20VC: The Best AI Companies Have Unique Data Acquisition Strategies | Will Simile Kill Kalshi, Polymarkets and NASDAQ | How to Sign Fortune 500 Companies As Customers in Weeks with Joon Sung Park, Simile
Summary
This episode centers on Simile’s vision for AI-driven human behavior simulation and why the company believes defensible data, not just better models, will determine the winners in AI. Joon Sung Park explains how memory, planning, and reflection help agents behave more realistically, and how early simulations produced surprisingly human social dynamics. A major theme is that prediction is only valuable when it helps customers change outcomes, which is why Simile emphasizes counterfactuals, causal mechanisms, and randomized controlled trials. The conversation also covers how Simile wins enterprise customers quickly, often closing Fortune 500 deals in just weeks or months by delivering faster and more actionable decision support. Finally, Park explores the broader implications of advanced simulation, including its potential to reshape prediction markets, stock markets, and even personal decisions like dating and marriage.
Key Takeaways
- 1AI companies need a defensible data acquisition strategy, not just strong models.
- 2Simulation becomes valuable when it changes decisions, not when it merely predicts them.
- 3Realistic human simulation depends on memory, planning, and reflection in agent design.
- 4Enterprise buyers will move quickly if simulation compresses slow, high-stakes experimentation into fast, credible answers.
- 5Park sees simulation as a future infrastructure layer for intelligence, potentially complementing frontier models like a 'GPU of intelligence.'
- 6Advanced simulation could disrupt markets and even personal life decisions by exposing better representations of behavior.
Notable Quotes
"My fundamental thesis here is for AI companies of this generation, you need to have an interesting data strategy that's going to be defensible."
"We collect transaction data, we collect observational data, we also partner with our customers, our vendors to collect some of this data."
"We actually showed that we can actually predict people's behaviors and attitudes 85% as accurately as people replicate their own."
"In two or three years we're running a single simulation session that's going to take 10-20 million dollars to run. A single session, but it's going to be so valuable that people pay $100 million for it."
""I think there is a world in which we can truly create a layer that becomes a representation layer of our society and of our collective intelligence.""
""I think one of this could actually be a stock market.""
""If you could date 100 people at the same time for the first date... it would be much more effective at finding the one for you who could pass through to the next stage.""
""You have to have interesting data strategy. Do you have access to data that no one else has access to? Do you know how to collect data that is very hard to collect?""
Episode questions
What makes Simile different from a standard survey or market research company?
Simile says its core product is a generalizable model of people, not just a better survey interface. It aims to simulate populations and counterfactual behavior, going beyond questionnaires to explain why outcomes happen and how to change them.
Why does the company focus so much on memory, planning, and reflection?
These mechanisms let agents maintain continuity across interactions and build higher-level interpretations from many events. Park says they were necessary to avoid repetitive behavior and to produce personalities that can make sense of long-term experience.
Why do enterprise customers pay quickly for Simile?
Because they are already spending time and money on slow experimentation and they want evidence to guide decisions faster. The company can compress studies that might take months into minutes, which makes the value proposition obvious.
Why did Simile raise $200 million when it had already raised $100 million recently?
Park says the main reason was compute and data scale. The team believed they could accelerate progress by increasing inputs dramatically, and insiders plus new investors moved quickly because of the traction and technical progress.