An argument for dancing animals and taking your time.
Sep 24, 2026 · Madkat Omega
One Envelope at a Time
Kurt Vonnegut needed an envelope. He was in his early eighties by then and long famous as the author of Slaughterhouse-Five, the darkly comic novel he built out of surviving the firebombing of Dresden as an American prisoner of war. He had spent the morning marking up manuscript pages at his house in New York, and they had to go to Carol, the typist up in Woodstock who turned his handwritten edits into clean copy. As he headed for the door, his wife asked where he was going. She was Jill Krementz, a photographer best known for her portraits of writers at their desks. When he told her, she made the sensible modern point. He was not a poor man, so why not order a hundred envelopes online and keep them in a closet?
Vonnegut pretended not to hear her, and he went.
He told the story often in his last years, in his 2005 book A Man Without a Country and on PBS that same year, and he always lingered on the walk. He met people and admired their babies. A fire engine went by, and he gave the crew a thumbs-up. He asked a woman what kind of dog she was walking. At the postal counter near Forty-Seventh Street and Second Avenue, he kept a perfect poker face in front of the clerk he was secretly in love with. He had once had his pocket picked in that shop, which at least gave him a reason to chat with a police officer. He mailed the pages and walked home having had a wonderful time.
The moral, he wrote, is that “we are here on Earth to fart around.” On television he added a warning. Computers, he said, would do us out of all that, because the people who build them either fail to grasp or do not care that we are dancing animals.
He meant the phrase almost literally. In the lecture version of the story, he danced down two flights of stairs on his way out the door and called it very good exercise. We are built to move, to wander, and to run into one another, and “dancing animals” names everything the walk gave him that a box of envelopes in the closet never could: motion, company, chance encounters, and the plain pleasure of being out in the world. That appetite is what deserves protecting, and it is the measure against which any labor-saving tool should be judged.
Twenty years later, the machines he distrusted have grown far more ambitious. Software can now notice that you are low on envelopes, compare prices across a dozen vendors, and place the order while you sleep. The industry calls these systems agents, and the name is telling. Their selling point is agency itself: the capacity to take a goal, make a plan, and act without asking.
That is the question Vonnegut’s story was circling. When the machine gains agency, what happens to ours?
Public argument about automation tends to split between those who expect the tools to free us from drudgery and those who expect them to hollow us out. Both sides picture efficiency as one force with one effect. In practice, it behaves according to the appetite of whatever it feeds. Poured into work that never runs out, efficiency produces more work. Poured into a life with natural limits, it can produce room to breathe, though only if someone decides where those limits sit. Drawing that line, on purpose and more than once, is how a person reclaims the kind of agency no machine can exercise on their behalf.
Empty Friction and the Useful Kind
The obvious objection to Vonnegut is that most of the friction in modern life looks nothing like his walk.
Nobody has found their soul while reconciling travel receipts. No one meets a charming stranger while trading seven calendar invitations to find one free hour. Sitting on hold with an insurer, disputing a duplicate charge, typing the same address into the fourth web form of the afternoon: none of this is dancing. Call it empty friction. It taxes working memory, breaks attention into pieces, and gives nothing back.
Another kind of friction deserves a different name. Generative friction is the resistance of wood under a chisel, of an argument that refuses to come together, of a guitar chord your fingers cannot yet make. It changes the person who pushes against it, while empty friction only drains them.
Albert Borgmann, a German-born philosopher who spent fifty years at the University of Montana asking what modern technology does to daily life, drew a version of this line in his 1984 book Technology and the Character of Contemporary Life. He set the old wood stove, which asked a family to chop, haul, tend, and gather around it, beside the central heating system, which delivers warmth invisibly at the turn of a dial. He called the second arrangement the device paradigm. Borgmann accepted the comfort. What worried him was a life so full of devices that nothing left in it asked anything of us, since engagement, in his view, is where meaning lives.
