Jun Sung Park

1. Key Themes

Human Behavior Simulation as the Next AI Frontier

Simile is building what Jun describes as a "foundation model of human behavior" — distinct from frontier LLMs whose goal is rational superintelligence. Simile explicitly wants models that are biased and fallible in the same ways humans are. "If we have a person make a mistake in this context, we want our models to make the same kind of mistake. We want our models to be biased in the same way humans are. In a way, we want to be a representation of people's values, preferences, and taste, sort of their subjective half of their brain." 00:10:28

The Real Customer Need Is Shaping the Future, Not Predicting It

Jun reframes the entire market research and prediction industry around causality rather than forecasting. Enterprises don't just want to know what will happen — they need to know how to change it. "No one really cares about prediction. No one really cares about what's going to happen in the future unless you're trying to predict the stock market. What people actually care about is they want to shape the future... What they want to know is how can we prevent it? What do we need to do now to change the future? And there, what you really need is causal mechanism." 00:11:41

Defensive Data Strategy as the Core Moat for AI Companies

Jun argues that the distinguishing feature of the best AI companies of this generation won't be algorithms or compute — it will be proprietary, hard-to-replicate data. "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... 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? When you see those opportunities, I would invest." 00:00:00 and 00:57:49

Synthetic Panels Will Exceed Human Panels in Market Research

Jun predicts synthetic panels — AI simulations of people — will not just replace human panels but expand the total ceiling of research possible. "What I see today in the market is actually quite broken. We have so many questions we want to ask about our market... You are looking at maybe 5% of those ideas get answered. The rest of the 95% we never bother experimenting with because we either don't have the ability to do them." 00:25:16 He predicts synthetic panels will be larger than the current human panel market within three years.

Simulation Will Follow the Same Compute Escalation as LLM Inference

Jun draws a direct analogy between the rise of "thinking models" spending heavily on inference tokens and where simulation is headed: massive compute spend justified by massive ROI. "I think there is a world in which in about two, three years, we're running a single simulation session. And that's going to take 10, $20 million to run a single session. But it's going to be so valuable that people will pay $100 million for it." 00:48:59

The World as Ground Truth Creates a Uniquely Powerful Learning Flywheel

Unlike other AI tasks where feedback loops can be unclear, simulation has the entire real world as its validator. "Every single day, we can be generating tens of thousands of hypotheses. Each hypothesis is mapped onto an end statement... A month goes by, we generated a million hypotheses, X percentage of them came true. This is the best way to learn about the world." 00:20:03

Fortune 500 Enterprise Sales Can Close in Weeks When Pain is Acute Enough

Jun fundamentally challenges conventional wisdom about slow enterprise sales cycles. The acute pain enterprises feel in making gut-based, data-poor decisions caused them to move at startup speed. "We actually saw some of the largest customers in the world move at lightning speed for enterprise, where we saw them close deals within three months." 00:29:26 A key tactic was replaying a study that previously cost 3–6 months from a large consulting firm — and producing the same answer in two minutes on the first call.

Simulation as a Representational Layer for Society

Jun's long-term vision goes well beyond enterprise software: simulation becomes the infrastructure through which every individual's perspectives and values are represented at scale in collective decision-making, from corporate strategy to government policy to climate action. "I actually don't think this is a limitation we have to suffer through in the future. I think there's a world in which we can truly create a layer that becomes a representational layer of our society and of our collective intelligence." 00:54:38

The GPU Analogy for Collective Intelligence

Jun offers a structural insight about where AI architecture is heading: LLMs are the "CPU" of intelligence — one powerful centralized reasoner. Simulation represents the "GPU" equivalent — massively parallel, diverse, collectively emergent intelligence. "What I see coming and what I think simulation as a field can offer is the GPU of intelligence unit... What we care about is creating models that are as smart as we are... when they come together as a large collective, the emerging phenomena that we're able to draw out is some of the most wonderful thing that we can see in our world." 00:51:44

2. Contrarian Perspectives

Observational Behavioral Data Is Overrated for Business Decisions

Most AI companies chase transactional and behavioral data as their crown jewel. Jun argues this data is good only for correlation and prediction — which isn't actually what businesses want. What's truly valuable is randomized control trial and A/B test data that reveals causal mechanisms. "A lot of observational behavior data... what they're amazing at is actually helping you create a correlation of the observation and what could happen in the future... But my take here after interacting with so many of our customers and also being in research, no one really cares about prediction." 00:11:13 Simile specifically collects RCT and A/B test data rather than relying on observational web data.

