The Windstorm

A weekly dispatch from the eye of the storm

Week of October 5, 2026

Welcome to the Eye of the Windstorm.

Last week a Reuters investigation, carried by CNBC, pulled the curtain back on something that has been running quietly across nearly 14,000 McDonald's restaurants: an AI pricing engine that reads millions of daily transactions and recommends the “optimal” price for every item, store by store. In one Fresno neighborhood it settles on $5.69 for a Big Mac at one location and $6.89 at another two miles down the road — a 21% gap for the same burger. Here is the part worth sitting with from the calm center. The algorithm did not decide to charge you more. A human told it what “optimal” was allowed to mean, and the machine obeyed, brilliantly, all the way down to your lunch. The capability was never the hard part. Deciding what to point it at is the entire job.

1. McDonald's taught a machine to read your wallet

The engine is built with a firm called Tiger Analytics. It weighs how much people in a given area seem willing to pay, and even pulls the posted prices off the Wendy's and Burger King down the street. McDonald's calls the portal “a tool, not a mandate,” though five franchisees told Reuters they felt pressured to follow it. That gap is the whole lesson. The model can price ten thousand stores before lunch; what it cannot do is decide whether this quarter you are chasing more customers or fatter margins, or whether a $6.89 Big Mac in a struggling neighborhood is a win or a wound. Those are human calls wearing a machine's confidence. The intelligence is rented by the hour now. The judgment is still yours, and still scarce.

2. The smallest companies are moving fastest

On September 30 OpenAI published a look at who is actually using this stuff. In a single week in mid-September, roughly four million people at small firms used its tools — and nearly one in five of them worked somewhere with fewer than ten employees. More striking: by August, agentic work made up two-thirds of small businesses' AI output, up from one-third in April. In finance, supply chain, legal, and marketing, those jobs make up at least three times as large a share of small shops' agent use as they do at the big firms. Remember the decades it took electricity to remake the factory? Buying the power was never the hard part; rewiring the building and teaching everyone to flip the switch was. This time the small shops are rewiring in months, precisely because there is less building to rewire. Size is no longer the advantage. Speed of wiring is.

3. The agents stopped waiting for you

At its DevDay on September 29, OpenAI introduced “Dots” — always-on agents that each get their own cloud computer, reach across 4,000-plus apps, and keep working toward your goals around the clock instead of sitting idle until you type. (There was also GPT-6.1 Sol, a far cheaper model that nearly matches the flagship, plus a faster Ultrafast tier. But Dots was the headline.) An agent that never sleeps is a wonderful thing pointed well and a quietly expensive thing pointed poorly, because now it compounds its instructions twenty-four hours a day. Which means the old superpower matters more, not less: the quality of the question you hand it, and the guardrails you set before you walk away, are the entire difference between leverage and liability.

The Eye

Notice what every one of these stories is really about. McDonald's did not discover a cleverer machine; it discovered it had to say, out loud and in numbers, what it actually wanted. Four million small-firm workers did not get smarter overnight; they just started wiring the switch. The always-on agent does not need a better brain; it needs a better brief. Strip the noise away and the same truth sits at the center: the supply of raw intelligence has gone vertical, and the one thing that did not scale with it is knowing what to want.

That is the great skill of this decade, and almost no one has mastered it yet — the winnowing. Not “can the machine do it,” but “of the thousand things it can now do, which handful should it, and toward what end.” The engine will price your Big Mac, draft your contract, chase your leads at three in the morning. It will never tell you which of those is worth doing, or who gets hurt if you optimize the wrong number. That sentence — the one where you name what “better” means — is the most valuable thing you will write all week. The storm is everyone racing to adopt. The eye is the quiet minute where you decide what you are adopting it for.

Do this with your agent this week

Take one real decision on your desk right now — a price, a hire, a cut, a bet. Before you ask your AI for its recommendation, type a single sentence first: “Here is what I am actually optimizing for.” Then ask it to make the strongest possible case for the opposite of what you are leaning toward. You are not outsourcing the call. You are forcing yourself to name the target out loud before the machine quietly picks one for you.

From the eye of the storm,
Grant