Mirror Particle's Ambitious Quest: Building a 'World Model' of Human Behavior with AI

The landscape of technology startups is currently witnessing a surge in companies dedicated to predicting human behavior.

Notable examples include Simile, which recently secured $200 million at a $2 billion valuation, Aaru, which raised $88 million at a $1 billion valuation, and Humans&, an AI startup that launched Persimmon to model human behavior after a massive $480 million seed round valuing it at $4.48 billion.

However, amidst this trend, San Francisco-based Mirror Particle, a two-year-old AI startup, challenges the prevailing approach that heavily relies on large language models (LLMs) for human behavior prediction.

The co-founder and CEO of Mirror Particle, Abhivyakti Ahuja argues that using LLMs in this manner is fundamentally flawed.

She likens it to "bringing a super soaker to Niagara Falls," suggesting that attempts to fine-tune LLMs with small datasets to role-play as target demographics are ineffective.

Ahuja emphasizes that LLMs, trained on vast amounts of written language, cannot truly perceive the world as humans do, who rely on visual perception, spatial reasoning, and social intelligence.

Therefore, insights derived from LLMs may miss crucial aspects of human behavior, focusing instead on what humans might not consciously notice.

Mirror Particle is pioneering an alternative strategy: building a proprietary foundation model, or "world model," from the ground up. This model is designed to simulate the underlying reasons for human actions and track how human behavior evolves over time.

Ahuja explains, "We don’t want to capture the static person. We want to capture the changing person."

This involves analyzing longitudinal data to understand how individuals change, what triggers these changes, and the degree of their impact. Even a lack of change is considered a significant signal.

The startup employs a unique combination of data sources, including clients’ customer data, current events, pop culture trends, and social media interactions.

This diverse data pool allows Mirror Particle to model demographic segments as dynamic systems, continuously tracking shifts in motivations as individuals navigate various experiences.

A core focus is on "revealed behavior"—what people actually do—rather than potentially biased self-reported survey responses.

Mirror Particle's initial go-to-market strategy, similar to its competitors, targets established budget areas such as market research, brand strategy, and product development.

For example, Mirror Particle could assist a beauty brand not just in crafting effective ad copy for a product like makeup appealing to Gen Z, but also in determining if that demographic genuinely desires the product.

Ahuja illustrates this: "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."

Beyond predictions, Mirror Particle's engine provides customers with the "why" behind current or future behavior, offering insights into motivations, constraints, and contextual factors that justify its recommendations.

This empowers brands to make more informed decisions, and in an early pilot, a prominent pet food brand sought to optimize packaging imagery (e.g., chicken, beef, vegetables).

Mirror's technology revealed that the imagery was not the problem; instead, the brand's widespread recognition made it seem mass-market and inexpensive, leading to plateaued sales until this perception issue was addressed.

Ahuja describes their model's evolution as akin to a baby learning about the world, progressing from vision to language to body awareness and social intelligence.

Her deep interest in modeling the human brain stems from her academic background in neuroscience and computer science at the University of Toronto, where she was inspired by AI pioneer Geoffrey Hinton.

After university, Ahuja worked at Amazon Robotics, building robots that construct other robots, where she met her co-founders, Will Song and Thomson Yen.

Song has extensive experience in developing sales personalization engines, while Yen specialized in using deep learning to understand how AI agents comprehend human behavior.

Mirror Particle has successfully secured an angel round and is nearing the close of its first venture round. The company is also slated to compete in the prestigious Startup Battlefield 200 at TechCrunch Disrupt 2026 in San Francisco from October 13-15.

The startup's long-term aspiration is to establish itself as the "general layer for anticipating human behavior," eventually transitioning from broader population-level analyses to granular, individual-level insights.

Ahuja underscores the broader societal need, stating, "We just need a better model of humans if we’re going to work alongside AI and with each other."