A week ago most companies posted some version of the same message: thank you to our hardworking team, enjoy the long weekend.
It's a kind sentiment attached to a holiday that means nearly the opposite. Labor Day was not given to American workers as thanks. It was taken, slowly, in a fight, one state at a time.
The Department of Labor's own history lays out the sequence plainly. Oregon passed the first state law recognizing the holiday on February 21, 1887. Four more states followed that same year, then three more by the end of the decade. By 1894, twenty three additional states had adopted it, and only then, on June 28, 1894, did Congress pass an act making the first Monday in September a legal holiday.
Congress went last. That detail is the whole holiday.
The dividend always had to be claimed
Here's the pattern worth noticing once the long weekend is far enough behind us to think straight about it.
Every major productivity wave of the last century and a half produced a surplus, and in each case some portion of that surplus was converted into human time. The ten hour day, then the eight. The two day weekend. Paid leave. Not one of those conversions happened because productivity rose and management passed the savings along. Output per hour climbing does not automatically become hours falling. It becomes more output, unless somebody intervenes.
The gains were real. What happened to them was a decision.
Keynes was right about the math and wrong about us
In 1930, John Maynard Keynes wrote Economic Possibilities for our Grandchildren, predicting that a century of compounding productivity would leave his grandchildren working about fifteen hours a week. His line was that "three hours a day is quite enough to satisfy the old Adam in most of us."
We are now four years from his deadline. The productivity arrived roughly as he forecast. The fifteen hour week did not.
That's not a story about Keynes being naive. He got the economics right and the sociology wrong. He assumed a surplus would be spent on leisure because leisure is obviously valuable. Instead it was spent on more, because more is what a system optimizes for when nobody specifies otherwise. The dividend showed up on schedule and got absorbed.
Which brings us to the one currently landing.
What this dividend is being converted into
Watch what happens when an AI deployment works at your company. The hours saved rarely become hours. They become expanded scope, or a role not backfilled, or a quota raised at the start of next year to reflect the new capacity. Every one of those is a legitimate business choice. None of them is inevitable, and almost none of them is being made deliberately. They're defaults, and defaults are what you get when nobody decides.
The objection is that giving time back costs output. That has actually been tested now, which is more than can be said for most claims in this argument.
A 2025 study in Nature Human Behaviour followed 2,896 employees across 141 organizations in six countries through six month trials that cut hours without cutting pay. Burnout fell. Job satisfaction, mental health and physical health improved, with the largest gains among people who cut the most hours. In the UK pilot run out of Cambridge, revenue across reporting organizations was roughly flat, up 1.4% on average, turnover dropped 57%, and 92% of the companies kept the change after the trial ended.
Be honest about what that evidence is: organizations that volunteered, short horizons, sympathetic researchers. It is not proof that a shorter week works everywhere. But it comfortably clears the bar it needs to clear, which is that converting a productivity dividend into time does not automatically damage the business. The tradeoff most executives treat as settled is not settled.
Most of our readers are both sides of the table
There's a version of this that matters more for the people I actually talk to, who are usually founders, solo sellers, small teams, with nobody to bargain with because they are management and labor in the same chair.
For them the conversion isn't negotiated. It's decided every week by default, and the default is brutal: any hour AI gives back gets immediately refilled with more pipeline, more outreach, more accounts worked. The capacity gain is real and it vanishes on contact, because there is no counterparty whose job it is to claim it.
That's not discipline. That's just the same absorption Keynes missed, running at the scale of one person.
So here's the Labor Day thought, arriving late and meant seriously. The workers who won this holiday did not ask whether the productivity gains of the industrial era were real. They were. They asked a better question: who decides what they turn into? Then they answered it themselves, state by state, for seven years, until Congress caught up.
The gains from this wave will be argued about for a decade. What they become is being decided right now, mostly by nobody, mostly by default.
Somebody should decide on purpose.