Ditch the Chatbots
Ever wonder why brands seem to guess wrong about what we actually want? Startup after startup is trying to solve this by using LLMs to play-act as consumers, but San Francisco-based Mirror Particle says the status quo is fundamentally broken. CEO Abhivyakti Ahuja claims current AI models are like “bringing a super soaker to Niagara Falls,” noting they just don't have the depth to capture real human decision-making.
The 'World Model' Approach
Instead of just predicting text, Mirror Particle is building a foundation model from scratch. Their goal? A "world model" that tracks how people change over time, factoring in everything from pop culture and current events to social media and customer data. It’s not just about what we buy; it’s about the why—the motivations, constraints, and context behind our choices.
In one pilot, a pet food brand asked what pictures to put on a bag to boost sales. Mirror’s model basically said, “It’s not the imagery, bestie.” It realized the brand had a ‘cheap’ reputation problem that no amount of fancy packaging could fix. Real talk, the tech prioritizes “revealed behavior” (what we actually do) over those sus survey answers we all fill out without thinking.
The Master Plan
Ahuja, who has a background in neuroscience and robotics, built the startup alongside co-founders Will Song and Thomson Yen. They want to move from general population trends to individual-level insights. The long-term vision? Being the “general layer for anticipating human behavior.”
Why it matters
If we’re going to be working alongside AI long-term, we need models that actually get how humans think, rather than just regurgitating word patterns. Mirror Particle is aiming to be the decoder ring for why we make the choices we do, which is a major W for brands tired of throwing marketing spaghetti at the wall to see what sticks.