Researchers Are Better Entrepreneurs When Driven by Impact, Not Problems

The standard assumption is that the best academic founders are deeply specialized and problem-obsessed. Jun inverts this: problem obsession makes for science projects, not companies. "The thing I actually would look for is, are they married to a problem or are they married to impact? Sometimes researchers are very much focused on a problem and something about that problem fascinates them. But oftentimes it's just not a good company... You want to find researchers who are in that category" — driven by reach and revenue. 00:41:33

Asking People to Pay You Is the Best Form of Research Feedback

Against the conventional startup wisdom of rapid free user acquisition, Jun's mentor Pat Hanrahan's advice at Tableau was to charge early. "The best way to get feedback is to actually ask people to pay you. That was their core philosophy at Tableau." 00:23:12 Jun applies this directly: enterprise pricing creates the highest-quality signal loop for improving simulation accuracy.

Short-Term Paranoia and Long-Term Religious Belief Must Coexist in Founders — and Can't Be Separated

Most investors and operators treat optimism and paranoia as a tradeoff. Jun argues the best builders hold both simultaneously and that the paranoia is actually causally responsible for the success. "If you are short-term paranoid, then you're likely going to be very pessimistic about your future... If you're religious, you have the opposite problem, which is you're complacent... Balancing those two needs somebody who is broken in some ways." 00:36:42

Neo Labs Without Clear Impact Vision Are Overheated Science Projects

At a time when new AI labs attract enormous capital, Jun is skeptical of the category as a whole. "I do think neo labs, without a clear vision for how they're going to impact the world, I do genuinely think there is some risk that they will turn out to be interesting research projects, but not a viable company." 00:57:13

3. Companies Identified

Simile

AI simulation company building a foundation model of human behavior for enterprises. Jun Sung Park is CEO and founder. Raised $300M total — $100M round ~five months ago, followed by a $200M round led by Green Oaks, with Index Ventures (Shardul Shah) as insider and lead of prior round. Mentioned throughout as the subject of the conversation. "We are a company that is creating foundation model of human behavior that can then be used to create simulations of individuals, simulation of subpopulations, and down the line, the simulation of the entire ecosystem and even the market." 00:10:04

CVS

Fortune 500 pharmacy and health company. Mentioned as a named Simile customer that moved at "lightning speed" for enterprise, closing a deal in roughly three months. Jun specifically named their VP of Insights, Sri, as an exceptional counterpart. "The pain they were feeling in their day-to-day work was so real... when they realized that there is or there could be an answer in this market for addressing some of those pains, they are ready to drop everything and try us out." 00:28:56

Index Ventures

Venture capital firm. Shardul Shah leads Simile's prior round and is described by Harry as "one of the best investors of the last decade" who has backed companies including Wiz. "Shardul has sort of this comment that he every once in a while makes where he's seen some of the fastest growing market. And his track record does show that he truly has seen different markets. He has quite never seen this kind of traction, this kind of pull." 00:42:21

Green Oaks Capital

Growth equity firm. Neil Meadows led Simile's $200M round. Jun describes Green Oaks as having done deep diligence on the simulation market before being approached. "It turns out his team actually has been looking deeply into this market and all the players, how the market is going, and were actually prepared to make the investment." 00:43:17

Tableau

Data visualization company co-founded by Pat Hanrahan, Jun's Stanford colleague. Mentioned for the principle that charging customers early generates the best feedback. "The best way to get feedback is to actually ask people to pay you. That was their core philosophy at Tableau." 00:23:12

OpenAI

Frontier AI lab. Mentioned as evidence that ambitious AI visions can go from being ridiculed to near-trillion-dollar businesses. "These are people who have literally seen OpenAI being the laughingstock in Silicon Valley to becoming a nearly trillion dollar business." 00:39:04