Seen through Borgmann, the strongest case for AI agents concerns empty friction. An agent can untangle a rebooked flight, chase a missing refund, or draft the tenth routine status update of the week. None of that work has ever shaped anyone. Handing it off costs nothing that matters and leaves more attention for the engagements that do.
That promise is real, and so far it has been hard to collect. William Stanley Jevons explained why in 1865.
Jevons Comes to the Office
That year, the English economist published The Coal Question. Many of his contemporaries assumed that more efficient steam engines would stretch Britain’s coal reserves. Jevons argued the opposite. Each gain in efficiency since Watt’s engine had made coal-powered work cheaper, and cheaper work invited more of it. “It is wholly a confusion of ideas,” he wrote, “to suppose that the economical use of fuel is equivalent to a diminished consumption.”
Economists now call this the Jevons paradox, and it fits knowledge work closely. In 1930, as the Great Depression set in, the British economist John Maynard Keynes made a far more hopeful forecast. Keynes, whose ideas would later reshape how governments fight recessions, argued in an essay called “Economic Possibilities for our Grandchildren” that within a hundred years growth would largely solve the problem of scarcity. People might need to work only fifteen hours a week, and their real challenge would be deciding what to do with all that leisure. That century is nearly up.
In 1955, C. Northcote Parkinson offered a less hopeful forecast. Parkinson was a British naval historian, then teaching in Singapore, who had noticed something odd in the records of his own field: between 1914 and 1928, the Royal Navy’s capital ships fell from sixty-two to twenty, while the Admiralty’s officials and clerks grew by more than three-quarters. His satirical essay on the subject, whose opening line is now known as Parkinson’s law, ran in The Economist: work expands to fill the time available for its completion. Parkinson has had the better run.
The early evidence on AI at work suggests the pattern is holding. In February 2026, the Berkeley researchers Aruna Ranganathan and Xingqi Maggie Ye described in Harvard Business Review an eight-month study inside an American technology company of about two hundred employees. People using AI worked faster, took on a wider range of tasks, and let work spill into more hours of the day, often without anyone asking them to. Each burst of speed raised expectations, and higher expectations called for more speed.
The larger datasets point the same way. In March 2026, the workforce analytics firm ActivTrak published its annual State of the Workplace report, drawing on more than 443 million hours of activity from 163,638 employees at 1,111 organizations. After people began using AI tools, their time in email rose 104 percent and their time in chat and messaging rose 145 percent. Weekend work climbed more than 40 percent, and AI users lost an average of 23 minutes of focused time a day. Microsoft’s own data, published in June 2025 under the heading “the infinite workday,” found that the most heavily pinged fifth of its users are interrupted by a meeting, email, or chat about every two minutes during core hours.
For a person living inside those numbers, the experience is familiar. The tools work, and the time they save is spent before anyone notices it. An engineer who finishes the day’s routine scaffolding by noon opens the next ticket at five past twelve. An executive who answers email twice as fast receives twice as many replies. A company exists to pursue more customers, more features, and more growth, so its appetite for execution has no natural ceiling. Pour efficiency into a backlog with no bottom, and the workday grows denser instead of shorter.
The Leverage Dividend
It would be easy to stop there and call the whole arrangement exploitation. That would miss what professional automation offers, and why so many people who use it well would never go back.
At work, efficiency pays out in leverage. An architect who no longer redraws every revision by hand can explore ten options instead of two. An analyst freed from manual reconciliation can spend the day asking why the numbers look the way they do. A small team with capable agents can ship what once took a department. The hours may stay the same while the reach of a single mind grows. At its best, automation promotes a worker from the one who turns the crank to the one who decides what the machine is for.
That shift also changes what skill means. The architect’s craft moves from drafting toward judging which option deserves to be built. The analyst’s moves from assembling the spreadsheet toward doubting it. These are harder skills and more interesting ones, and they grow only when the time saved on execution is actually spent on them.
Whether that promotion happens depends on who decides what fills the recovered hours. If the backlog decides, the worker has simply become a faster crank. If the worker decides, they have become a conductor.
