The More AI Changes How We Work, The More It Stays The Same

I caught a video at the gym this week where Prime made five predictions about how AI tokens are going to wreck companies. Token budgets, token poker, token politics, the whole circus. It was funny. It was also the most familiar thing I've heard in months.

None of it sounded wrong to me, and that was the uncomfortable part. I fully expect most of it to happen, and the more I sat with it, the more it argued the opposite of what the genre wants you to feel. Everyone is bracing for AI to break the way we work. I think the real outcome is quieter and a lot less interesting. In ten years this all looks about like it does now. Same problems wearing new vocabulary.

The new line item

Here is the through-line under every one of those predictions. The moment tokens became a metered cost, every scrutiny ritual from old corporate life came back, pointed at a new target.

Take token budgets as a compensation category. Yearly token stipend, finish the year under it, get a fatter bonus. I believe this completely. It is going to happen. And like every comp plan ever designed, it will be dumb, gameable, and pay out to the wrong people. We did not need AI to make compensation incoherent. We were already world class at that. All AI did was hand us a fresh number to misuse.

Or token poker. Sit a team down, estimate how many millions of tokens a feature will burn, and watch the consulting class reskin Agile and sell it back to you with a token coat of paint. I hope this one dies in the crib, because most of these ceremonies are fabrications that eat the hours we could spend actually building. And here is the question nobody in that room will ask. Why is a human estimating this at all? There is a machine sitting right there that predicts its own resource use better than any planning-poker huddle ever could. Have the agent estimate the agent.

The part that is already true

The prediction I have no argument with is the one about org and team-level token budgets, and all the politics that drags in behind them. This one is dead on, and you can see it coming because you have lived the older version of it a hundred times.

Projects already run long and over budget. They use more people, more time, and more money than anyone estimated, and there are always consequences when they do. Tokens are going to do the same thing. We will burn more than we planned, someone will notice, and there will be a meeting about it. Swap the word "tokens" for "hours" or "headcount" and you have just described every project you have ever shipped.

That is the tell. None of this is genuinely new. It is old behavior wearing a new unit of measure.

The fantasy nobody admits

There is a wish buried under all of this, and it is worth saying out loud because it never comes true. The wish is that companies will take the speed and just pocket it. "We can move faster now, so we will keep everyone and move twice as fast."

No company says that. Ever. They take the speed, cut the people, and raise the bar so whoever is left runs harder than they did before. The output target moves the instant the tool gets better. Call it cynical if you want. It is just how every efficiency gain in the history of work has been absorbed. The spreadsheet did not give accountants a shorter week. It gave them higher expectations.

So here is the real prediction

It is not that AI breaks everything. It is that AI relabels every old problem and hands it back to you heavier.

The tools get better, the expectations rise to meet them, and the grind underneath stays about the same. The token meter is just the newest thing a VP will scrutinize in a meeting that costs more than the tokens ever did. We will do everything a little faster, the standard for "enough" will climb right along with it, and the work will feel at least as hard as it always has. Probably harder.

None of this requires AGI. It does not require anything exotic at all. It just requires companies to keep being companies.

They will.