mirror Particle views this methodology as fundamentally flawed.
Abhivyakti Ahuja, co-founder and CEO of Mirror Particle, a company offering an AI engine to predict consumer behavior and its underlying causes for brands, likens using LLMs for this purpose to “bringing a super soaker to Niagara Falls.” She explains, “LLMs have been trained on hundreds of billions of data points. How much can you influence its behavior by [fine-tuning] with such a small amount of data? It’s still stuck in the past.”
Ahuja contends that LLMs lack a human-like understanding of the world. She asserts, “LLMs are modeling written language, but humans are made of visual perception, spatial reasoning, social intelligence.” Consequently, she argues that depending on LLMs yields insights centered on what people fail to observe, which defeats the purpose of predicting human behavior.
Mirror Particle is adopting a different strategy: developing a foundational model, or what Ahuja calls a world model, constructed from the ground up to simulate human motivations and the evolution of their behavior over time.
Ahuja explained, “We don’t want to capture the static person. 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.” She further noted that a lack of change itself provides a valuable signal.
Mirror Particle has successfully completed an angel funding round and anticipates closing its initial venture round soon. Additionally, the company is set to participate next week in Startup Battlefield 200, TechCrunch’s prestigious startup contest held during TechCrunch Disrupt 2026 in San Francisco from October 13-15.
This startup utilizes a unique blend of data, encompassing client customer data, contemporary events, pop culture, and social media, among other sources, to construct models of demographic segments. It views these segments as dynamic systems that evolve through experiences, meticulously tracking shifts in motivations. A significant emphasis is placed on “revealed behavior”—observing actual actions instead of relying on self-reported survey responses.
Similar to its competitors, Mirror Particle’s initial market entry strategy targets areas with established budgets for such insights, namely market research, brand, and product strategy. For example, Mirror could assist a beauty company not only in crafting more effective ad copy for makeup appealing to Gen Z but also in ascertaining if that demographic truly desires the product in the first place.
Ahuja questioned, “What if [the target demographic] doesn’t want eyeshadow palettes? Maybe blush is a better option to go for if you want to sell a product to this market.”
Additionally, Mirror Particle’s prediction engine offers clients insight into the underlying reasons for present or future behavior—including motivations, limitations, and contextual factors—which validates its recommendations and assists brands in making more informed choices.
During an initial pilot program, a prominent pet food company sought advice on packaging imagery to enhance sales, pondering options like chicken, beef, or vegetables. Mirror’s technology revealed that the brand was misidentifying the core issue. The imagery itself was irrelevant; the actual challenge stemmed from the brand’s widespread recognition making it perceived as mass-market and inexpensive, indicating sales would stagnate until this perception was rectified.
Ahuja explained, “The way we see our model evolving is like how a baby learns about the world,” highlighting that infants progress from visual perception to language acquisition, then body awareness, and finally social intelligence.
Ahuja’s foundational interest in simulating the human brain stems from her academic background in neuroscience and computer science. Hailing from India, she pursued her studies at the University of Toronto, where she found inspiration in the pioneering work on neural networks by AI luminary Geoffrey Hinton.
Following her education, Ahuja joined Amazon Robotics, where she developed robots designed to construct other robots. It was there she connected with her co-founders, Will Song and Thomson Yen. Song dedicated a significant portion of his career to developing sales personalization engines, while Yen concentrated on leveraging deep learning to understand how AI agents comprehend human behavior.
The startup envisions itself becoming the “general layer for anticipating human behavior” in the long term, transitioning from analyses at a broad population level to insights tailored for individuals.
Ahuja stated, “We just need a better model of humans if we’re going to work alongside AI and with each other.”
Discover Mirror Particle and numerous other innovative startups, all rigorously evaluated by TechCrunch’s editorial team, next week at Disrupt in downtown San Francisco. The victor of this year’s Startup Battlefield will be chosen by our panel of VC judges on the afternoon of Thursday, October 15.