Miserable Coffee. Leaves Me Alone.

Suppose I could have a little program that remembered where I liked to sit. I am sketching it here, not reporting an app I have built. Its entire user base would fit in my trousers, and would occasionally forget why it had entered the kitchen.

There would be a place for the name of a diner. A place for a note. Perhaps a button large enough to find without my glasses. I would press it and learn that I had been here before, paid for something, and formed an opinion which I had since mislaid.

Here is an imaginary entry:

Miserable coffee. Leaves me alone.

I would go back.

Ian Bogost has made a less imaginary program for remembering beans and coleslaw. In an October 12 Atlantic essay, read in a copy archived that evening, he describes using AI to build an iPhone app called Beanslaw. He and his family kept forgetting which barbecue places served the kinds of sides they liked. He had a programming background; the machine made a small, otherwise unappealing amount of work worth doing.

Then he showed it to a friend.

“Where are everyone else’s slaw and bean ratings?” the friend asked.

I can hear the reasonable tone. More information would improve the thing. A roomful of people knows more than one man. This is especially persuasive when the one man has forgotten why he came into the kitchen.

Bogost added community ratings in about an hour. ChatGPT pointed him toward Apple’s CloudKit and helped write the code. The difficult extra job had become an easy extra job. His private food journal could now serve a crowd. He never released it to the public.

It is the hour that bothers me. Not because it was too little labor. I don’t want a programmer made to suffer before I let him remember his lunch. It bothers me because too much trouble is a very useful answer, and the machine has made it harder to give.

I would have had that answer ready. Everyone else’s reviews? Sorry. Can’t do it. I would have sounded like a man with a technical limitation rather than a man unwilling to improve. Now the improvement can be finished before the argument about whether it belongs there.

Consider my lousy coffee.

A public rating would ask me to settle the matter. Good place? Bad place? How many stars? I could write a paragraph explaining the bargain, but I hadn’t wanted another writing assignment. I had wanted a table for the writing assignment I already had.

The coffee could be bad and the diner right. Nobody clearing away my cup while I was still working. Nobody hovering over the empty plate. A seat where I could spread the pages without borrowing space from the next fellow’s breakfast. I might sit there for an hour, getting one paragraph less wrong, and leave a perfectly honest note about the coffee.

Everyone else could hate the place for exactly the same reason. Slow service. Couldn’t get attention. Had to wave twice. I would believe them. They wanted breakfast brought promptly and the bill brought after it. They weren’t wrong because I wanted to sit there like a stain somebody had decided to leave until closing.

Their information could be accurate, plentiful, and increasingly useless to the question I had asked.

This isn’t an argument against reading other people’s reviews. I read them. If a dozen people say a place is closed, my private affection won’t get the door open. But will it suit me this afternoon is a different question from did most people approve. I don’t need to win the second argument before eating in answer to the first.

Now suppose my own note changes.

Same coffee. Wanted the bill. Nobody came.

There would be no scandal in my little program. I wouldn’t need to reconcile the entries, decide which was the real review, or establish that the diner had declined. The afternoon was different. I had somewhere to go. The thing I liked before was in my way.

Keep both notes. I can bear learning that I am not a consistent restaurant critic. I would rather the program remembered this than helped me arrive at a reliable-looking opinion I could defend in public.

A notebook can do it, of course. So can a scrap of paper. I know. People become wonderfully resourceful about old methods when you ask for a new one that isn’t going to interest everybody. But I might enjoy pressing the big button. I might enjoy changing it when it annoyed me. Bogost describes that pleasure too: AI made small projects and even unsuccessful experiments less costly, more worth trying.

I want some of that. Not a business. Not the satisfaction of announcing that I have escaped the businesses. A thing I can meddle with until it fits an undignified little part of my day.

There are still real jobs involved in making software work. Bogost’s experience isn’t a promise that I can mutter at a screen and receive a reliable program. But if the help makes this particular job possible, I don’t want to repay it by inventing a larger one.

In my sketch, I add one more note.

Table at the back. Left leg short. Fold a page under it.

At last. Something to do with the first draft.

Source: The Tragedy of AI Coding (published October 12, 2026; inspected copy archived October 12, 2026, 21:20:56 UTC).