Katherine Johnson approximated John Glenn's re-entry trajectory by hand, one calculation per second, using Euler's method. Glenn refused to board the rocket until she checked the numbers. She got it right. He came home alive.

Today we run that same math trillions of times per second, across thousands of processors, on trillions of times more data. We have built a universal approximator and a universal generator. Two new tools for science and business.

Feed a model enough low-level data over time and something remarkable happens. Pixels over time, and it learns physics, fluid motion, emotion, cinema. Nucleotides over time, and it learns how proteins fold, how genes express, how cancer progresses. The same architecture, the same math, working across physics, biology, and creativity.

That shouldn't be surprising. We've been approximating the world for centuries. Newton was wrong, but close enough to build bridges and predict tides. We traded pen and paper for calculus, and it gave us the modern world.

Now we're doing it again, at a scale Katherine Johnson couldn't have imagined. Instead of one equation solved by hand, these models use high-dimensional functions with billions of parameters to approximate the observable universe. They're still wrong. Newton was wrong too. But they're less wrong than anything we've had before, and getting less wrong fast.

A healthy life is a path through the space of valid lives. The edges are illness and death. A gorgeous video is a path through time, with characters, places, things, and emotions. A drug binding to its receptor is one shape fitting into another. The math doesn't know which one it's solving. It just finds the path.

I've been in the room for four of these shifts. I advised Lou Gerstner as IBM bet the company on the internet. That work grew into a $15B business. I was Chief Technology Advisor to Bob Moritz as PwC moved 6,000 partners and 160 territories to cloud. I built Google Cloud's AI business from NASA to healthcare to genomics, growing it into one of Cloud's fastest segments. Now at Alphabet I work on generative AI, where the same pattern is playing out in content and advertising.

Every time, I go hands-on first. I learn the technology deeply, figure out where it's headed, and bring what I find to CEOs and boards. That's what I do.

Copernicus wrote down the position of the sun and moon every day. Not to publish. To see. The data told him we weren't the center of the universe.

I started doing the same thing with AI. Every day I document one deployment that moved a number. A short-form video. One data point. I started with the hyperscalers and ran out after a few hundred. The well ran dry. I needed a bigger lens.

So I built the AI Radar. I pointed AI agents at 3,338 public companies, scanning every earnings call, every 10-K, every press release. They extracted 4,332 verified AI deployments. Thousands already in production. It blew my mind. The gap between the companies that understand this phase change and the ones doing window dressing is enormous.

Now I document the most interesting and impactful examples daily, looking for the patterns. Hundreds of data points so far. Thousands to come. AI Money Moves is the daily log. One AI deployment, one number, one video. The weekly newsletter pulls back and connects the dots.

Here's what the data tells me.