Instinct raises $1B at a $10B valuation for its personal AI agent

Why it matters

Instinct is being valued on the promise that an assistant can act across a user's digital life. Scaling that promise depends on reliable execution, clear permission boundaries and economics that work beyond a large funding round.

Instinct said on September 28th that it raised $1 billion at a $10 billion valuation, giving founder Noah Shinn capital to build a personal agent that acts across people's apps and devices. The Series C included Sequoia Capital, Benchmark and Coatue, according to Reuters. Reuters did not identify a lead investor.

Shinn came to Instinct after working as a research scientist at Sierra and conducting machine-learning and programming-language research at Northeastern and MIT. As a Northeastern student, he co-developed Reflexion, a framework that has a language model check its own output and use feedback to improve subsequent attempts. Northeastern reported that the method reached 91% accuracy on the HumanEval coding benchmark, 11 percentage points above the result then attributed to GPT-4. The research offers a useful clue to Shinn's approach: getting an AI system to do a task is one problem; getting it to recognize and recover from mistakes is another.

That problem is now personal. Instinct's website describes an assistant that connects to email, messaging, a user's screen, audio and location. People can text or call it; the assistant is designed to use phones and computers rather than ask users to learn a new interface. Reuters said Instinct had recently released products including a concierge service for bookings and phone calls.

A valuation that moved faster than the product's public record

The financing marks a sharp step up from Instinct's previous round. In August, TechCrunch reported that the company had raised $250 million in a Series B at a $2.5 billion valuation, bringing reported total funding at that point to $350 million. The new valuation is four times that August figure in roughly a month. The speed of the repricing is clear; it does not, by itself, show how much revenue, paid usage or customer retention Instinct has built.

The latest announcement does not include user or revenue figures. The Information reported on September 15th that Instinct had passed 100,000 users, citing a person familiar with the company, and that demand had strained capacity. That is a reported user count, not a measure of paying customers or how often the assistant completes tasks successfully. Capacity strain is especially relevant for an agent that may need to make repeated calls to models and external services to finish a single request.

In a September 28th post promoting his conversation with Shinn, Patrick O'Shaughnessy said the invite-only product was growing roughly 10% a day without marketing spend. The post did not give a starting user count or specify the period used to calculate that rate, so the figure cannot be reconciled with the reported 100,000-user count or treated as an audited company metric. The conversation's listed topics include buying compute ahead of demand and Instinct's business model, both relevant to the costs and revenue behind that growth.

The timing also follows a rapid fundraising sequence. RuntimeWire reported in August that Shinn's startup was seeking at least $200 million at a valuation above $2.5 billion, with Benchmark and Index Ventures in discussions around the earlier financing. TechCrunch later reported the $250 million Series B. The September Series C brings Sequoia and Coatue into the named investor group alongside Benchmark, but the announcement does not say which firm led or whether every earlier investor participated.

The harder task is earning permission

Instinct's design asks users to grant an assistant access to the context that makes personal help useful: communications, accounts and other connected services. That access is also where a mistake can have consequences beyond a bad answer. A booking, message or purchase made on a user's behalf must be correct, authorized and recoverable when it is not.

TechCrunch reported in August that testers had raised concerns about Instinct's permissions and terms while the product was in private testing. RuntimeWire has since covered reports of slowdowns and capacity warnings, as well as a user's allegation that the assistant surfaced another person's financial details. Screenshots cited in that report did not establish whether another user's information had been exposed or the assistant generated the details; Shinn denied a data breach. Those reports do not settle the product's security, but they show why reliability and permission controls are part of the product itself, not a later polish pass.

The financing gives Shinn resources to expand a product whose ambition is to act across services rather than stop at advice. The next test is whether Instinct can make those actions dependable as usage grows, and turn a large reported audience into a business with repeat use and a workable cost structure. The round establishes investor confidence in that possibility. It does not establish the outcome.