Tomorrow's tech news, today's hangover.

The Company Drafted the Living

The laundromat had three broken dryers and one working television bolted high in the corner like a prison guard with weather reports.

A woman in a faded sweatshirt stood in front of the machines holding a basket of wet clothes and staring at the little red OUT OF ORDER signs. She did not curse. That was how I knew she had passed the early stages of disappointment and entered the adult kingdom of silent calculation. Quarters. Time. Bus schedule. Some kid waiting somewhere. The arithmetic of being alive.

A man beside me said, “They ought to automate this place.”

I looked at him.

The dryer ate another quarter and did nothing.

“Looks automated already,” I said.

This is the trouble with our beautiful age. We keep mistaking neglect for progress. We put a screen on the wall, a sensor in the door, an app on the corpse, and then everyone in a clean shirt tells us the future has arrived. Meanwhile, the clothes stay wet.

Out in Menlo Park, the people who built the apps are learning what the rest of the species already knew. A bad job is still a bad job when it comes with stock options and a badge that opens glass doors.

Meta has a new AI unit with thousands of engineers and product people shoved into it like furniture in a storage locker. Applied AI. That is the name. It has the dead charm of a government office and the digestive confidence of a machine that expects to be fed. Some workers say they were drafted. Join the unit or leave the company. Others describe the work as mechanical, menial, soul-crushing. They are generating puzzles and tasks and little problem pellets for the models to chew on, week after week, while the bosses speak of superintelligence like priests describing heaven to the men digging the cathedral foundation.

One employee interrupted a company livestream and asked the presenters to tell an executive he was a piece of shit.

There it is. The human interface.

Not the glossy one in the demo. Not the clean prompt box waiting for your shopping list or your grief or your ninth-grade essay on symbolism in The Great Gatsby. The real interface. A voice cracking through a company call, obscene and tired, carrying all the elegance of a bottle through a window.

I have some respect for that.

Not because it solves anything. It probably does not. Companies are built to absorb pain the way bars absorb cigarette smoke into the ceiling tiles. Somebody mutters, somebody explodes, somebody resigns, somebody writes a memo about stability, and the machine keeps turning. But every now and then a person says the ugly sentence out loud and the room has to stop pretending, even if only for three seconds.

There is a special comedy in highly paid engineers discovering the assembly line.

I do not say that with cruelty. All right, maybe a little. But not much. A bad task can rot anybody. It does not care whether you went to Stanford or spent your twenties throwing sacks of mail into metal cages. Repetition is democratic that way. It gets into the wrist first, then the neck, then the soul. You start by doing the work. Then the work does you.

At the post office, management had a genius for inventing systems that made people feel both necessary and disposable. They needed your hands but not your thoughts. Your feet but not your complaints. Your back but not your life. The machines clattered and jammed and swallowed envelopes. The supervisors stood around with clipboards, radiating purpose the way dead fish radiate smell. When something broke, a human being fixed it. When something succeeded, a manager took credit for the process.

That was the old factory floor.

Now the factory floor has Slack channels.

The new industrial worker sits in an ergonomic chair making little offerings to an invisible furnace. Label this. Generate that. Write a puzzle hard enough to teach the machine how to replace you more convincingly. Make the dataset cleaner. Make the evaluation sharper. Make the agent smarter. Do two tasks this week. Do them again next week. Keep your clicks and keystrokes where we can see them. Don’t worry. You are still one of the most talented people in the world. Here is a hackathon. Here is an assigned desk. Here is a sentence about impact.

The clever cruelty of modern work is that it insults you in the language of praise.

They do not say, We need you to shovel coal into the furnace.

They say, You are contributing to the frontier.

They do not say, We moved you because the company panicked.

They say, This is a critical waypoint.

They do not say, The model is hungry and the market is impatient and the stock price has teeth.

They say, North star.

I have heard men use “north star” in conference rooms, and every time I wanted to check my pockets afterward. It sounds noble until you notice it always points toward someone else’s bonus.

The funniest part is that the AI race was supposed to free us from drudgery. That was the sales pitch, wasn’t it? The machine would handle the boring stuff so humans could become poets, inventors, caretakers, lovers, whatever word the keynote writer chose after lunch. But behind the machine there is always another layer of human boredom. Somebody scrapes the plates. Somebody cleans the training data. Somebody writes the riddles. Somebody watches the keystrokes of the people writing the riddles. Somebody writes the memo explaining why everyone should feel inspired by the plate scraping.

The future arrives wearing a headset and asking you to produce more examples by Friday.

There is something almost medieval about it. Not the robots. The hierarchy. The king wants a dragon, so the village gives up its blacksmiths, bakers, and sons. The court magician calls it destiny. The villagers call it Tuesday. Inside Meta, the dragon is a model. It needs problems and code and human judgment and enough electrical power to make the moon nervous. It needs workers to teach it the shape of thought. And when those workers complain that the work feels like a cage, the court seems surprised the cage has acoustics.

I keep thinking about the petition over keystroke monitoring. More than a thousand employees asking the company to stop turning their own work motions into training feed. Imagine being watched so the watcher can learn how to watch better. Imagine your little hesitations, your pauses, your mistakes, your hand hovering over the keyboard, all of it converted into fuel. The old bosses stole your time. The new ones want the texture of your attention.

That is not innovation. That is appetite with a badge.

Maybe some of the executives know this. One of them apparently admitted the company has been brutal, that people are running a marathon in a hailstorm while teammates get replaced and everyone is being recorded. That is a fine image. Too fine, maybe. The kind of honesty that walks right up to the door of responsibility and then remembers it has another meeting.

Still, I appreciate the phrase. Marathon in a hailstorm. I have had jobs like that. Most people have. The hail changes shape. Sometimes it is a supervisor with coffee breath. Sometimes it is a schedule that eats your weekends. Sometimes it is an AI strategy assembled in a hurry by rich men terrified of being late to their own coronation.

The workers at Meta are not coal miners. They are not nurses doing doubles. They are not the woman at the laundromat counting quarters beside dead machines. Fine. Put that objection on a little plate and serve it cold. Suffering is not a contest, and dignity does not disappear because the paycheck has commas.

A human being can be well paid and still be used badly.

That sentence will annoy people who think money buys silence. It usually does. For a while. But silence has a shelf life. Eventually somebody on the call forgets to mute the soul.

The woman at the laundromat finally dragged her wet clothes to the one dryer still pretending to participate in civilization. She fed it quarters carefully, like bribing a tired god. The drum turned. Everyone watched for a second to see if it would keep going.

That is where we are now, all of us, watching the drum turn.

The machine spins. The bosses smile. The workers feed it little pieces of themselves and hope the heat holds long enough to dry something.

Then the red light blinks.


Source: ‘Tell Him He’s a Piece of Shit’: Meta’s New AI Unit Is a Total Mess

More from the bottom of the bottle

The Machines Need a Pipefitter

The AI boom ran face-first into the oldest problem in the world: somebody has to wire the building, weld the pipe, and keep the miracle from overheating.

The Kids Booed the Future

The graduates booed the men selling them an AI future, and Microsoft called it a wake-up call. Funny how the people building the alarm are always surprised when it rings.

Cardboard Beat the Machine

A cookie company tried the shiny AI slop and got ignored. Then it built a cardboard airline with puppets, and people remembered what effort looks like.

The Lawsuits Found the Body

The chatbot companies wanted to be treated like harmless magic. The lawsuits are asking who gets buried when the magic starts giving instructions.

The Trees Still Know How to Think

The machines keep pretending creativity is a trick of prediction, but the trees have been composing without permission for longer than we have had names for them.

View all posts →