Let’s talk about the decision the numbers said no to — and the person who made it anyway.
Picture a room. Could be a boardroom, could be a kitchen table, could be a garage. There’s a spreadsheet, or a report, or just a consensus. And the numbers are clear. The market says no. The data says no. Everyone sensible in the room says no.
And one person says: I’m doing it anyway.
We have a word for that person. We call them unreasonable. And mostly we mean it as a compliment, because we know — somewhere under the eye-rolling — that everything we use, every company we work for, every product we carry, started as a decision the evidence did not support.
The evidence never supports the new thing. That’s what makes it new.
George Bernard Shaw wrote it more than a century ago. The reasonable man adapts himself to the world. The unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man.
He was talking about people. We have since built the perfect reasonable man. And we’re putting it in charge.
Here’s what a machine is, under all the marketing. It’s trained on everything that already happened. It’s calibrated to consensus. It’s rewarded for giving the answer most likely to be right, which is the answer most like every other answer. The machine is reasonable the way water is wet. It cannot be anything else.
And it can never want. That’s the part nobody says out loud. It optimizes — beautifully, relentlessly — but you have to hand it the goal. It doesn’t have one of its own. It has never once looked at the world and thought: I want something that doesn’t exist yet.
So we’ve built the most reasonable thing in history, and we’re slowly handing it the job of deciding what’s worth trying.
People think the danger is the machine being wrong. It’s not. The danger is the machine being reasonable.
Watch how it works in practice. The proposal goes through the machine first. The machine, trained on what worked before, flags everything that hasn’t worked before. And it’s not even hostile — it’s just honest. It says: low probability, weak precedent, no market evidence. Which is true. Which is exactly the problem.
Every new thing that ever mattered had low probability, weak precedent, and no market evidence. That’s the definition of new. The machine isn’t rejecting a bad bet. It’s rejecting the category that bad bets are born in.
The average has never built anything. The average maintains. The future doesn’t come from the average — it comes from someone who refused to be reasonable about what was possible, and then went and found out.
Now the operator part, because this matters if you actually build things.
The unreasonable direction needs a reasonable auditor. You want the machine costing the attempt, stress-testing it, pricing the failure honestly. It’s brilliant at that. The best auditor we’ve ever had.
The mistake is reversing it. The moment the reasonable machine decides what gets attempted, you haven’t added judgment. You’ve added a filter that only lets through what’s already been done. And you’ll never notice, because the filter is polite, fast, and almost always right. It will just quietly close the frontier.
That’s what an all-reasonable world looks like. Nothing new. Everything maintained. Every hunch declined, every risk priced, every what-if answered with a probability. The future doesn’t get cancelled by force. It gets cancelled by veto.
So here’s the principle I’d keep, if I had to keep one: the right to be unreasonable.
I don’t mean reckless. Reckless ignores the cost. Unreasonable knows the cost and goes anyway. The founder who hears the numbers and goes anyway. The engineer who builds the thing nobody asked for. The doctor who tries the treatment the protocol doesn’t cover. The person who wants a life the data on their life doesn’t support.
That wanting is the whole engine. It’s the one thing the machine cannot fake, cannot optimize, cannot even register.
So don’t hand it over. Use the machine for what it’s good at — the honest price of the attempt — and keep the unreasonable part for yourself.
The future has never been built by the average. It gets built by whoever wanted something badly enough to be wrong about it. And the machine, for all its speed, has never once in its life wanted anything at all.