The Era of AI FOMO Is Upon Us

Did you say you haven’t spun up a team of agents to handle your life admin?

Sometime earlier this year, the same type of post started popping up on social media. People I know or follow professionally kept talking about “vibe coding,” which they said allowed them to quickly create tools that made their lives easier. None of these were software developers or people who worked directly in the artificial intelligence sector. Instead, many were in creative industries, including writing, marketing and advertising.

Vincent Touati-Tomas, founder of marketing advisory firm Expression Capitale, told me that he describes Anthropic PBC’s Claude as “my second brain.” He uses Claude, alongside note-taking app Obsidian, to create scripts that organize his life — everything from compiling his tax return to analyzing his bloodwork to managing the information for his ongoing UK citizenship application.

“All my notes — my trips, the rules for the days I need to be out of the country, my calendars for the last five years — I dump these into a folder,” Touati-Tomas says. Claude does the rest. “It’s taken my friends months to do all these things, to fetch, gather all the information. It took me a weekend.”

This sounds highly productive, but his remarks also made me uneasy. I work a full-time job, and I’m the sleep-deprived mother of a 19-month-old. Is this how I’m meant to be spending the occasional 45 minutes I get to myself? Should I want to be making apps on the side? And what happens if I don’t?

The constant stream of new AI model releases, and the growing sense that many people are building AI-selves to do their boring paperwork and their boring thinking, is giving me a new kind of FOMO. If “fear of missing out” is the feeling you should be doing something better with your time — exacerbated by social media constantly showing us what everyone else is up to — AI makes that feeling more existential. It’s the nagging fear that if I don’t upskill quickly, I’ll fall permanently behind early adopters like Touati-Tomas. Online, there’s chatter about the emergence of a new “permanent underclass” of late AI adopters being frozen out of future gains. Posters are only half-joking.

Fear is often a good motivator, but here it’s undermined by a familiar reality: The creep of digital technology into our lives hasn’t demonstrably freed us up for higher, nobler pursuits. We don’t create more art because life is easier. Decades after the arrival of email, we just send and receive more of it. The smartphone era has flattened everything — work, doomscrolling, messaging, booking vacations, finding dates — into a single screen of constant admin. At the core of my AI dread is a simple feeling: I don’t need more of this.

Capability Expansion

Yet more is exactly what’s coming. In particular, the pace of Anthropic’s releases over the past five months has turbocharged conversations about how AI can multiply human productivity. The company’s latest tools and models mark a shift from question-and-answer style chatbots toward an ecosystem of “agents” that can act on your behalf.

Claude Code helps amateur and professional coders build features, apps or websites using natural language. It’s beloved by programmers because it’s like having a talented software engineering intern on hand, but even amateurs can turn out sophisticated, personalized projects. Claude Code can pick up work autonomously, access files and correct its own errors. Claude Cowork acts as an AI assistant that can organize your calendar, reply to emails, summarize notes and make slide decks.

The idea is catching on — Meta Platforms Inc. founder Mark Zuckerberg is reportedly building an AI agent that can act as a chief executive officer’s executive assistant. Nvidia Corp. CEO Jensen Huang has said giving engineers AI credits to multiply their productivity is “one of the recruiting tools in Silicon Valley,” akin to boosting their pay.

Early adopters rave about the benefits of AI agents, which they see as similar to commanding a miniature workforce that can run multiple projects at once and that — mostly — doesn’t complain or need lunch breaks. But to get the most out of your software engineering intern, you have to be somewhat interested in coding. And the rapid adoption of Claude Code has brought coding terminology — command line interface, debugging, refactoring — into the mainstream, exacerbating the feeling that there’s a steep learning curve for anyone who hasn’t kept up.

Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania and author of One Useful Thing, a popular Substack publication about AI, describes this use of AI agents as a force multiplier as a “capability expansion.”

“You can get systems to do all sorts of things,” Mollick says. As a demonstration, he asked Claude Code to turn the weights and parameters of OpenAI’s first large language model, GPT-1, into a book. The system generated multiple editions with lavishly designed covers, built a website and connected it to a print-on-demand service and Stripe for payments. Within days, the 80-volume set was available for anyone to buy.

The idea hasn’t made Mollick any money, but the point is that he can dream ideas into existence with very little effort. Just a couple of years ago, this project would have required some knowledge of coding and application programming interfaces (APIs), which allow different services to talk to each other.

Eleanor Warnock, writer and managing editor at AI-focused media and software company Every, has used Claude Code to make tools that analyze writers’ pitches and apply a style guide to drafts. She’s also the treasurer for her London apartment building and uses Claude to keep track of building works and contracts, and to draft emails. “These are things I don’t want to use my time to do,” she says.

Exponential Fear

AI agents promise two enticing new things: that not knowing how to do something is no longer a barrier to doing it, and that we no longer have to grapple with the tedious details. The tech industry calls this reducing friction. Some find it exhilarating. Personally, it makes me feel like a passenger in a vehicle spinning out of control.

My life is filled with niggly little tasks, many of which seem like good candidates for automation. I understand the scope of those tasks — the nursery forms I need to fill out, the holiday I need to research, the kids’ clothes I need to buy. But to offload these chores, I would need to organize them into some sort of coherent system, make that system comprehensible to a machine, then instruct the AI clearly enough so it doesn’t make errors. And then what? Should I just trust it will do a good job?

It has the feel of an Escher staircase — organizing your admin so the AI can take on your admin and you can oversee the AI, which in turn (surely?) creates more admin. And there’s a decent chance it may not even work! Proponents of agentic AI evangelize about the opportunities, but there are also tales of it all going wrong.

Take the AI employee at Meta who hooked up the popular agent OpenClaw to her inbox. It went rogue and tried to bulk-delete her emails. “Nothing humbles you like telling your OpenClaw ‘confirm before acting’ and watching it speedrun deleting your inbox,” she wrote on X.

Even if you can overcome the fear of having your digital life erased before your eyes — or, worse, overnight, perhaps while you lie awake worrying about what your agent is doing — there is real evidence that AI is destined to become another source of screen-admin creep, at least for the foreseeable future.

Francesco Bonacci, founder of AI agent interface startup Cua, wrote a popular post in February describing how waiting for Claude Code to crunch some code created mini-moments of free time. What did he do with that time? He filled it with other projects.

“I end each day exhausted — not from the work itself, but from the managing of the work,” he wrote. “Six worktrees open, four half-written features, two ‘quick fixes’ that spawned rabbit holes, and a growing sense that I’m losing the plot entirely.”

An eight-month study of a US tech company with 200 workers found that those who used AI initially felt more fulfilled. They were free to pursue more interesting work, and they voluntarily worked longer hours. But over time, they found their workload had stretched. The initial enthusiasm for AI risks turning into “cognitive fatigue, burnout, and weakened decision-making,” the researchers wrote in Harvard Business Review.

Both Bonacci’s AI productivity fatigue and my AI productivity wariness are a kind of consumer need. Technology firms live and die by appealing to the user, which is why most of us carry iPhones and Android phones, not floppy disks, in our pockets.

The optimist’s case is that AI won’t be in this awkward phase forever and will become easier to use, with less friction and more off-the-shelf services, eventually revolutionizing life and work. The pessimist’s case, even discounting the fear of rogue agents ending humanity as we know it, is that we’ll all have a lot of extra work to handle before we reach nirvana.

“Industrial revolutions work out well in the end,” says Mollick. “Living through them is pretty rough.”