John Rector · 11 November 2026 · An essay on shared time
The Hour We Have in Common
AI can give you an evening back.
Whether anyone can join you is a different achievement.
Imagine four friends whose personal AIs have each found them three free hours on Thursday. Twelve hours recovered. Four excellent reports. No dinner.
Anna is free early. Ben is free a little later. Cleo can leave after that. By the time Dev finishes, Anna has another commitment. Everyone has gained time. The four of them still cannot meet.
We have become very good at describing freedom as something that happens inside one person’s calendar. A meeting disappears. An errand is handled. A document gets written while its owner sleeps. The empty space looks like the result.
But an hour has a second property beyond its length. It has a position. Move it far enough from the people you love and it becomes a different kind of hour.
This will matter more as AI takes on the arrangements of daily life. My prediction is that one of the most valuable things a personal AI will do is help people become available to one another. The achievement will be measured in evenings that actually happen.
01 / The arithmetic of an evening
A dinner lives in the overlap.
Some time is useful precisely because it is yours alone. Reading, sleeping, walking without explaining where you are going: solitude needs no coordination committee. More private time is a real gain.
A dinner with four friends has a different structure. Each person must be available during the same interval. Adding their free hours tells you how much time they possess collectively. It does not tell you whether they can sit down together.
The distinction is between a sum and an intersection. Twelve separate person-hours can contain zero hours when all four people are free. Rearrange those same hours and they can contain a whole evening.
A small experiment
Same hours.
Different evening.
Choose an arrangement. Every person keeps three free hours. Only their timing changes.
An illustrative schedule, not a statistic. This rearrangement assumes the affected commitments can move by agreement. A shared interval makes a gathering possible; it does not oblige anyone to attend.
02 / A calendar can separate people
The day off that everyone had separately.
In September 1929, the Soviet Union introduced a continuous five-day work cycle for industrial and office workers. The plan divided workers into five groups. Each group worked four days and rested on the fifth, with different groups resting on different days. The schedule could keep production running while individual workers took their turns away. Researcher Maria Yushmanova examines the social reception of this arrangement in her research on the continuous workweek.
The structural problem is easy to see. A person could have a rest day while a spouse or friend was at work. Rest remained on the calendar. A common rest day was no longer built into the arrangement.
This is what a calendar can conceal. Viewed from inside a factory, staggered availability looks useful. Viewed from a kitchen table, it can make an ordinary visit difficult. The same schedule has different properties depending on whose activity you are trying to make possible.
Today’s personal AI is pursuing a very different purpose from that industrial system. But any system that arranges people’s time can encounter the same mathematical fact: improving the parts does not necessarily improve the overlap.
03 / Why “when” has value
Your free hour changes mine.
Economist Daniel Hamermesh found that spouses’ work schedules in U.S. data were more synchronized than random pairing would predict. His study, “Timing, Togetherness and Time Windfalls,” treats the placement of time as something that deserves its own explanation.
In separate research, Hamermesh, Caitlin Knowles Myers, and Mark Pocock studied how television schedules and time zones affect daily routines. Using the American Time Use Survey, they found evidence that people adjust the timing of work and sleep partly to coordinate with others. The study of timing and coordination helps explain why a schedule is never entirely private in its effects.
My inference for AI is straightforward. If a friend becomes free when I am free, something about my hour improves even though its duration has not changed. The benefit crosses from one calendar into another. A system that reports only what it saved its own user cannot see the whole result.
This is familiar without any economics. The difficult part of a reunion can be finding a date. A choir needs its singers together. A parent’s unexpectedly free afternoon is more useful for a school performance if it happens to be the right afternoon. Time carries the possibility of other people.
04 / Give the delegate a better purpose
Protect the evening, then arrange the errands.
Return to the four friends. This time, they have each told their AI that dinner together matters. Within the authority each person grants, the delegates compare acceptable windows and identify which commitments are actually movable.
Anna can move a delivery. Ben’s evening already fits. Cleo can move an errand without changing her appointment. Dev can finish an administrative task earlier. The useful proposal emerges from those particulars. It does not require everyone to become infinitely flexible.
The exchange can be modest. “Thursday, six to nine, is available if confirmed by Tuesday” is enough to offer a window. Nobody needs to distribute a complete calendar or disclose why another evening is unavailable.
Once the people agree, the common interval becomes a commitment. This matters because an agent continually improving one calendar could otherwise dismantle the arrangement: a more convenient delivery for Anna, a better appointment for Ben, and the dinner disappears one individually sensible change at a time.
The instruction must therefore survive the optimization: this is time we have chosen to share. A proposed change has to account for the people who are relying on it. An empty square may already contain a promise.
That also gives the agent a more useful definition of completion. Finding a slot is one step. There must be enough time to get there, somewhere to meet, and a plan that still works after an ordinary delay. A three-hour opening with a two-hour journey at each end is an arithmetic success and an evening that cannot happen.
05 / The next social invention
A little time held in common.
Once an evening works, a further possibility appears. The friends can choose to keep it. The first Thursday becomes a standing invitation, with room for absence and a way to change the arrangement together.
Recurrence changes the burden. Instead of making a fresh negotiation prove the value of the friendship every month, the agreement carries that value forward. Other tasks arrange themselves around something the people have already decided is worth doing.
A neighborhood can do this with an open supper. A family can do it with a weekly call across time zones. A workshop can keep an afternoon when people bring unfinished things and know someone else will be there. These are small institutions built out of shared expectation.
They need not use the same day. A universal weekend has never fit every shift, every duty of care, or every tradition. The point is to make common time possible for particular people, including those whose schedules give them the least room to negotiate. Some constraints require an employer, a service provider, or another person to change an arrangement. An AI cannot manufacture consent or staffing by moving a rectangle.
There is freedom in being able to decline, too. A good delegate should protect a solitary evening with the same seriousness as a dinner. Shared time has value because people choose it. Attendance extracted by an efficient system defeats the purpose.
Thursday · 6:00 p.m.
Four chairs.
One evening.
At six, the four friends arrive. There is no report on the table explaining which tasks were moved or how many messages the delegates exchanged. Someone is late enough to be teased. Someone has a story that takes longer than expected. The food cools a little.
None of that appears in the original calculation of hours saved. It is what the calculation was supposed to make room for.
AI can do much of the work that gets us to the table. The life we wanted begins when we sit down.