## The Company That Clones Your Customers The premise sounds like science fiction: build a digital copy of a real person, accurate enough that companies can test decisions against it before making them in the real world. Simile makes this literal. Founded by Joon Sung Park, Michael Bernstein, Percy Liang, and Lainie Yallen — all with Stanford ties — the company builds AI models trained on individual-level data: interviews, purchasing behavior, psychological profiles, and behavioral science research. The result is a "digital twin" that simulates how a specific person would respond to a product change, a marketing message, or a business decision. Park's academic pedigree is directly relevant. His 2023 paper "Generative Agents" — which placed 25 autonomous AI characters in a simulated town called Smallville — won Best Paper at ACM UIST and became one of the most cited AI papers of the year. It demonstrated that language models could simulate believable human behavior with surprising fidelity. Simile is the commercial application of that research. ## How It Works Simile partners with real people to build high-fidelity models of how they live and make decisions. The training data is granular: interviews, transaction histories, behavioral assessments, and contextual data. The output is a model that can answer the question: "How would this specific person react to X?" The applications are immediately valuable to any company that currently relies on focus groups, surveys, or gut instinct: **Retail:** CVS has used Simile for five months to simulate customer reactions — influencing decisions about what items to stock and how to display them. **Investor Relations:** Companies rehearse earnings calls against digital twins of analysts, testing how specific questions land and refining their messaging before the real thing. **Litigation:** Legal teams model how judges or jurors might respond to different arguments, allowing them to optimize trial strategy. **Policy:** Organizations test policy changes against simulated populations before implementation. ## The Funding Story Simile raised $100M in a Series A led by Index Ventures, with participation from Bain Capital Ventures, A* and Hanabi Capital. The round is notable for its angel investors: Fei-Fei Li (Stanford HAI co-director, World Labs founder) and Andrej Karpathy (former Tesla AI lead, former OpenAI) both backed the company. When Li and Karpathy both invest in the same company, it's worth paying attention. Li's involvement signals academic credibility in human-AI interaction. Karpathy's signals technical conviction in the approach. The company emerged from stealth after seven months of building, having already signed enterprise customers. That's fast for a company handling sensitive behavioral data. ## The Provocation Simile raises the question that every AI company eventually confronts: where is the line between useful simulation and uncomfortable surveillance? Building accurate behavioral models of individuals requires intimate data. The company's value proposition — predicting how real people will react — is precisely what makes it ethically complex. The difference between "market research" and "surveillance" is often just a matter of consent and transparency. Simile's approach of partnering directly with real people (rather than scraping their data) is an intentional design choice that addresses some of these concerns. But as the models get more accurate, the ethical questions get harder. ## Why It Matters Simile represents a category shift in how companies understand their customers. Focus groups are small and biased. Surveys measure what people say, not what they do. A/B tests are slow and expensive. Digital twins, if accurate, compress all of that into software. The market for customer insight and decision support is enormous. If Simile's approach works at scale, every Fortune 500 company will want one. The question is whether the accuracy holds outside controlled environments — and whether the ethical framework can scale as fast as the technology. **Watch for:** Expansion beyond enterprise into political polling and public policy simulation. That's where the technology gets both most valuable and most controversial.