Mirror Particle is transforming how we predict human behavior with its distinctive approach. Rather than depending on large language models (LLMs), which mainly focus on written language, this San Francisco-based startup is developing a foundational model that mimics the intricacies of human behavior over time. Co-founder and CEO Abhivyakti Ahuja points out that LLMs miss the essence of human experience, which is deeply rooted in visual perception, spatial reasoning, and social intelligence. “It’s like bringing a super soaker to Niagara Falls,” she remarked, highlighting the limitations of fine-tuning LLMs with insufficient data.
Central to Mirror Particle’s mission is the goal of understanding and predicting how individuals change. Ahuja explained, “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.” This strategy goes beyond static models, acknowledging that a person’s motivations and behaviors can shift over time. Even if individuals remain unchanged, that fact itself is a significant signal to consider.
The startup utilizes a proprietary blend of data sources, including clients’ customer data, current events, pop culture, and social media. By merging these diverse inputs, Mirror Particle creates a dynamic model of demographic segments. This model treats human behavior as an evolving system, examining how experiences shape motivations and actions. Ahuja emphasized the importance of “revealed behavior,” concentrating on what people actually do rather than just relying on self-reported survey responses.
Initially, Mirror Particle’s market strategy aims at established budgets in areas such as market research, brand strategy, and product development. For example, they could help a beauty brand not only refine ad copy for Gen Z makeup products but also determine if that demographic is truly interested in the product. “What if [the target demographic] doesn’t want eyeshadow palettes?” Ahuja asked, suggesting that alternatives like blush might be a better match for that market.
Additionally, the prediction engine developed by Mirror Particle provides valuable insights into the motivations and constraints that drive consumer behavior. This context enables brands to make more informed decisions. An early pilot with a well-known pet food brand illustrates this well. The brand sought guidance on which imagery–chicken, beef, or vegetables–would boost sales. Mirror’s analysis revealed that the brand was asking the wrong question, as the imagery didn’t matter until they addressed the perception of being mass market and cheap.
Ahuja compared the evolution of their model to the developmental stages of a baby, stating, “The way we see our model evolving is like how a baby learns about the world.” This analogy reflects their commitment to thoroughly understanding human behavior. Ahuja’s background in neuroscience and computer science, along with her studies at the University of Toronto under AI pioneer Geoffrey Hinton, has shaped this vision. Her previous role at Amazon Robotics, where she helped build robots, further fueled her interest in modeling intelligent behavior.
Co-founders Will Song and Thomson Yen bring complementary skills to the team. Song has focused much of his career on developing sales personalization engines, while Yen has concentrated on leveraging deep learning to understand how AI agents interpret human behavior. Together, they aim to position Mirror Particle as a crucial layer for anticipating not just population-level trends but also individualized insights into human behavior.
Mirror Particle’s innovative approach highlights the need for a model that captures the complexities of human behavior rather than relying on traditional methods. By focusing on longitudinal data and revealed behavior, the company aims to provide brands with insights that reflect actual consumer motivations and actions. This shift towards understanding the evolving nature of individuals could significantly impact how businesses engage with their customers, moving beyond static assumptions to a more dynamic, responsive strategy.
