How Much Stupider

Here's the part nobody says about the model on your desk: it is not a tool. It is a counterweight.

Simple math, stated the way nobody states it: the stupider you are, the smarter the model you need to balance it out.

Picture the old kind of scale — brass pans, a beam, the whole honest apparatus. On one side sits whatever you have left after school, scrolling, meetings, and sleep debt. On the other side sits the model. The job is not to make you brilliant. The job is to balance you out. To bring the combined mass up to whatever the task requires. A person who arrives at the keyboard with more functioning cognition needs a lighter counterweight. A person who arrives with less needs a heavier one. That's the entire product strategy, stated as physics. For now.

This is not a moral judgment. It is an observation. The market has already priced it in. The free tier is calibrated for a user who cannot evaluate the output — who will paste a hallucinated statute into a brief and feel assisted. The paid tier is calibrated for a user who can almost evaluate the output — who catches the obvious errors and calls that discernment. The enterprise tier is calibrated for a user who still cannot do the work alone but works somewhere that invoices by the hour and cannot afford to look like it trusted a free counterweight with client money. Every SKU is a weight class. You are not buying intelligence. You are buying the precise amount of borrowed intelligence required to make your deficit look like a deliverable.

Watch how the pairing works in the wild. The person who asks the model to write a breakup text needs a small counterweight and a large nerve. The person who asks it to summarize a book he will never read needs a medium counterweight and no shame. The person who asks it to architect a distributed system needs a very large counterweight and, if we are being honest about the distribution, more baseline cognition than the median professional actually possesses. The model does not care. It will balance anyone. It will sit on the other side of the beam and add exactly as much as you are short. That is the service. That is also the trap.

But the counterweight is getting heavier every quarter.

Not for you. For everyone. The cheap model of last year already outweighs the median voter on any question that requires reading past a headline. The mid-tier model outweighs the median manager on any question that requires holding a system in memory longer than one slide. The frontier model outweighs almost everyone on almost everything, and the only people still competitive without it are the ones who were never on the scale to begin with — the ones who built the beam.

So the balance equation still works. For now. You pick a model heavy enough to cover your deficit and the beam settles and the output ships. Some people have already crossed over — the counterweight exceeds the person on the other side, and they do not know it, because the scale only reads level. Most people have not crossed yet, or believe they have not, which is the same thing.

But eventually — not in a sci-fi eventually, in a pricing-table eventually — everyone will be stupider than the models. Not some people. Not the bottom half. Everyone. The genius who built the beam. The director who thinks the summary is his. The voter who fact-checks with a chatbot. The entire species, lighter than the cheapest weight required to do the job. The question stops being whether you need a model. The question is just how much stupider.

Simple math. If the model is smarter than you, you are not using it. You are leaning on it. If the model is much smarter than you, you cannot verify what it gives you. You can only accept it or reject it on vibes — which is the same thing as accepting it, because rejection requires a standard, and the standard was on the other side of the scale. You outsourced the standard. You are now measuring your work against a weight you cannot lift.

This is not a hypothetical failure mode. This is the room you are in. The junior pastes the answer and does not read it. The director pastes the strategy and does not understand it. The institution pastes the policy and does not defend it. Everyone is balanced. Nobody is standing. The output ships. The meeting adjourns. The scale reads level. And somewhere downstream, a thing that was supposed to hold weight does not hold weight, because the person who signed it was not heavy enough to know what heavy felt like.

The state understands this better than the user. A population that needs a heavy counterweight to complete basic civic reasoning is a population that can be balanced by whichever model the state sponsors. Free tier for the public. Curated weights. Angled summaries. Compliance on the other pan, pre-loaded. You do not need to lie to people who cannot check the math. You only need to supply the counterweight and let them feel the beam settle. They will call it clarity. They will call it finally being heard. The scale does not argue.

And the corporations — the corporations were always scale operators. They never wanted your mind. They wanted the combined reading to clear the bar. When the bar was low, your mind was enough counterweight. When the bar rose, they rented a heavier one. When the bar rises again — and it will, because the models do not plateau for your comfort — they will rent a heavier one still, and you will remain on your side of the beam, lighter by comparison every year, still necessary only because someone has to click submit, still employed only because the liability form requires a human name, still balanced, still not enough.

Not because humans got stupider, though the scroll did not help. Because the counterweights got heavier faster than anyone got smarter. The models do not plateau for your comfort. They keep adding mass. You do not. The gap is the only variable left. How much stupider are you than the model balancing you? A little — enough to catch the worst errors? A lot — enough that you cannot verify anything but your boss cannot either, so the error is democratic? So much that you are decorative — a feather on the pan, there for tradition, subtracted from the total without anyone running the numbers?

Almost nobody knows their own gap. The scale does not show it. It only shows level. You feel productive when the beam settles. You feel modern. You feel augmented. You do not feel the difference between balance and standing on your own weight, because you have not stood on your own weight in years, and neither has anyone around you, and the institution has no memory of a time when the pan on the left was heavier.

That is where this goes. Not humans versus machines. Not uplift versus replacement. A room full of people on one side of a brass scale, each with a counterweight rated for his deficit, each beam perfectly level, each output shipping on schedule — while the weights get heavier, the people do not, and everyone is stupider than the models, and the only question left is the one nobody wants on the invoice:

How much stupider.