Noah Shinn raised his first $100 million before he turned 24. By September, Instinct had raised $350 million, the last slice of it at a $2.5 billion valuation. On September 28, the company announced $1 billion more, at $10 billion. The check is not the story. The multiple is.
Instinct, the company Shinn founded in 2025, builds a personal AI agent that works over text messages and phone calls. You message a number. It books the reservation, negotiates the bill, cancels the subscription, orders the groceries, using its own phone line and its own computer.
The company has not disclosed a dollar of revenue. Shinn has told reporters he does not want to charge users at all.
Twenty-three, and a $10 billion mark
Shinn founded Instinct in 2025 after a stint at Sierra, the customer-service AI company run by Salesforce co-CEO Bret Taylor. He was 23 when the product shipped. It skipped the App Store. It launched as an invitation-only web page passed around venture networks, and within weeks it passed 100,000 users.
We're building Instinct to be the best personal agent that can handle the deeply personal nuances of everyday life.— Noah Shinn, founder, Instinct
That is the pitch. The round is the evidence investors bought it. Sequoia Capital, Benchmark Capital and Coatue put money into a company with no published retention curve, no growth chart and no revenue line.
The demand looks real: more than 100,000 users on an invite-only list, with the service short of compute.
The strongest competitor charges nothing. Meta's Muse is free.
What the agent actually does
An AI agent is not a chatbot that answers questions. It takes actions. The agent replies to messages, books travel, makes restaurant reservations, pays bills, cancels subscriptions, runs research and orders groceries. The agent places the call itself and types into the same web forms a person would, from a phone number and a machine that belong to the company.
Over the summer the company shipped features that push it past simple task-running. Instinct Concierge handles high-touch requests: the restaurant that never has a table, the call that needs a human voice. A Trusted Person Network lets two users' agents coordinate directly and exchange files. Location-aware suggestions surface inside iMessage.
Trust is the open question. On September 24, a user reported that the assistant had surfaced what looked like a stranger's private data. Shinn said the agent invented the detail, a hallucination rather than a real leak. The company's answer is a feature it calls an active detective system, which screens responses for subtle fabrications before the agent sends or executes them. Every serious agent lab is racing the same problem. An assistant with access to your messages, your cards and your calendar cannot be wrong quietly.
The compute wall behind the invite list
The invitation list is not only marketing. The company has run into compute limits. During peak load, users were told the service was at capacity and that responses might be slower. A personal agent that places calls and browses the web burns far more compute per task than a chat reply. Every reservation is a chain of decisions, not a paragraph of text.
Part of the $1 billion buys hardware. The rest buys time in a market where GPU capacity is the scarce input and the largest buyers are the same hyperscalers building rival agents.
Sierra, Shinn's old employer, sells agents to companies. Instinct sells one to individuals. The margin math is different at every level.
Meta gives its agent away
Two weeks before Instinct's round closed, Meta released Muse, a free personal assistant that rose to the top of the U.S. app stores. Meta has spent the past year assembling what it calls a personal superintelligence, an internal effort insiders describe as an assistant for daily errands. It ships to billions of users at zero price.
It has no mobile app. Its whole surface is a phone number. That could be a distribution advantage or a ceiling, depending on how you read it.
As we wrote in September, Harvey's $15.5 billion round put a price on legal AI agents. The pattern repeats here: investors are funding the interface and the habit, not the current revenue.
What the $10 billion is really for
The tasks the agent performs sit next to spending decisions. Reservations, grocery orders, bill payments, subscription cancellations and product research all touch commerce or money management. An agent that handles those steps owns a gate people pass through many times a week. The $10 billion prices the option on that transaction layer. It does not price this year's revenue, which does not exist.
Dealroom puts the round in the 99th percentile of U.S. enterprise software Series C deals, measured against a sample of 1,750. That is a narrow band of companies. Most of them share Instinct's shape: large raises, ten-figure marks, thin public disclosure.
This funding helps us bring Instinct to more people and continue building the future of personal AI. It's an exciting, creative time, and we're just getting started.— Noah Shinn, founder, Instinct
The risks are stacked and visible. Meta can bundle a free rival into its existing apps. Compute costs scale with usage, and the product is free. A single trust failure, a fabricated message or a wrong payment, can end the habit faster than any competitor.
Shinn's bet is that the interface wins first and the money follows. His investors just paid $10 billion for that ordering. The next twelve months will say whether the instinct was right.