Instinct is a four-month-old company with a $2.5 billion valuation and no app to install. You reach it by text or phone call, and it will book your flights, cancel your subscriptions, and argue with your insurance company while you sleep. That is the pitch. It is also the problem.
Silicon Valley is in the middle of a personal assistant frenzy, and two startups sit at the center of it. Instinct, founded by 23-year-old Noah Shinn, just closed a $250 million Series B co-led by Index Ventures and Benchmark, bringing its total funding to $350 million. Across town, a company called Town is closing in on a $1 billion valuation in a round led by the same Index Ventures. Both are betting that the next wave of AI will not live in a chat window but in your life, acting on your behalf across your apps, your devices, and your calendar.
Town takes a different approach. It offers customizable characters, or "Townies," that users can assign to handle specific tasks. The userbase ranges from investors to plumbers, each deploying their own digital stand-in to manage workflows, answer questions, or coordinate schedules. Instinct skips the characters entirely. It operates through texts and calls, learning your preferences by watching how you use your phone and computer. The interface is your existing interface. That is the appeal. It is also where the unease starts.
The Accessibility Trade
Both companies position their AI agents as personal assistants rather than developer tools. This is a deliberate shift. For years, AI agents were built for engineers who could write code, configure APIs, and manage permissions through a dashboard. Town and Instinct are built for everyone else. The promise is that you will not need to learn a new system. You will just tell your phone what you need, and it will happen.
The excitement around this accessibility is hard to miss. Early users describe a sense of relief at handing off tedious tasks to an agent that already knows their email, their calendar, and their shopping habits. The friction of switching between apps disappears. The agent becomes a layer over your digital life, smoothing out the gaps. For people who live in their phones, this feels like progress. It feels like the technology is finally catching up to how we actually work.
But the same features that make these agents useful also make them invasive. To function, Instinct requires access to emails, text messages, screen content, and location data. Town's characters need similar permissions to act on a user's behalf. This is not optional. An agent that cannot read your email cannot reply to it. Without seeing your screen, an agent can't move through your apps. The access is the product.
Privacy concerns have surfaced quickly. Security researchers point out that granting an AI agent persistent access to personal data creates a single point of failure. If the agent is compromised, the attacker gains a window into everything the user has shared. There are also questions about how the data is stored, who can access it, and what happens when the company changes hands or shuts down. These are not hypothetical risks. They are the direct result of the design.
The documented engagement around these stories splits cleanly. On one side, there is enthusiasm for the convenience and the speed. On the other, there is caution about the privacy implications. The numbers tell the story as clearly as any quote could. Instinct's valuation jumped from $50 million to $2.5 billion in a matter of weeks. Town is approaching $1 billion. Investors are betting that the convenience will win. Critics are betting that the risks will catch up.
What remains unknown is how these companies will handle the inevitable incidents. Data breaches do not announce themselves in advance. They happen when a system is stressed, when a permission is misconfigured, or when an employee makes a mistake. The question is not whether these agents are useful. They are. The question is whether the utility is worth the exposure.
Silicon Valley has a history of shipping first and fixing later. The personal assistant wave feels like a repeat of that pattern. The technology is impressive. The use cases are compelling. The privacy model is an afterthought. That is the tension. It is also the bet. Investors are wagering that users will accept the trade. Users are still deciding.
The last word has not been written. The products are still in private beta. The valuations are still climbing. The privacy concerns are still unresolved. This is the moment where the story could go either way. The technology is ready. The question is whether we are.
Browse: Tech