Here’s something to think about.

The most popular product on Earth—something Trump now demands we call “Super Intelligence!”—is basically free.

Consider: ChatGPT alone has roughly 900 million weekly users. About 5% pay for it. 

Nothing this popular has ever been this free. 

The PC took years to reach the office, and it was insanely expensive. AI got there in a third of the time, and most paid almost nothing. By last May, an astonishing 45% of American workers used it on the job, according to a survey tied to the St. Louis Fed.

So here’s the question that matters for your money and everyone else’s… 

What happens when 900 million freeloaders start paying? And who collects?

The answer has everything to do with one of the LAST great windows of opportunity in the AI trade. 

Watch the Clock 

METR, a research group that stress-tests AI agents, tracks how long a task an AI can finish on its own. Since 2023, that length has doubled roughly every four months. Lately, closer to three.

Here’s what that looks like. 

A programmer leaves a job queued at 6 p.m.—the kind that would eat two full workdays. She goes home. By morning, the AI has finished it. 

Two years ago, it handled tasks measured in minutes. Now it can work around the clock. 

People might pay $20 a month for a chatbot. They’ll pay far more for a worker. Somewhere on this curve, the toy becomes staff.

Of course, there are objections to this. 

Factories. Kitchens. Streets. 

One objection: AI keeps getting cheaper. Epoch AI estimates the cost of a fixed level of AI smarts is falling between 13x and 50x a year. Faster than electricity ever fell. Faster than batteries. Faster than DNA sequencing.

How does anyone get rich selling something that gets 50x cheaper every year?

It’s simple: the same way companies did during the Electric Age. As electricity got cheap, people wired everything. Factories. Kitchens. Streets. 

The price per unit collapsed and the total bill exploded. Cheap AI follows the same law. But 100x faster. Every price cut opens up work that wasn’t worth automating last year.

And the bill’s already exploding.

Enterprise spending on generative AI tripled in 2025, to $37 billion. Menlo Ventures called it the fastest-scaling category in software history.

This year, one product shows how fast it’s still moving. In September 2025, Anthropic’s coding tool, Claude Code, was bringing in revenue at a pace of $500 million a year. By February, that pace hit $2.5 billion. Five months. Five times the money.

That’s the freeloader-to-payer shift. It’s happening at scale. Right now.

Hunting For the Next 100x 

Most AI catalysts are guesses: earnings whispers, product rumors, leaks. This one’s written on a calendar. 

In 1965, Gordon Moore drew a line. Chip power would double on a schedule, roughly every two years. It held for half a century. Whole industries planned around it. The investors who believed it early got rich.

METR’s line is the new Moore’s Law. It doubles every three to four months, about six times faster. Since early 2023, AI went from tasks measured in minutes to two full workdays.

Read it forward. 

One more doubling gets you almost a full work week. Two more get you two weeks. A month of work arrives by mid-2027. 

Each doubling converts another slice of the freeloaders into customers. Each one moves the money a little further toward the companies sitting on the chokepoints.The further downstream it goes, the smaller the companies that catch it. 

THAT is where you find the next 100x. But this window won’t be open for long. The calendar always turns whether you’re watching or not.