BREAKING: Stanford PhD Built… | Sourcery Newsletter Summary

1. Key Themes

Theme: Human behavior is predictable enough to simulate commercially, and that is the company's foundation

Digital twins built from interviews, not demographics, hit 85% of human self-consistency

The core scientific claim is that agents built from qualitative interviews can replicate real people's answers nearly as well as those people replicate their own.

"The agents replicated participants' General Social Survey responses 85% as accurately as participants replicated their own answers 2 weeks later. That figure came from 68.85% agent accuracy divided by 81.25% human self-consistency."

Note the nuance: "85%" is a ratio against human test-retest reliability, not raw accuracy. The company itself frames it as the foundation:

"At the end of 2024, for the first time really in the field, we could actually show that we can predict people's behavior with 85%-plus accuracy with these agents. That's ultimately what gave this company foundation for today's technology."

Interview-based agents also reduced bias

"The interview-based architecture also reduced accuracy biases across racial and ideological groups compared to agents given only demographic descriptions."

Theme: Simulation compresses market research from years and millions of dollars to a query

Collapsing the cost and time of traditional research

"The traditional analysis that researchers had to run today, that would've taken them sometimes hundreds of thousands of dollars, sometimes millions of dollars over many years, we can run them in simulation basically almost instantly. They are just a query away."

Early enterprise adoption shows concrete use cases

"Deloitte uses the platform as a substitute for focus groups, and Wealthfront reported a 15x expansion in qualitative research scope. Banco Itaú in Brazil uses simulations to test policies and messaging."

CVS uses it to screen research before running pilots:

"CVS uses the simulations to screen research questions, review customer journeys and test approaches to medication adherence before running real-world studies or pilots."

Theme: Proprietary human data and panel access are the moat

Real-people panels feed and validate the models

"Today, Simile have partnership with over 40 million real people across the panel partnerships that we have. This is across the globe, and at any given moment, we're talking to tens of thousands, if not hundreds of thousands of real people every single week."

Customers contribute large consented datasets

"In a case study published by Simile, CVS's agents were built using 2.9 million consented responses from more than 400,000 participants across over 200 behavioral scenarios."

Data quality controls are part of the product

"Simile controls for panel bias through spacing between survey invitations to reduce fatigue, studies measuring response quality across new and repeat panelists, and attention checks such as simple arithmetic questions."

Theme: The market is rewarding the category fast

Rapid funding and revenue ramp

"The $200M Series B followed on July 30, led by Greenoaks... The round valued the company at $2B and brought total funding to $300M+." "The company reported fivefold revenue growth in the 5 months after its Series A."

Elite academic team plus strategic investors

"Bernstein, a Stanford human-computer interaction professor, is Chief Product Officer, and Liang, director of Stanford's Center for Research on Foundation Models, is Chief Scientist."

CVS Health Ventures participated in the Series B, aligning a customer with the cap table.

Theme: Simulation is early, with long-term upside in policy and societal modeling

Park positions the category at an early inflection point

"Simulation is basically in its GPT-4, GPT-4.5 era, where the technology's here to make impact, but this technology is changing and evolving every single month."

Park expects simulation to reach the maturity of today's coding agent market in about 1.5 to 2 years.

Next frontiers: coordination problems like climate and AGI policy

"Climate change is hard, not just because of the physics of it, but because it's about coordination. There are so many actors with competing incentives who have their own things they're trying to optimize for."

2. Contrarian Perspectives

Surveys measure stated intent, which diverges systematically from behavior

Most market research treats survey answers as signal. Simile argues they follow a different distribution than real behavior, and that the gap is the core modeling problem.

"Survey answers about going to the gym this week tend to follow a normal distribution. Actual gym attendance follows something closer to an exponential curve with a sharp drop-off, because the activation energy to act is higher than people expect."

The reasoning is that real decisions involve opportunity cost, which abstract surveys omit:

"This is where people have this optionality, so now they have to think. They have to think about the opportunity cost of taking that option. So that's actually what really governs people's behavior, that, yes, I can go exercise this week, but if I don't, I can sleep for one more hour, or I can take one more meeting."

Implication: simulating the environment (competing products, alternatives) alongside the person may beat traditional polling.

Synthetic respondents can be trusted for high-stakes decisions if they carry a calibrated confidence score

The common objection to AI-simulated respondents is hallucination. Simile's answer is a jointly trained accuracy model that scores each output.

"Its job is to output 0 to 100 range probability on how confident we are in this particular response on the basis of all the data that we use to train our model and the question that we got." "If confidence is higher than 90% or usually 85%, our customers have found that to be robust evidence that can actually be used to make very high stake decisions."

Causal data (RCTs) matters more than forecasting

"Simile weights its training data toward randomized controlled trials... Customers ask less about forecasts and more about which actions change an outcome, which requires causal data."

This is a differentiated bet: the product is about "what lever do I pull," not "what will happen."

3. Companies Identified

Simile

Description: AI startup building foundation models of human behavior and digital twins of real people; $2B valuation, $300M+ raised.
Why mentioned: Subject of the article and interview.
Quote: "Simile is a company that is creating foundation model of human behavior that basically can take as input a description of a person or a group of people, and then simulate what these people might do in different context."

