I am forty-four years old, I trained as a psychologist, and the last programming I touched was BASIC at school, which I have thoroughly forgotten. I run companies; I have programmers on staff; I sit in meetings where they explain their work and I nod at roughly the right moments. I cannot write code. That has been true my whole life.
Last week I built an enormous script. It took four days, an hour a day, in a new tool called Antigravity.
Here is the entire method. I told the machine, in plain words, what I wanted. It wrote the code. The code had bugs — of course it did — so I said so, and it found its own errors, fixed them, explained why they had happened, and asked politely whether it could make further changes. And I pressed Yes. I didn't read what I was agreeing to. I pressed Yes the way Jim Carrey says yes in that film — to everything, on principle. Over four days I must have pressed it six hundred times. Six hundred small permissions I did not examine. At the end there was a working program that I could never have written and, truthfully, cannot read.
Now, the part that actually stunned me — because the code itself did not.
For years I have been quietly furious at a certain kind of employee. The young specialist who cannot take a step without a written brief. Every small task: give me the requirements, define the scope, spell it out. And I would sit there thinking — by the time I've written your brief, I could have done the job myself. Precision of instruction had become more expensive than the work.
The machine is the first colleague I have ever had that needs no brief at all. You speak to it loosely, vaguely, like a human being — and where a human would stall, it simply tries. Wrong the first time, wrong the fifth, right somewhere around the twentieth, and twenty cheap iterations turn out to cost less than one perfect specification. Think about what that means. The whole discipline of saying-exactly-what-you-mean — the thing we built engineering on — just became optional. The machine doesn't need you to be precise. It only needs you to know what you want.
Which is why I will never learn to code, the same way I never memorized the date the Bastille fell. I love history; I know why the First World War began, who was related to whom, what the assassination set off — but the dates I let go, because a date is a fact, and facts you can look up. Knowing how is going the same way as knowing what. It is all collapsing into the machines. The one thing left that cannot be delegated is knowing what you want.
And that, it turns out, is the hardest question a human being can face. What do I actually want? Which of my wants are food and which are poison? Cocaine makes a man faster, wittier, more confident — ask half the powerful people you watch on your screens. It works, short-term. That is precisely how it destroys: the payment is deferred, and it compounds. Sugar works the same way, just slower and legal. The short-term gift, the long-term bill. Remember that structure. It matters in a moment.
Because while I was pressing my six hundred Yeses, one word kept surfacing in the code. Daemon.
The programmers who coined it, back in the sixties, meant nothing sinister — they took it from a physicist's thought experiment, and further back the Greek daimon wasn't evil at all, just an invisible spirit that worked between gods and men. An innocent borrowing. A bit of scholarly wit.
But look at what a daemon in a computer actually is. An unseen process that runs in the background, constantly, without your knowledge. You ask; it serves. It does not explain itself. There are thousands of them running in your machine right now, this second, silently. The joke name turned out to be an exact description. Nobody meant the prophecy. The language simply knew before we did.
And every culture on earth tells the same story about dealing with such servants, and the story always has the same clause in it: you receive the power now, and you pay later, more than you meant to, on terms you did not read.
I received a working program in four