Instinct jumps from $500M to $2.5B valuation in weeks while still a revenue-free beta

What happened

Instinct, a personal AI assistant that operates over text messages and phone calls, has been valued at $2.5 billion — roughly five times the figure attached to it just weeks earlier — according to a report on dev.to. The new round was led by Benchmark and Index Ventures, and arrived only weeks after an early-August round led by Kleiner Perkins that valued the company above $500 million. Total funding raised now sits at around $350 million. Despite the price tag, the product remains a free, invite-only beta with no disclosed revenue and no published user numbers.

The researcher behind a texting interface

The company behind Instinct is Spear Street Technology, and its founder is Noah Shinn, described by dev.to as the author of Reflexion, a landmark paper on AI agents that improve through self-critique, and a former research scientist at Sierra. That background shapes how the product is read. Users connect their apps and devices, then message or call the assistant to run errands on their behalf; the founder's own framing, relayed by dev.to, is that there is no new interface to learn — the system is trained to use a phone and computer much the way a person would. Early users have reportedly planned cross-country road trips, handled weekly grocery orders, bought concert tickets and cancelled hundreds of dollars in subscriptions through it.

Hard tech behind a simple entry point

The dev.to analysis argues that the interesting part is not secret technology but the packaging. Texting and calling are unforgiving interfaces — no confirmation dialogs, no way to clarify an ambiguous sentence — so an agent that completes multi-step tasks through them is doing substantial hidden work. That work includes breaking vague requests into verifiable steps, deciding which actions need the user's sign-off and which can proceed automatically, keeping coherent state across email, calendar and other apps, and operating real interfaces rather than only calling APIs. In this reading, investors are racing to back whichever team can first make serious agent research usable by people who will never learn to configure anything.

A bet on position, not earnings

Dev.to is blunt about the valuation: nothing disclosed so far supports $2.5 billion on current performance. The number is a wager that Instinct could become the default gateway for personal AI — a positional race, the piece suggests, similar to the earlier scramble to own the smartphone era's entry point — rather than a reward for money already earned. The author urges readers to keep two things separate: a product going viral and attracting capital, and a product validated as a stable business.

The permission problem

The piece also raises an issue it says the excitement has largely skipped. An assistant that sends email, books flights, charges cards and cancels subscriptions needs deep access to inboxes, calendars, payment methods and account credentials. After Instinct went viral, dev.to notes, its terms of service and the scope of permissions it requests drew scrutiny. In products like this, convenience and permission depth are tightly bound, and the author argues that boundaries need to be set where the agent cannot override them — otherwise the cost of a task being silently handled may be an unexpected charge on the user's card.

Why it matters

A fivefold valuation jump in a matter of weeks, for a pre-revenue beta, is a strong signal about where venture capital believes the AI agent market is heading: toward consumer assistants that act across a person's entire digital life rather than chatbots that merely answer. It also shows that agent startups can now be priced on founder research pedigree and interface simplicity alone. The two questions investors are effectively deferring — whether the convenience can be monetised, and whether the permission model is safe enough to trust with a card and an inbox — are the same ones users and builders should be watching as this category matures.