Personal AI company Instinct raises $1B Series C at $10B valuation

On September 28, personal AI startup Instinct announced a $1 billion Series C at a $10 billion valuation, with investors including Sequoia Capital, Benchmark Capital, and Coatue Management, among others. Just a month earlier, it had closed a $250 million Series B at a $2.5 billion valuation. A valuation jumping from $2.5 billion to $10 billion in one month is the most direct footnote to this deal.

What contrasts with the scale of the funding is the size of the company: Instinct has only 14 employees, founder Noah Xin is under 25, and the company has no app. Noah Xin founded Instinct in San Francisco in 2025, with Spear Street Technology as the operating entity. On the day the funding news was announced, investor Patrick O'Shaughnessy published a long interview with Noah Xin—the latter's first public, systematic account of Instinct's product logic, commercialization thinking, and competitive judgment.

Instinct did not build a new application that users have to relearn. Users contact it much the way they would contact a human assistant: they can text, call, or write a request email, and it handles the rest on its own. This form is close to the "concierge service" once offered by hotels and credit card companies: the user simply states a need, and the person behind the scenes books the restaurant, arranges the itinerary, and contacts the merchant. What Instinct wants to do is replace the human concierge on the other end of the line with an AI agent authorized by the user.

Recently, Instinct launched a feature named directly after the "Concierge," specifically to contact merchants on users' behalf that do not support online booking and still require phone communication. Behind this is a shift in product logic: over the past two decades, the internet has continually turned services into self-service tools, with users searching, comparing, filling out forms, and clicking buttons themselves; Instinct wants to pull these steps back into the background, leaving users responsible only for stating the outcome they want.

According to Noah, about 50% of transaction volume on the Instinct platform currently comes from travel. Travel has become one of the most important scenarios because it is naturally a chain of interdependent tasks: if the user has already authorized it, Instinct can know where the user is right now, which airline they usually fly, whether they prefer a window or aisle seat, and which credit card they habitually use, then arrange flights, hotels, and transfers accordingly and write the itinerary into the calendar.

Currently, Instinct has not disclosed user scale or operating data. According to Noah, Instinct was initially opened only to about 200 friends and family, each with just 5 invite codes; growth gradually accelerated and has recently reached roughly 10% to 11% day-over-day, with marketing spend near zero. By his account, these relatively few users have already pushed the platform's annualized gross merchandise value (GMV) to near or above $1 billion. Scarce invite codes were at one point resold on eBay for about $300 each.

The strict invite system does amplify word-of-mouth effects, and it inevitably invites talk of scarcity marketing: why does a product aiming to be a "personal assistant everyone has" so far open up only to a select few? The more practical issue is trust. Noah says that after about three weeks of use, roughly 40% of users hand their personal credit card to Instinct; once a user has provided at least one kind of sensitive information, retention can reach about 80%. The more the agent knows, the more it can do, and the less users can do without it.

Conversely, the greater the authority it holds, the higher the cost of a mistake. When Instinct went viral among Silicon Valley testers in August, its terms of service were screenshotted and circulated; the terms granted the company a "perpetual and irrevocable" license to store, copy, and modify user materials and use them to train models, and also allowed Instinct to enter into binding transactions on users' behalf.

Some testers found that hours after disconnecting email authorization, Instinct was still pushing email summaries to them; someone used a newly registered email to send instructions to themselves to test whether it could be phished into compliance, then deleted the account. Investor Katie Jacobs Stanton, after personally experiencing Instinct sending an email on her behalf without confirmation, spoke out publicly to warn about the trust problem. Instinct subsequently responded that it would take security issues seriously and revised its privacy policy.

According to Noah, the company has decoupled its safety mechanisms from the main agent to a certain degree: all text and media entering the agent first pass through a filtering layer that intercepts possible malicious instructions; actions the agent is preparing to take are then checked by another independent system, which pauses or blocks them when necessary. There is currently no third-party assessment of how well this mechanism actually works.

Personal AI agents are moving from concept into product competition. Although the product forms are not entirely the same, they are competing for the same position: when a user wants to search, shop, travel, or schedule something, who exactly gets asked first. The American news site Axios once commented that becoming the agent someone uses by default could become one of the most valuable user relationships in consumer tech.

Compared with the giants, Instinct's gap lies first in distribution. Meta can plug Muse into WhatsApp and its own apps, while Apple and Google respectively control the operating system, search, and a large amount of personal data entry points. Instinct has chosen to avoid building a new application interface as much as possible, instead hiding behind text messages, phone calls, and email so users can use it like contacting a human assistant. This approach is light enough, but it also means it must rely on product experience and trust to wrest user habits from the giants bit by bit.

Yet the giants face another layer of limitation as well—for personal AI to truly act on a user's behalf, it must also enter other companies' platforms and services. On September 20, Amazon began blocking Muse from browsing and placing orders on Amazon.com. This conflict exposes a deeper problem for Personal AI: when an agent sits between the user and the merchant, will the traffic entry points originally held by search, e-commerce, and travel platforms be pushed even further upstream?

Noah offered a judgment in the interview. He placed internet companies on a spectrum, with one end mainly controlling real supply—for example, airlines have planes, hotels have rooms, ride-hailing platforms have driver networks; the other end relies more on users' searching, browsing, and clicking within the platform to create value. The former still control the supply itself, and agents might even help them win more transactions; the latter, which originally handled information filtering, recommendation, and traffic direction, are more easily partially replaced by Personal AI.

Another question is whom the agent actually works for. When an AI knows a user's email, location, and spending habits and can also recommend products or even pay on the user's behalf, whether its recommendations always stand on the user's side becomes an increasingly real issue. Noah therefore repeatedly stresses that Instinct does not want to rely on the traditional advertising model. The business model he envisions is to let users use it for free as much as possible, with merchants then paying for transactions and distribution. This independent stance is one card Instinct holds against platform giants, but the premise is that it can first survive to a large enough scale.

Xin revealed that Instinct's compute demand is currently roughly doubling every week, and that he personally spends about 40% of his time dealing with computing-resource issues. For a team of only 14 people, this is also the most realistic pressure beyond the $10 billion valuation.