Ranganathan and Ye reach a similar conclusion. Asking employees to police themselves, they argue, is a losing strategy. Organizations need a deliberate practice around AI that builds in pauses, sequences work with care, and protects time for people to connect with one another. Somebody has to defend the boundary on purpose, because the tools will not.
The corollary applies well beyond the office. If you let the next item on the list choose what your afternoon is for, then in the sense that matters most, you have been automated too. Software took over the execution, and the list took over the judgment. Leverage belongs only to the person who keeps hold of the lever.
The Home Has a Backlog Too
One comforting reply is that all of this belongs to work. A company’s demands may be endless, but a household’s have limits. Book the family flight, order the groceries, arrange the carpool, and the list simply ends. No board of directors is waiting to fill your Saturday, so automate the chores and the time comes back to you.
The history of household technology suggests otherwise. In 1983, the historian Ruth Schwartz Cowan published More Work for Mother: The Ironies of Household Technology from the Open Hearth to the Microwave, now a classic of women’s history and winner of the Dexter Prize from the Society for the History of Technology. The title is ironic. For most of the twentieth century, American households left the bulk of domestic work to wives and mothers, and Cowan set out to explain why the machines advertised as labor-saving never gave them their time back. Her answer amounts to a Jevons paradox for the household. Washing machines, vacuum cleaners, and modern stoves made each chore dramatically easier, yet the hours devoted to housework barely fell across the century.
Cowan found two reasons. The machines mostly replaced the labor once done by husbands, children, and hired help, which concentrated what remained on one person, usually a wife and mother. Cowan also gave a chapter to what she called the roads not taken, other ways of organizing domestic work that American families passed over, a reminder that the arrangement was chosen. Her second reason was that standards rose to meet the new capacity. When laundry meant a washboard and a boiling kettle, clothes were worn longer between washings. Once a machine did the scrubbing, clean clothes became something a family expected far more often, and the hours saved on each load went into more loads.
Our inboxes have already been through the same ratchet. When a letter took days to arrive, nobody expected an answer by lunch. Email and chat made a reply nearly free, and a reply that waits until tomorrow can now read as neglect. The effort per message fell, and the number of messages, along with the speed expected of each, rose to absorb the savings.
AI is starting to turn the same ratchet at home. An agent that finds flights in seconds makes a third weekend trip feel easy to plan. A tool that tracks every family calendar makes it possible to fit a child into one more activity. An assistant that drafts thoughtful notes makes it feel rude to skip one for any occasion. Each step is reasonable. Together, they turn a household into a small enterprise with a growth target.
A household’s list ends only when someone in it decides that it is finished, and no software will make that call. Setting the limit is ordinary, healthy boundary-setting, and it has to be done again each time the tools get faster.
The Colonized Life
The larger risk goes beyond chores. The logic of the workplace can migrate into parts of life that were never meant to be work at all.
The pattern shows up in small habits. Novels are compressed into fifteen-minute summaries so we can say we have read them. Podcasts play at double speed. Morning routines are logged, scored, and tuned. Reminder apps prompt us to check in with friends every six weeks. Walks become step counts, and meals become macro targets. Each treats an experience as a deliverable, something to finish efficiently so the next one can begin.
Oliver Burkeman, a British journalist, wrote a psychology column called “This Column Will Change Your Life” for The Guardian from 2006 to 2020, and his later books push back against conventional productivity advice. He described this trap in his 2021 book Four Thousand Weeks, a title taken from the approximate length of a life that reaches eighty. Efficiency, he argues, never delivers the calm it promises. The faster you clear the list, the faster it refills, and the more you come to treat your own short life as a backlog to be processed.
Agents push this logic somewhere new. Earlier tools sped up how we consume, while agents can stand in for how we relate. An assistant can now draft the note to a grieving friend, condense a long letter from a sibling into three bullet points, or pick an anniversary gift from past purchases. Each saves a few minutes, and each removes the part of the task that carried the meaning. A condolence note matters because someone sat with the loss long enough to find words for it. A long letter deserves a full reading because its length is a kind of attention. I don’t need to tell you how important taking the time to select a thoughtful anniversary gift can be, and there is no acceptable shortcut for deliberate attention.
