AI Agents Want to Run Your Life. Instinct Shows Why the Right Form Matters

The experiment began when I quit my full-time job at WIRED and gave Claude Cowork access to my email and calendar. After reporting on artificial intelligence for years, I was curious about what the technology could do for me. I imagined Cowork as a highly capable digital assistant, but I was disappointed to find that it felt like interacting with a regular chatbot—albeit one with additional tools. “What do you want to do?” it seemed to ask. I didn’t understand.

Then Instinct, an invite-only AI agent that communicates with users through iMessage and WhatsApp, started popping up around the Bay Area. Connect it to your email, calendar, and messaging apps, and it begins to resemble OpenClaw for the masses. The company launched a private beta in February. The Information reported that the company was in negotiations to raise $1 billion in addition to the $350 million it had already raised, potentially bringing its valuation to $10 billion.

Instinct’s moment in the spotlight comes as Meta experiences an unexpected explosion in popularity for its AI assistant, Muse, which it launched earlier this month. As of this writing, Muse is the most popular free app on Apple’s App Store, with more than 900,000 downloads, according to third-party estimates. Despite serious security vulnerabilities in X that allow “attackers to do ‘anything’ they want on a victim’s Mac,” according to Ars Technica, the platform’s commentator community seems genuinely excited about the app. I don’t plan on handing over a large amount of my personal information to Meta, but if Muse can keep me logged in to Bloomberg or unsubscribe me from Hot Yoga São Paulo—the mailing list I’ve been emailing once a week since 2017—I may reconsider.

Why AI Agents Feel Different From Chatbots

The gulf between people who use AI agents for everything and those who have never used one has rarely felt wider. This may help explain why technology company CEOs were almost caught off guard by the backlash against data centers. The costs and disruption of building data centers may seem like an acceptable trade-off if AI agents can automate people’s tedious administrative tasks and significantly increase productivity. But if you use chatbots primarily as a more elaborate form of Google, the bargain is much less appealing.

The central problem is that most people’s problems are not software-based—and even when they are, people often don’t recognize them as software problems. Technology journalist Jasmine Sun writes that “most people’s problems are not software-based, and even if they are software-related, most people don’t notice.”

Like chatbots, AI agents still require substantial input and direction from the user. If you know what you want—help with homework, for example, or a virtual girlfriend—AI tools are eager to provide it. But if you don’t know what you want, well… welcome to the permanent underclass.

I wanted to avoid that fate. I may not be agentic, but I consider myself action-oriented. I was able to identify several software problems in my own life. I had previously used Claude to create freelance invoices, saving about 10 minutes. It worked well, but it was hardly life-changing. I decided to give Instinct a try.

Instinct’s Biggest Advantage: Its Form Factor

When you talk to AI researchers about Instinct, you often hear the term “form factor.” Instinct has the right form factor. There are no open text boxes—only iMessage, WhatsApp, and a few helpful prompts.

I cannot overstate how important this is. Instead of trying to figure out how to make a chat experience more convenient, you simply text with an agent, suggest something it can do, and let it take action. Unlike other agents in this field, Instinct does not text constantly. I tried a competing agent called Lindy for a week, but it started texting me every morning and joining Zoom meetings without my permission. I deleted it.