Instinct holds its own as Meta's Muse and OpenAI's Dots crowd the AI agent market

A startup's head start meets Big Tech

In August, a startup called Instinct shipped an AI agent with an unusual launch: invite-only, zero marketing, and barely a website. It nonetheless became the most talked-about product in AI, earning praise for a plain, text-message interface and a knack for chores like booking DMV appointments and chasing follow-up emails, according to The Verge.

That lead is now under direct attack. Meta launched Muse and OpenAI followed with Dots — broadly similar agents backed by far larger companies. But in The Verge's hands-on testing, Instinct held its own, handling everyday online tasks about as capably as its rivals. Founder Noah Shinn, an early employee of the startup Sierra, argues the difference is focus: while OpenAI and Meta build AI products for many kinds of customers, his company works on a single thing, a personal assistant for everyday life.

Investors appear to share that confidence. The Verge reports Instinct was valued at $10 billion in late September, after Muse had already arrived. Access still requires an invitation from an existing user or waitlist approval — The Verge was admitted within a couple of hours — and there is no subscription fee for now.

No app, no mascot, just texts

Where Muse ships a cuddly bear mascot and Dots anthropomorphized blobs, Instinct has no character at all. Conversations run over iMessage, WhatsApp, or email, with short replies, the occasional emoji reaction, and more doing than talking. The agent connects to services including Google Workspace, Slack, and Notion, and when it needs to log into a website it sends a link to a "secure vault" for credentials. That approach makes it available anywhere texts work, including CarPlay and smartwatches — a quiet omnipresence the reviewer found equal parts convenient and unsettling.

Capable, with human-web blind spots

The first impression was rough: the agent took the reporter's Cincinnati area code as a cue to send local news links and flight search results. Given meatier tasks over a couple of weeks, though, it mostly impressed. Asked to find age-appropriate swimming lessons, it returned workable options and pivoted to a neighborhood community center on request, with dates and times that all checked out — though it missed a prerequisite class hidden in a pop-up window, a reminder that agents cope well with clean data but stumble on the messy, human-built web.

Other chores went smoothly, including emailing an eye doctor's office to arrange an appointment and completing the digital paperwork for an Ikea return. It even dug through past Ikea purchases and identified a product described only as vertical, white, and branch-like (a Tjusig rack, for the record). More striking was the unprompted helpfulness: it confirmed it had the right doctor by citing a 2024 confirmation email, and flagged a potential calendar conflict while working on an unrelated errand. That kind of proactive relevance, The Verge argues, is a real step beyond a glorified autofill machine.

The monetization question

Instinct is free for invite-only users, and Shinn says the company intends to keep it as affordable as possible. Techsponential analyst Avi Greengart tells The Verge that a venture-backed startup's priority right now is growth — establishing something worth paying for — with revenue from the customer base or an acquisition by a larger company as possible endpoints. Meta's Muse is free to use, while OpenAI's Dots are limited to paid subscribers. Greengart speculates that free agents may eventually cut deals with the businesses users ask them to reach, because if you are not paying for a service directly, you are paying for it indirectly. IDC senior director Greg Ireland sees advertising as a natural fit, given how shopping- and transaction-heavy these agents already are.

Why it matters

The consumer AI agent market went from a startup novelty to a three-way contest with Meta and OpenAI in a matter of months, and Instinct's $10 billion valuation shows investors believe a focused underdog can survive the collision. The Verge's testing suggests the core technology is converging quickly, with all three agents handling life admin comparably — which means distribution, pricing, and trust may decide the winner more than raw capability. And how these agents eventually make money, whether through subscriptions, commerce deals, or advertising, will shape whether users can trust them to act in their interest rather than a sponsor's.