The leverage collapse

Ordinary people have never had much power. What they have had is leverage.

The mechanism#

The leverage was always the same thing in different clothes. Factories needed hands. Armies needed bodies. Bureaucracies needed clerks who could read. None of those institutions wanted to share anything, and all of them did, because the alternative was that the work stopped.

Universal education was not a gift. It was a purchase. Governments bought literate populations because literate populations were the input to industrial power. This is the mechanism underneath almost every gain ordinary people made in the last two centuries. Not persuasion, and not moral progress, but the simple fact that the people at the top needed the people at the bottom to show up.

The intelligence curse argument is that AI dissolves this dependency. If cognitive work can be bought by the token, the institution no longer needs the clerk. If physical work follows, it no longer needs the hands either.

The thing to notice is that leverage does not degrade smoothly. It collapses at a threshold. A firm that can automate ninety per cent of a role still needs the remaining ten per cent, and those workers keep most of their bargaining power, sometimes more of it. A firm that can automate the whole role needs none of them, and their bargaining power goes to zero at once. The curve is a cliff.

What would have to be true#

The automation has to be broad rather than deep. A technology that replaces one occupation shifts workers into others, as every previous wave did. A technology that replaces the general capacity to do cognitive work has nowhere to shift them to.

The replacement has to be cheap enough to beat subsistence wages. This sounds like a high bar and is not. Inference costs have fallen by orders of magnitude on every capability level that has been held fixed long enough to measure.

The owners of the replacement have to be few. If everyone owns a model, the leverage reappears in a different form, because capability itself becomes the thing people hold.

What would count as evidence against it#

Real wage growth concentrated in occupations with high AI exposure, sustained over several years. That would suggest the technology is complementing labour rather than substituting for it, in which case leverage survives in a modified form.

Rising union density or strike effectiveness alongside rising AI exposure. The leverage story is wrong somewhere if workers are winning at the exact moment their position was supposedly becoming hopeless.

The emergence of a new scarce human input that is neither cognition nor muscle, which labour then organises around. I cannot rule this out, and I notice that every previous wave produced one.

Why it is listed separately#

Every other problem in this series is downstream of this one. The rentier pivot happens because the leverage is gone. The broken ladder matters because it forecloses the route back. The loyalty substitute is this same mechanism applied to the specific case of coercion.

If leverage survives, the curse does not land. If it does not, nothing else in the series is optional.