Happy Monday.

We love it when a Unicorner story comes full circle: a former Unicorner writer is now a co-founder of a startup we’re covering!

Rehearsals is building digital twins of real people so companies can test a decision (like a price increase) before it ever reaches an actual customer. Keep reading to uncover the whole story.

In other news, we’re announcing another event to close out 2026: a founder dinner in SF with Mercury’s VP of Design.

Arek and Ethan 🦄

Rehearsals helps companies test business decisions with digital twins of real people. The Rehearsals team interviews participants and uses their answers and personal details to build a digital twin for each person. Research and marketing teams can then simulate how hundreds of potential customers might respond to a new product. A team considering a price increase, for example, can see how hundreds of digital twins might respond before changing the price for actual customers. The company says those results can be delivered in under an hour, giving companies a way to return to the same audience as new questions come up rather than recruiting a new research group for every study.

Check it out: runrehearsals.com

If you’ve already started building what’s next, your previous company may still have filings, tax obligations, and assets to resolve.

SimpleClosure manages the shutdown process from start to finish. Through AssetHub, founders can also explore selling eligible assets or continuing to license company data, allowing what they built to keep generating value.

If closing before year-end is the goal, now is the time to get started. Unicorner readers receive 10% off SimpleClosure dissolution services.

This is sponsored content.

Rehearsals sells subscriptions to businesses based on monthly interview volume. One credit covers one twin answering up to 20 questions, and all plans include unlimited seats. Higher-tier plans offer greater volume and more human-to-twin validation, which compares a digital twin's answers with those of the real person it represents. They also provide expanded governance features and integration options with additional support for customers on those plans.

When a company wants to research an audience that is not already represented in the panel, Rehearsals says it can recruit and interview new participants, typically within five days when feasible. It also offers a standard 30-day pilot, giving companies an opportunity to test the platform with their own business questions and target audience before committing to a subscription.

  • Backed by a16z speedrun, 25madison, Factorial Fund, and Plug and Play Ventures, alongside angel investors from OpenAI, Google DeepMind, and Scale AI

  • Customer count was doubling every two weeks in November 2026, per CEO Liam Bolling

  • Some Fortune 50 companies use Rehearsals daily, with individual teams running multiple full simulation sets in a day

  • Some customer simulations now involve 500 to 1,000 digital twins, allowing teams to test broad product questions

  • Launched Human Behavior Bench with Google DeepMind and Kaggle in October 2026, bringing together 1,244 U.S. adults whose responses generated 95,553 data points across 1,199 evaluation tasks

Before starting Rehearsals, Liam Bolling spent seven years at Google across Maps, Search, and Labs where he helped launch the first Gemini model and served as the founding product lead for Project Mariner, Google's experimental AI agent that could navigate websites and complete tasks directly in a browser. 

During his time working on Google Maps, he noticed that users were repeatedly moving map pins hundreds of miles, sometimes changing an address from one city to another. It looked like bad data, and Bolling spent six months trying to understand what was happening before realizing that many of these users had moved homes and were simply trying to update their own addresses, unaware that changing the pin could alter the location shown to other Google Maps users. The system recorded what they were doing, but it had no way of understanding why.

Bolling kept thinking about that as AI models grew better at interpreting human behavior. He became interested in whether models could do more than generate plausible responses and instead help companies understand how people might react to a decision. He eventually left Google to pursue the idea, bringing in his childhood friend David Schmetterling, who had led computer vision and engineering teams at Wayfair. Daniel Katz, a former Unicorner writer, later joined the founding team after working on AI coding agents at Databricks and building products at Microsoft.

The first version of Rehearsals was built around website analytics. The team tracked mouse movements and screen activity to identify where customers were getting stuck or abandoning a purchase, hoping to understand what was driving their decisions. But one research session exposed a problem with that approach. A participant had left a checkout page, and the team initially interpreted it as a sign that something in the shopping experience was not working. Her response was a surprise. Turns out, she’d simply needed to use the restroom. The analytics had captured exactly what she did, but completely missed why she did it.

From there, they began interviewing people directly, collecting the context behind their decisions and using those conversations to create digital twins. The shift took Rehearsals from observing customer behavior to simulating it, giving companies a way to explore how people might respond to decisions before putting them in front of actual customers.

Rehearsals initially took its product to engineering teams, positioning it as a way to identify friction on websites and improve user flows. But those teams had limited budgets for the problem. 

The founders began looking beyond the people responsible for building websites toward the teams whose work depended on understanding customers. Marketing and consumer insights teams already had budgets for research and were making decisions that depended on knowing how people would respond. By shifting its focus to those buyers, Rehearsals found a much stronger reason for companies to pay: beyond identifying problems with a website, Rehearsals could tell teams early how customers might respond to a decision.

Companies have far more questions about their customers than they have time or budget to investigate them.

Traditional research can take weeks to be organized and completed, which makes it difficult to justify every change a team is considering. Companies often reserve formal studies for major launches and rely on internal judgment for smaller decisions, even when getting customer feedback beforehand could prevent an expensive mistake. 

Rehearsals is trying to make research fast enough to use while a decision is still being shaped. A team considering a price increase, for example, can test how its audience might react before changing prices across its actual customer base.

General-purpose AI models have made it easy to generate a plausible response from a hypothetical customer, but a model prompted to act like a target demographic is still drawing on patterns in its training data, not the experiences of the direct people a company wants to reach. 

Rehearsals starts with interviews from human participants, using their responses and personal context to build digital twins that can be questioned. Building a digital twin is only useful, though, if it can carry that person's preferences into questions they have never been asked. Rehearsals reports 94% agreement when it asks a cohort of digital twins and the real people behind them the same questions, comparing their answers person by person. Rehearsals has also tested its predictions against what customers actually did. In one flight-pricing test, its twins predicted that 93% of travelers would still book after a 10% fare increase, close to the 91% who actually did. 

Rehearsals has also started putting its approach to the test against other AI models. In October 2026, it launched Human Behavior Bench with Google DeepMind and Kaggle, using 95,553 responses from 1,244 U.S. adults across 1,199 evaluation tasks. Its Engram 1.0 model scored 94.1 on distribution agreement, compared with scores between 76.2 and 80.6 for six general-purpose models.

Rehearsals builds digital twins of real people from interviews, letting companies test how audiences might react to business decisions before launch.

One of the questions asked how adults over 60 years old in the Midwest typically pay at grocery stores. 25% of participants said they use a phone or watch, and Engram predicted 23%. The other models predicted between 3% and 6%. It’s a small example, but a telling one: Engram came much closer to what this particular group of people said in the end, while the general-purpose models fell back on general assumptions. 

The score measures how closely Engram reproduces the way groups of people respond, rather than predicting the behavior of any one individual. The benchmark uses questions with predefined answer choices, so it is an early measure of how well the model captures audience preferences, not a test of whether a particular customer will make a specific purchase. For Rehearsals, the results provide encouraging evidence that its approach can capture differences between audiences that general-purpose models often miss.

Bolling has said Rehearsals has already reached more than 250 customers, including Fortune 50 companies. Some enterprise teams are already using the platform daily, with individual teams running multiple simulations involving hundreds of digital twins. Teams can return to the same digital twins as a product develops, without having to start a new research study each time.

Rehearsals has an opportunity to become part of the standard product process for many companies. Research conducted only before a major launch can tell a company whether an idea is likely to work, but by then, much of the time and money has already been spent developing it. Giving teams access to customer feedback while a product is still taking shape allows them to make changes earlier, before those decisions become expensive to reverse. With its growing customer base and early performance results, Rehearsals may succeed at becoming the go-to for just that.