20VC: How Simile Uses Unique Data to Predict Human Behavior

What if the biggest AI breakthroughs come not from bigger models, but from better data about how people actually behave? In this condensed summary of Harry Stebbings’ 20VC conversation with Jun Park, founder and CEO of Simile, you’ll hear how a Valentine’s Day simulation experiment grew into a company building human-behavior models, why “reflection” and memory matter in agent architecture, and how transactional, observational, and experimental data can unlock causality, counterfactuals, and better decision-making. Compared with the original episode, this version saves you time by distilling the key ideas in minutes. Learn why Jun believes the best AI companies need defensible data acquisition strategies, how enterprises are using simulation to replace months-long research with rapid testing, and why synthetic panels could reshape market research, A/B testing, and the future of AI. Listen now to get the key ideas in minutes.