Anthropic

Frontier AI lab. Mentioned alongside OpenAI as validation that deep research vision companies can achieve massive scale in five years. 00:39:04

Edged

Chip/hardware AI company that recently came out of stealth. Jun identified it as one of the most exciting teams in the chip/inference layer space. "The recently Edged came out of their stealth, quite bullish on their team. I think they're going to be exciting." 00:58:16

Crosby

AI law firm mentioned as a sponsor. Described as combining AI with elite human attorneys from top-10 law firms and engineers from Ramp and Stripe, turning red lines around in under four hours. Customers include Cognition, Ramp, and Clay. 00:01:43

Kalshi

Prediction market company. Mentioned in the episode title in the context of whether Simile's simulation capabilities could disrupt or compete with prediction markets. 00:31:58

Polymarket

Prediction market platform. Also mentioned in the episode title in the same competitive/displacement context as Kalshi. 00:31:58

4. People Identified

Founder and CEO of Simile. Former Stanford PhD researcher and lead of the Smallville generative agent simulation experiment (2023), one of the earliest multi-agent AI architectures. No prior research experience entering academia — got his foot in through cold outreach and the kindness of a Stanford professor. "In the past six years, all my core team members never left. Then we all moved from one project to the next, to the next together." 00:40:07

Shardul Shah

Partner at Index Ventures. Described by Harry as one of the best investors of the last decade, with a track record including Wiz. Introduced Harry to Jun and led Simile's prior funding round. Led the preemptive insider round. "Shardul Shah, one of the best investors of the last decade at Index Ventures, emailed me late one night... and said that he had a company that he was leading around in and that he'd never seen sales and growth like it before, even though he'd been in companies like Wiz and many other incredible Decacorns." 00:00:28

Michael Bernstein

Co-founder of Simile and Stanford professor. Co-authored the Smallville simulation paper and is a leader in human-centered AI. Notably, was Jun's doctoral advisor and followed his former student into the company. "Michael was one of the co-authors of the ImageNet that really kickstarted the AI revolution, and he's been a leader in human-centered AI." 00:26:44 Jun also cites him as an archetype of the "common denominator of success" — reinventing himself from crowdsourcing to generative AI agents across his career. 00:34:24

Percy Liang

Co-founder of Simile and Stanford professor. Credited with coining the term "foundation model." Also Jun's doctoral advisor who joined Simile. "Percy was the person who literally coined the term foundation model." 00:26:44

Lainey Allen

Co-founder of Simile leading go-to-market. Introduced to Jun by Mike Volpe. Cited by Jun as the embodiment of short-term paranoia paired with long-term religious belief — an archetype Jun considers essential for company building. "She's paranoid... unless we put everything on our table today and do everything possible, we'll lose. But long term, she's religious. This is somebody who fundamentally believes the world is stacked for her." 00:36:42

Pat Hanrahan

Co-founder of Tableau, Stanford graphics professor, Turing Award winner. Jun's office neighbor at Stanford. Gave the advice that shaped Simile's enterprise-first go-to-market: charge customers for the best feedback. "The best way to get feedback is to actually ask people to pay you. That was their core philosophy at Tableau." 00:23:12

Neil Meadows

Partner at Green Oaks Capital. Led Simile's $200M round. Described by Jun as deeply analytical but also highly intuitive — the rare dual-superpower profile Jun looks for. "Deeply analytical, but he's very intuitive. And I think that's how he makes investment that happens to be very successful." 00:36:15

Mike Volpe

Runs his own VC firm (formerly associated with Aster). Led Simile's seed round alongside Shardul Shah. Introduced Jun to Lainey Allen and acted as an operating mentor for Jun as a first-time CEO. "Mike actually introduced me to Lainey, who ended up becoming instrumental as I thought about the business." 00:46:07

Sri (last name not mentioned)

VP of Insights at CVS. Named specifically as an exceptional enterprise champion: "Extremely forward-looking, extremely ambitious, extremely hardworking, amazing counterpart, two vision-like simile." 00:28:56