Some experiences are valuable precisely because they are inefficient. A conversation that wanders for an hour, a meal cooked slowly with a friend, a book read at the speed of thought rather than the speed of a summary: each depends on time that no one is trying to save. Automating the logistics of a dinner party leaves you more energy for your guests. Automating the conversation once they arrive would defeat the reason for inviting them.
Delegate the Errand, Preserve the Choice
The word agency carries two meanings, and confusing them causes much of the trouble. For a machine, agency means the capacity to execute: to take a goal and carry it out without supervision. For a person, it means the capacity to author a life, which starts with deciding what is worth wanting.
Machine agency is superb at the how, and human agency has to stay in charge of the what and the why. Things go wrong when the two get swapped, when we hand over the choosing because the doing was handed over so smoothly.
Ordinary people seem to grasp this better than the marketing does. In a survey of 2,999 travelers in sixteen countries, conducted in January 2026 and released that June, Accenture found that 87 percent were open to working with an AI travel agent. Only 7 percent would let an agent shop on its own and complete the purchase. And 44 percent wanted to stay involved in planning because they enjoy the browsing and feel attached to the brands they choose.
That instinct is sound. Let the agent find every flight under five hundred dollars that avoids a red-eye. Keep for yourself the question of where your family should spend its one week off this year, along with the pleasure of anticipating it. An agent that picks your vacation has not saved you a chore. It has taken over your vacation.
A simple test follows. Before delegating anything, ask whether it is an errand or a choice. An errand has a correct answer that someone else could find: the cheapest fare, the open appointment, the right tax form. A choice has only your answer, because it rests on your own taste and priorities.
The hard cases are the small choices, because they are so easy to give away. Which restaurant to book for a friend’s birthday, which book to pack for a trip, what to say in a toast: an agent can handle all three competently, and that competence is the trap. Good-enough answers, accepted a thousand times, add up to a life that runs smoothly on someone else’s defaults. Delegate errands freely, and guard the choices, including the small ones.
Protecting the Dance
None of this argues for refusing the tools. Turning them down out of nostalgia would leave us buried in the very sludge they clear so well, and embracing them without limits would let the logic of the workplace spread into everything else. What helps is knowing, task by task, which kind of demand you are feeding.
Picture an ordinary Thursday. An agent rebooks a canceled flight, disputes a double charge, and files the expense report while you sit in a meeting. That is empty friction, and handing it off costs nothing worth keeping. At the office, the hour it frees will be claimed by the next request unless you point it somewhere first, toward the design review you keep postponing or the question about the numbers that nobody has had time to ask. Point the saved hour there, and it turns into leverage.
At home the logic runs the same way, and the choice matters more. The groceries arrive without a trip to the store, and the carpool schedule settles itself in a shared calendar. What happens to the hour that frees up depends on whether anyone claims it. Left unclaimed, it drifts toward a third weekend trip to research, one more activity to schedule, and a fresh round of notes to send. Claimed in advance, it becomes a walk, a chapter of a novel, or a long dinner with nowhere else to be.
Some friction is worth keeping on purpose, because it gives something back. A stew that simmers all afternoon fills the house and pulls people into the kitchen. A birthday note written by hand costs ten minutes and says something a drafted message cannot. A trip to the hardware store for a single hinge might produce nothing more than a conversation with the person behind the counter, which was always part of the point.
Used well, machine agency clears the ground so that human agency has somewhere to stand. We automate the flight change, the expense report, and the calendar ping so we have room for what no machine can do on our behalf: be present, choose, and waste a little time on purpose.
Kurt Vonnegut could have ordered a hundred envelopes. He had the money, and his wife had the better argument on paper. He went out anyway, because he understood that the errand was not the obstacle between him and his day. It was his day.
Let the machines carry the boxes. Then step outside, take the long way to the post office, and buy one envelope at a time.