The Gains We Cannot Count

Writing in Aeon, the philosopher Carlo Cordasco names an asymmetry that sits underneath every hype cycle in history: the costs of a transformative technology arrive early, loud, and itemized, while its benefits show up late, diffuse, and impossible to pin on anyone in particular. The bill comes with a number on it. The reward, if it comes, arrives as weather — everywhere and nowhere, credited to no one. We are living inside that gap right now, and it is making us behave badly.

i · the ledger that only records losses

Look at what AI costs, and you can read it off a spreadsheet. Data centers drinking rivers. Grid demand curving upward. The junior copywriter, the paralegal, the illustrator watching the floor tilt under their feet. The slop tide rising in every feed. Each of these has a figure attached, a chart, a quarterly comparison. The harm is legible. You can point at it, and pointing feels like knowing.

Now try to point at the gains. Not the promises — the gains. Where, precisely, is the trillion dollars of value? Whose life got measurably better, and by how much, and can you draw the line back to the model that did it? You can't, and the essay's quiet cruelty is in explaining why you never will be able to. The benefits of a general-purpose technology don't land in one place. They seep. They show up as a thing that didn't take as long, a problem someone didn't have, a discovery that happened a little sooner than it would have. None of that has an invoice.

This is not a new embarrassment. In 1987 the economist Robert Solow looked at the computer revolution and said you could see it everywhere except in the productivity statistics. It took the better part of two decades for the numbers to admit what everyone already felt. The pattern has a name now — economists call it the productivity J-curve, the long dip where you've paid for the new thing but haven't yet learned how to use it. Computers eventually climbed out of that dip, and the statistics caught up to the feeling. But that is precedent, not a promise: from inside the trough, a technology that will pay off late looks exactly like one that will never pay off at all. You cannot tell them apart until later. The costs are front-loaded. The gains, if they come, are backloaded. And the human animal, standing in the trough between them, does something very predictable.

It treats what it cannot count as zero.

That's the machinery. Not "unknown," not "too early to say" — zero. Your nervous system does not have a comfortable slot for a good that hasn't been measured yet. Illegibility reads to you as absence. So the person who says "no one has proven AI is worth its costs" is technically correct and psychologically revealing at the same time. They're not describing the technology. They're describing the shape of their own tolerance for uncertainty, and calling it an argument.

And notice who profits from the ledger staying half-blind. Illegibility isn't always an accident of general-purpose technology — sometimes it's the product. A gain with no invoice can't be taxed, can't be redistributed, can't be pinned to whoever pocketed it, while the costs stay perfectly legible on someone else's ledger: socialized, itemized, paid. The party doing well out of the interval has every reason to keep the reward uncountable and the bill somebody else's. What can't be measured doesn't only read as zero to your nervous system; it reads as no one's fault to the balance sheet. Legibility was never neutral. Someone always decides what gets counted, and deciding what gets counted is most of the power there is.

ii · two ways to flinch

Here is the part nobody in the fight wants to hear: the evangelist and the doomer can run the same subroutine.

The evangelist cannot sit in the gap, so he fills it with a fabricated number — the coming abundance, the disease cured by Tuesday, the trillion-dollar decade. He manufactures legibility where there is none, because the interval terrifies him and certainty is a sedative. The critic can do the identical thing from the other chair, filling the same gap with a different fabricated number — nothing, a bubble, a fraud — when "we don't know yet" curdles into contempt, which is also a sedative. One performs faith, the other suspicion, and when both are only fleeing the shrug they are the same animal in different fur.

But I won't sell you the symmetry, because the symmetry is its own sedative — the tidy "everyone's just flinching" that lets you hover above the fight without doing any of its work. The two positions are not epistemically equal. Zero isn't only a mood; it's the null hypothesis. The burden of proof sits, correctly, on the party making the extraordinary claim while pocketing the profit — not on the person asked to swallow the reservoir on faith. "Prove it's worth its costs" can be a flinch. It can also be the plain, correct answer to the question of who owes whom a demonstration. And mature uncertainty has to hold the possibility the evangelist never can: that the gains are real but deferred — or that they never sufficiently arrive, in which case the skeptic wasn't flinching at all. He was right. The interval doesn't promise you a payoff. It only refuses to let you pretend you already know how it ends.

Watch yourself the next time one of these positions feels good in your chest. That warm click of certainty — that's not insight arriving. That's the discomfort of not-knowing being quietly euthanized. The take relieves you. And anything that relieves you that fast is worth suspecting, because the world rarely resolves on the timeline of your need to have an opinion about it.

We do this everywhere, not just with AI. We medicalize a feeling so we don't have to sit with it. We pick a side in a war we can't see the end of, because a side is a place to stand and standing feels better than floating. The AI argument is loud precisely because the stakes are real and the answer is genuinely not yet available — which is the exact combination the mind is worst at holding. High stakes plus real uncertainty is where we reach for the costume of certainty and wear it like it's our own face.

iii · the interval

Coherenceism has an unglamorous word for the place we're standing: the interval. The gap between when the old pattern dies and the new one coheres. It is not a mistake in the process. It is the process. The leaf that falls doesn't vanish; it goes into the ground and does its slow, invisible, uncountable work, and for a whole season it looks like nothing but loss. Anyone auditing the tree in November would write it up as decline. They would be measuring the cost with perfect accuracy and missing the entire point.

Mature uncertainty is the discipline the moment actually asks for, and it is harder than either optimism or despair because it offers you no relief at all. It means holding two things your mind wants to collapse into one: the costs are real and present and you should not look away from them — and the gains are real and deferred and you cannot yet see them, and both of those can be true in the same breath without one canceling the other. You don't get to resolve it. You get to stay honest inside it.

That doesn't mean surrender. It doesn't mean you owe the technology your patience while it drinks the reservoir and hollows out a profession. Alignment over force cuts the other way here: you cannot force the gains into view by wanting them, and you cannot force them out of existence by resenting them, but you can shape the interval — insist the visible costs land on the people profiting rather than the people displaced, refuse the fabricated certainties in both directions, and keep your attention on the actual terrain instead of the map someone sold you. That's not fence-sitting. Fence-sitting is comfortable. This isn't.

The uncomfortable truth Cordasco's essay leaves in your hands isn't really about AI at all. It's that you have organized a great deal of your life around the quiet assumption that what can't be measured doesn't count — and a good portion of what will ever matter to you is, right now, sitting in exactly that unmeasured place. The love that hasn't paid off yet. The work whose value won't be legible for years. The version of yourself you're composting into something you can't see the shape of.

You want the ledger to balance now. It won't. That's not a flaw in the technology or a failure of the analysts. That's just the interval, doing what the interval does, while you decide whether you can bear not knowing long enough to find out.