There's an optimistic argument going around that I mostly agree with, and the part I don't agree with is the part that matters. The argument runs like this. For most of history, most people had to farm, because food was scarce and producing it took nearly everyone. Then productivity rose, the share of people farming collapsed, and the result wasn't a population of useless mouths. It was a population freed to do everything else, eating better than any peasant could have imagined, without knowing how to grow a thing. Intelligence, the argument says, is about to go the same way. Most of us will stop producing it and start consuming it, and the demand for it will turn out to be as bottomless as the demand for everything else once it got cheap.
I think that's right about demand. I think it's wrong about what you keep. You can eat without farming because eating doesn't require the faculty that farming builds. Consuming intelligence does. The thing you need in order to use a machine's thinking well is the same thing you build by thinking yourself, and if you stop building it, you lose the ability to tell good output from fluent output. There's a minimum amount of thinking you have to keep doing on your own for the rest of it to be worth anything. We call it the judgment floor.
The food analogy breaks at the point of use
Taste is a faculty you keep whether or not you cook. You can lose every farming skill your ancestors had and still know when a tomato is bad. The analogy relies on that. It assumes the ability to consume intelligence survives the loss of the ability to produce it, the way the ability to enjoy a meal survives the loss of the plow.
It doesn't, and the reason is specific. Judging a piece of reasoning means running at least part of it yourself. When a model hands you an analysis, you decide whether it's right by checking the assumption that matters, following the step that looks weak, noticing what it left out. Every one of those moves is a small act of production. You can't evaluate an argument without doing some arguing, the way you can evaluate a meal without doing any cooking. So the faculty that lets you consume intelligence well is downstream of producing it, and a person who produces none loses the ability to tell when they're being served something bad. Taste survives. The thing taste would have needed to rest on doesn't.
This is the medium changing what we value, again
I wrote a while back about AI as a medium, borrowing from Postman's argument that every dominant medium reshapes what a culture rewards. Print rewarded sequence and patience. Television rewarded speed and reaction. The claim was that AI is becoming the medium through which businesses understand themselves, and that nobody was asking how the machine shapes what the organization sees.
The same thing is happening one level down, to the people. AI rewards the question and punishes the slog. It makes the act of asking enormously productive and the act of grinding through nearly worthless. That's a fine trade for anyone who already knows how to ask well. The trouble is that asking well was built by grinding. The analyst who can spot the hole in a model's forecast learned to spot it by building forecasts, badly, for years. If the medium removes the years, it removes the holes-spotter too, and we get a generation that can command more intelligence than any human in history and can't tell when it's wrong.
The optimists are right about the upside, and it's bigger than they say
I want to be careful here, because this is where the piece could turn into something I don't believe. I'm not nostalgic for the slog. Most of it was waste. And the optimistic case has a point that's easy to miss if you're worried about atrophy. Abundant intelligence doesn't just let people consume more. It changes what people can want.
Someone who never had a reason to think hard about a problem, because they lacked the tools or the access or the time to do anything with the result, suddenly finds that a modest investment of attention opens a door. They can build the thing, investigate the thing, understand the thing. The return on paying attention goes up, and so does the incentive to pay it. I see this every week. People who were never going to write a line of code are building their own tools. People who would never have read a regulation are interrogating one. The medium is pulling more people into real thinking than any school ever did, because for the first time the thinking pays off immediately. That's the version of this I'm excited about, and it's a version where people think more, not less.
The same medium produces both outcomes, and the difference is the floor
Here's the part the optimists skip. The medium that raises the return on thinking also lowers the cost of not thinking, and it does both at the same time to the same person. The analyst who asks the model to find the weak assumption is building judgment. The analyst who accepts the forecast because it reads well is spending it. Same tool, same afternoon, opposite trajectories, and nothing in the interface tells you which one you're on.
What decides it is whether you keep a floor. The judgment floor is the share of your own thinking you refuse to delegate, kept not out of principle but because it's what keeps you able to evaluate everything you do delegate. It's small. It doesn't mean doing the work by hand. It means checking the step that matters yourself, forming your own view before you read the model's, and treating the model's output as an opening position rather than a verdict, the habit we described in the first-draft surrender. Below the floor, every gain in capability is borrowed, because you can no longer tell whether it's real. Above it, every gain compounds, because you can.
Why this is a leadership problem and not a self-help one
An organization has a judgment floor too, and it's the sum of the individual ones. A company where everyone stays above the floor gets the optimistic future, more people thinking harder about more things, with the machine as the lever. A company where people drop below it gets fluent output nobody can check, and the first time that output is confidently wrong in front of a customer or a regulator, the company finds out it has nobody left who can see why.
The danger isn't that AI makes people stupid. It's that the medium makes dropping below the floor feel like efficiency, and efficiency is the one thing every incentive in a business rewards. So the floor has to be installed on purpose. That means the standard for using a model is engaging with its reasoning, never just its answer. It means the people who push back on the output get recognized, not the people who ship it fastest. It means every workflow we design keeps a human doing the part that builds the judgment to oversee the rest, which is a different design question from where the human reviews. And it means being honest that the slog had one thing in it worth keeping, and finding a cheaper way to keep it.
The agricultural revolution freed people from farming and they went on to do everything else. This one can free people from the cognitive subsistence work and send them on to think about things that were never within reach. The only condition is that they keep thinking. Consume as much intelligence as you can find a use for. Just stay above the floor that lets you tell what you're eating.
If you want AI built into your operation in a way that raises the return on your people's thinking instead of replacing it, start with an AI Blueprint or reach us at contact@theyor.com