Mary Waters

Theory professor at Stanford. Responded to Jun's cold message when he had no research background, spent a full morning advising him, and connected him to his first research collaborators — a pivotal act of generosity that launched his research career. "She very graciously spent a full morning with me and just talking me through how I should think about AI space or how I should think about research." 00:59:41

5. Operating Insights

Replaying a Competitor's Past Work in the First Customer Call Closes Deals

Simile's most effective sales motion was not pitching future capabilities — it was taking findings from studies clients had already paid large consulting firms to produce (3–6 month engagements) and replicating those findings on the fly in two minutes during the first call. This collapses the "prove it" objection instantly. "One of the ways we actually got some of our first customers was in the first call, they actually had a finding from large consulting companies, and they basically queried our system... And we predicted the outcome of studies that took three to six months, but just within two minutes." 00:30:01

Hire for Two Uncorrelated Superpowers, Not One Strong Superpower

Jun's talent filter is not looking for depth in a domain but for people whose two strongest capabilities are logically contradictory — because that rarity signals an unusual mind. "Where I found things to be particularly compelling is if people have two superpowers, that's really contradictory." 00:35:23 His specific example: the world-class CMO is both rigorously data-scientific and wildly creative — two opposing mental orientations in one person.

The "Common Denominator" Test for Leadership Candidates

Rather than evaluating past roles in isolation, Jun maps career trajectories and asks whether this person was consistently the reason things succeeded across very different contexts. "Were they the common denominator? If the answer is yes, then what that suggests is a couple of things. That they have extreme degree of ownership... It also shows the ability to reinvent themselves." 00:34:24

Prevention Value Cases Close Faster Than Optimization Value Cases

When pitching AI tools to large enterprises, Jun found that the clearest and fastest-closing ROI framing is preventing catastrophic decisions rather than optimizing good ones. "One of the core premises and one of the ways that our customers are actually finding value at Simile is actually avoiding really damaging decisions that could have costed them hundreds of millions of dollars... That could have been a total disaster had we run that. That would have costed us half a billion dollars. We ran simulation and that prevented it. That's a no-brainer." 00:31:27

Build Retention by Letting People Express Their Superpower at Maximum Degree

In a market where researcher compensation reaches tens of millions at large labs, Jun argues that retention at startups comes not from matching salaries but from constructing a platform that allows elite researchers to operate at the full scale of their capability. "I actually do view the role of leadership to be that of obviously hiring amazing people, but also providing a platform where individual members can express their superpower to their maximum degree." 00:39:47

6. Overlooked Insights

Simulation's Inference Cost Curve Points Toward $100M Single-Session Contracts — A New Asset Class for Compute

Everyone in AI is watching the inference cost escalation from "thinking models" as a problem to solve. Jun briefly inverts this: the same dynamic that makes long-chain reasoning expensive is precisely what will make simulation enormously valuable — and therefore create a new category of ultra-high-value single-transaction compute contracts. This isn't incremental SaaS pricing; it implies that a single simulation session could become a standalone financial event comparable to a large M&A advisory fee. "I think there is a world in which in about two, three years, we're running a single simulation session. And that's going to take 10, $20 million to run a single session. But it's going to be so valuable that people will pay $100 million for it." 00:48:59 No one in the conversation pushed on the investment implication: this creates a new category of AI services company with deal economics resembling investment banking, not software.

Simile's Quant Talent Acquisition Signals an Internal Hedge Fund Is a Live Option

Jun mentioned almost in passing that quant professionals are joining Simile — not because of market research applications but because they see Simile's human behavior modeling as directly relevant to financial markets. Harry floated the hedge fund idea casually, Jun agreed it was "an interesting idea" and noted they already have quants in-house. "Some of the members who have joined actually do have more quants background... they actually see the vision of Simile very much well aligned with their passion and interest, which is to model the world. But down the line, I think it's actually an interesting idea." 00:53:37 This was treated as a throwaway exchange, but it is structurally significant: if Simile's human behavior model becomes accurate enough to predict market-moving consumer behavior ahead of earnings, the company faces either the opportunity or obligation to spin up a proprietary fund — and the talent is already accumulating organically.