The Past Got the Interview

The personnel file wishes to amend its statement.

In 2022, it says, I was already fluent in artificial intelligence.

I had not mentioned it at the time because I was busy doing whatever the hell it was I was paid to do. Moving numbers from one box to another. Fixing the printer. Writing the email nobody read. Smiling while a manager explained that a new software package would make everything more intuitive, which is a word people use when they have removed the labels from the doors.

But now it is 2026. The jobs want AI. The recruiters want AI. The little machine sorting the applications wants AI. So naturally I look back at the old years and discover that I was full of it.

Not lying, exactly. Let us not get too clean about this.

A working paper following 29.4 million LinkedIn profiles found that nearly one in five people had changed the title or description of a job they had already left. Since ChatGPT showed up, additions of AI words to old roles have risen more than sixfold. A snapshot taken now makes 2022 look about 30 percent more AI-skilled than it apparently was.

The researchers call it time travel.

That is a fine name. It makes it sound romantic. A man in a leather jacket gets into a sports car, floors it, and arrives four years earlier carrying a knowledge of large language models and a revised account of his accomplishments.

The actual scene is less glamorous. Somebody is alone at a kitchen table after midnight, clicking the little pencil beside a dead job. They have seen three listings that day requiring familiarity with tools that did not exist when they were working there. They know the recruiter will spend seven seconds looking. They know the filtering machine will spend less. They add a phrase because they did use something like it, maybe. They took an online course. They asked a chatbot to rewrite a paragraph. They watched the younger guy in marketing use it and thought, I could have done that.

Or they add the phrase because rent is due and the language of work has changed without asking whether anyone had time to catch up.

I do not have much appetite for sneering at that person.

A résumé has never been a police report. It is a sales brochure written by the product while the product is scared. Everybody knows this, including the hiring manager who has turned coordinated stakeholder alignment into three meetings and a pastry tray. The old lies were just more modest. “Managed” meant you stood near the problem. “Led” meant you were copied on the email. “Proficient” meant you once found the button.

Now the lie, or the hopeful translation, has a new word stamped across it: AI.

But here is where the little joke loses its cheap hat and becomes expensive.

People are not only editing their past for other people. They are editing it for the machines that decide which other people get to see them. And those machines, like every bad supervisor I ever had, may confuse a neat-looking record with the truth.

The career history gets scraped, counted, fed into labor-market studies, used to describe who has which skills and when they got them. It becomes training meat for systems that sort applicants. Then those systems may take the rewritten past as evidence of what the labor market looked like before the rewriting began.

A man changes an old line because he wants an interview today. Tomorrow some company’s hiring machine learns that people like him have always had that skill. The company decides the skill is common. The pay offer gets thinner. The job posting gets greedier. The next man has to claim even more just to get past the gate.

That is not time travel. That is a laundromat where everybody keeps putting dirty shirts into the same washer and then acting surprised the water has turned gray.

The workers did not invent the machine at the gate. They did not decide that every human life should be compressed into searchable nouns. They did not turn a career into a profile that must be constantly dusted, updated, optimized, and dressed for a party attended by software.

They are responding to the weather.

The people who built the weather, meanwhile, will give speeches about data quality. In Europe, AI systems used for hiring are now treated as high-risk under the AI Act. That sounds grave enough to require a man in a black robe. It means the training data is supposed to be relevant, representative, complete, and as free of errors as possible.

Good luck with that.

The raw material is a million anxious people repainting their own pasts to match the color of this week’s job market. The system wants a clean historical record, and it is collecting it from a billboard that changes when nobody is looking.

Maybe an old job really did involve the beginnings of what we now call AI. Language changes. A clerk using autocomplete, a designer using a recommendation system, a programmer building an old classifier—none of them were required to predict which stale acronym would become the admission ticket four years later. Calling all those later descriptions fraud would be stupid and cruel.

But calling them stable facts would be worse.

I remember the files at the post office. They had everything in them. Warnings. Attendance records. Training sheets. Complaints written in the brittle, cheerful language of a supervisor trying not to say somebody had made his life difficult. A file could make a man look like a saint or a disaster depending on which page you read first. That was with paper, dates, signatures, and somebody who could be asked what the hell happened.

Now we want an algorithm to infer a person’s value from a document they are continually rewriting under economic pressure.

Marvelous.

The hiring machine will not see the kitchen table. It will not see the bills beside the laptop, the mother asleep in the next room, the third rejection email, the small humiliation of realizing your old honest description has become invisible because it lacks the fashionable word. It will see “AI strategy” added to a job that ended in 2021 and nod its electric little head.

Then it will decide who looks prepared for the future.

Maybe the person who gets hired will be perfectly capable. Maybe they will be better than the machine’s tidy record says. People often are. The trouble is not that human beings revise themselves. We have been doing that since we learned to talk.

The trouble is that the machine takes the revision literally.

Somewhere tonight, another old job is being adjusted. A few words vanish. A few useful words arrive. The past sits up straighter, puts on a clean shirt, and waits by the phone.

It has an interview in the morning.

Source: Workers are rewriting their LinkedIn histories, and 2022 now looks 30% more AI-skilled than it was