Something prompted you to open this. A headline. A share from someone you trust. A notification you didn't ask for. You didn't wake up planning to read an essay about prompting. Something put it in front of you, and here you are.
That's not an accusation. It's just how information works now.
We've spent the last few years learning to prompt machines. There are courses for it, job titles, whole playbooks. Ask the right way and the model gives you what you want. Fine. But while everyone was learning to prompt AI, the bigger story was running in the opposite direction. Almost everything you consume is prompting you.
Call a prompt anything built to get a response out of you. By that definition, most of what reaches you in a day qualifies. Content is a prompt. A comment is a prompt. A like, a view count, a read receipt. The platforms are not in the information business. They're in the reaction business. Information is just the format.
And your reaction doesn't end with you. You comment, and the comment prompts someone else. You share, and the share lands in ten other feeds. You watch to the end, and the system learns that this works and makes more of it. Even doing nothing is a signal. The creator watches the numbers, and the numbers prompt the next post. Everyone is prompting everyone, all the time, in both directions. Nobody in this system is only a consumer.
Here's why it turns extreme. Prompts wear off. The thing that made you stop scrolling last month gets scrolled past this month. You build tolerance, the way you do to anything. So every new prompt competes against your immunity, and there's only one way to win that competition: escalate. Louder. Faster. More emotional. Nobody sits down and chooses this. The loop selects for it. Constant prompting becomes extreme prompting on its own.
Then generative AI arrived and removed the last constraint.
Until now, prompts were made by humans and distributed by machines. Human effort was the bottleneck. Someone had to write the post, film the video, think up the headline. That bottleneck is gone. Now a machine can generate the prompt, distribute it, watch your reaction, and generate a sharper one. Infinite volume at zero cost, tuned to you personally, in real time. The loop needs a human at exactly one point: the reaction. Yours.
And it's more personal than a feed ever was. A feed prompts a million people at once. A chatbot prompts you alone, in your own words, inside your own context. Every AI answer is itself a prompt back at you. We tell ourselves we're the ones prompting the machines. Look at the volume in each direction. It's not close.
I want to be transparent about where my own thinking moved on this. My first version of this idea had a clean ending: bad prompting burns out, good prompting wins. I liked that ending. I can't defend it. I don't know who wins.
What I'm sure of is different, and I think it matters more. The rate of change for humans is going to shoot up. For good or for bad.
The mechanism is simple. Humans change through a loop: stimulus, response, feedback. Prompting is the clock speed of that loop. Multiply the prompts and you multiply the speed at which people change. Beliefs, habits, skills, tastes. What took a decade to shift will shift in a year. Consensus will form and collapse in weeks. The half-life of what you know gets shorter. So does the half-life of what you're sure of.
And it cuts both ways at once. Learning gets faster and radicalization gets faster. Skills compound quicker and so do compulsions. The loop is an amplifier, not a filter. It doesn't check what it's amplifying.
The obvious objection is numbness. Won't people just tune out? Some will. But numbness at scale is not standing still. A generation going quiet would be one of the fastest cultural shifts in history, not the absence of one. And personalization exists for exactly one reason: to get past your tolerance. A billboard wears off. A prompt that adapts to you doesn't.
There's a harder question underneath all this, and I'll only pose it. If human choice is as free as we like to believe, why does any of this work? There is now a trillion-dollar economy whose returns depend on human decisions being predictable enough to engineer. It posts results every quarter. Philosophers have argued about free will for a few thousand years. The feed runs the experiment every day, at scale, and publishes the numbers. Believers in free will are going to have a hard time explaining this age. Maybe the will is fine and it's only the inputs that got captured. Maybe. But the burden of proof has moved, and it isn't on the skeptics anymore.
So where does it end? Maybe the optimists are right. There's a real case that bad prompting eats itself. It exhausts the attention it feeds on, and once AI makes manipulative prompts infinite and free, they become noise. Worthless. Trust and judgment become the scarce assets, and scarce assets win. There's an equally real case that bad prompting never dies. It just mutates, the way outrage always has, and we end up with a junk layer and a premium layer, the way food did.
I can't promise a direction. I can promise a speed. The rate is guaranteed. The direction is a choice.
And notice where that choice actually lives. Not in your reactions. Your reactions are exactly what the machine has learned to play. The choice lives one step upstream, in what you let near you. Which feeds you open. Which rooms you sit in. Which tools, which voices, which questions. You can't choose your reactions. You can still choose your prompts. Freedom used to feel like a default setting. At this speed it looks more like a practice.
Which leaves the only question that matters now, and you can check it tonight, the next time the screen lights up: what's prompting you, and did you choose it?