CVS Health

Description: Healthcare and retail giant; Simile customer and Series B participant via CVS Health Ventures.
Why mentioned: Flagship case study.
Quote: "CVS uses Simile to study consumers, store layout, messaging and decor."

Deloitte

Description: Professional services firm.
Why mentioned: Uses Simile to replace focus groups.
Quote: "Deloitte uses the platform as a substitute for focus groups."

Wealthfront

Description: Fintech/investing platform.
Why mentioned: Customer reporting expanded research scope.
Quote: "Wealthfront reported a 15x expansion in qualitative research scope."

Banco Itaú

Description: Brazilian bank.
Why mentioned: International customer using simulation for policy/messaging tests.
Quote: "Banco Itaú in Brazil uses simulations to test policies and messaging."

Gallup

Description: Polling and analytics firm.
Why mentioned: Partner and independent validator, and a source of panel data.
Quote: "Gallup is exploring simulated responses through a partnership with Simile and is independently validating the method; the 2 are developing new agent banks across different topic areas."

Index Ventures

Description: Venture firm; led the Series A (partner Shardul Shah).
Why mentioned: Lead investor and interview host site.
Quote: "Simile emerged from stealth with a $100 million Series A on February 12, 2026, led by Index Ventures partner Shardul Shah."

Greenoaks

Description: Growth investor.
Why mentioned: Led the $200M Series B at a $2B valuation.
Quote: "The $200M Series B followed on July 30, led by Greenoaks."

Other investors: Bain Capital Ventures, Hanabi Capital, A, Definition*

Description: Participating investors across rounds.
Why mentioned: Cap table detail.
Quote: "with participation from Bain Capital Ventures, Hanabi Capital, A, and individual investors."*

DoorDash

Description: Delivery platform.
Why mentioned: Source of Simile's head of engineering.
Quote: "The engineering organization is led by a head of engineering who joined from DoorDash."

4. People Identified

Joon Sung Park

Description: Co-founder & CEO of Simile; Stanford PhD.
Why mentioned: Interview subject and first author of the Generative Agents paper.
Quote: "Park completed his Stanford PhD advised by Bernstein and Liang and is the first author of the Generative Agents paper."

Michael Bernstein

Description: Stanford HCI professor; Simile co-founder and Chief Product Officer.
Why mentioned: Co-founder and Park's PhD advisor.
Quote: "Bernstein, a Stanford human-computer interaction professor, is Chief Product Officer."

Percy Liang

Description: Director of Stanford's Center for Research on Foundation Models; Chief Scientist.
Why mentioned: Co-founder and technical leader.
Quote: "Liang, director of Stanford's Center for Research on Foundation Models, is Chief Scientist."

Lainie Yallen

Description: Co-founder.
Why mentioned: Leads go-to-market and deployment.
Quote: "Go-to-market and deployment are led by Yallen."

Fei-Fei Li and Andrej Karpathy

Description: AI luminaries; individual investors.
Why mentioned: Signal of technical credibility.
Quote: "backed by Fei-Fei Li, Andrej Karpathy, Greenoaks, & Index Ventures."

Shardul Shah

Description: Index Ventures partner.
Why mentioned: Led the Series A.
Quote: "led by Index Ventures partner Shardul Shah."

Thomas Schelling

Description: Economist known for the segregation model.
Why mentioned: Park's research influence.
Quote: "Park cites Thomas Schelling, whose segregation model applied simulation to the defining social question of his generation, as a research influence."

Molly O'Shea

Description: Newsletter author and interviewer.
Why mentioned: Ran the interview and tested the simulation on herself.
Quote: "Before our interview, Joon simulated our conversation using AI, and the model correctly anticipated one of my opening questions."

5. Operating Insights

Build a confidence score into the product and coach customers on when to trust it

Simile treats trust as a product feature and a change-management problem.

"Park compares the score to a p-value in inferential statistics... Simile also runs change management with customers on when to use results, where more data is needed and how to validate outputs."

Use a forward-deployed model to drive enterprise adoption of novel technology

"The company runs a forward deployed model with customers, based on the success of forward deployed companies in agentic coding."

Stack your team with the research lineage that created the technology

"Roughly 20% are former lab mates and academic relatives from Liang's and Bernstein's groups... modeling, evals and simulation science make up 20% to 25% of the company."

A small, specialized team (about 100 people) plus social-science methodologists supports rapid iteration.

6. Overlooked Insights

Panel supply can be extended on demand

When a customer needs a population outside existing banks, Simile sources new participants quickly, turning data collection into a competitive speed advantage.

"When a customer needs a population outside those banks, Simile collects new data and trains the model to cover it, with new panel members typically sourced within 24 hours."

Consent and ethics guardrails are built into the data pipeline

"Participants who cannot give informed consent, including minors, are excluded, and every panelist can withdraw consent."

This matters for enterprise and regulatory acceptance, particularly in healthcare (CVS) and finance (Wealthfront, Banco Itaú).