Someone Moved the Cheese

The other night a buddy of mine told me he's been out of work since the spring. Senior engineer, the guy you want when the thing breaks at 2am, twenty-some years of it. He's put in more applications than he can count and heard back on almost none. And he said something that stuck with me. He said he doesn't even know what they want anymore. He used to just be good at the job.

That's the whole thing, right there. He didn't get worse at the work. The work didn't change. What changed is how you get to do the work, and nobody sent us the memo.

Think about how you used to get hired. You knew somebody. You talked to the boss, or to a guy who'd worked with you and would vouch that you knew your stuff. A person looked at you and made a call. It wasn't perfect. Plenty of good people got left out of that too. But there was a human in the chair.

That human is mostly gone now. In its place is a stack of machines, and you have to clear every one of them before anybody who can actually picture you doing the job ever sees your name. Somebody moved the cheese. And the guys I came up with, heads down, building things, are exactly the ones who didn't notice it move, because we were busy doing the work instead of watching the maze.

Here's the stack he's up against.

First gate is the résumé bot. The applicant tracking system. It reads your résumé hunting for the exact words in the job post, and if you don't have them, you're gone before a human blinks. And this isn't a few people slipping through the cracks. Harvard Business School and Accenture asked employers straight, and 88 percent of them admitted their own system rejects qualified, high-skilled people just for missing the keywords. They've even got a name for who it buries: hidden workers. Twenty-seven million of them. Want to work, can't get seen.

The honest version, so I'm not unfair: the machine didn't dream that up on its own. A person set those filters. Some folks in the industry will tell you it's not the robot's fault, it's the humans behind it. Fine. But that doesn't help my buddy. The gate's still shut, and there's nobody back there to argue with.

Say you clear the bot. Now there's a test. A timed coding puzzle, the kind with the clock ticking and a trick you either remember or you don't. Here's what nobody running these will say out loud: the test doesn't measure whether you can do the job. The people at interviewing.io checked their own data and found no connection between how you score on those puzzles and how you actually perform. The real research on what predicts good work puts the timed-puzzle stuff near the bottom. So a guy with twenty years of shipping real software gets knocked out by a riddle he'd never see on the job, beaten by a kid who drilled the riddle for three weeks. That's not finding the best worker. That's finding the best test-taker. Hell of a way to run a railroad.

Clear that too, and you get the interview. Except more and more it isn't a person, it's a camera and an AI grading your face and your voice. The big one, HireVue, used to literally score your facial expressions until a federal complaint made them quietly stop. But the parts they kept still reward the same thing the whole system rewards. Talking. Confidence. A quick, smooth answer.

And that's the part that gets me, because the best engineers I know are the opposite of that. They're quiet. They think before they talk. Ask them something hard and they go still for a second, because they're actually working it out, and a camera reads that pause as "weak." The research is plain: interviews reward the confident fast-talker, they barely predict who's any good at the job, and the quiet ones who'd do it better get passed over. We built a hiring machine that filters for the salesman and against the builder, and then we act surprised when the work turns to junk.

Now stack one more thing on top, the one my buddy won't say but I will. He's not young anymore. And the machine has a thing about that.

This isn't me being dramatic. The EEOC, the federal government, settled its first AI-hiring discrimination case against a company whose software automatically rejected women over 55 and men over 60. Just deleted them. Screened out more than two hundred people for being old and paid three hundred sixty-five thousand dollars for it. There's a bigger lawsuit running right now that says another widely used system does the same thing to everyone over 40, only quieter. And it happens in small ways too. The bots favor the new buzzwords, so a guy who describes his work the way he learned to twenty years ago gets ranked under a kid who sprinkled in the trendy terms. AARP found more than a third of workers over 50 ran into age discrimination just in the job search. In tech they'll call you old at 35.

So put it together. The gate that used to be a person is now four machines in a row, and every one of them is tuned to reject the exact guy you'd want building your stuff. The experienced, quiet, heads-down worker who's better at the job than he'll ever be at performing for a camera. Nobody sat down and decided to do that to him. That's almost the worse part. It just happened, one efficient little tool at a time, while we had our heads down working.

I don't have a clean fix to hand him. I wish I did. I'm not even sure there's one a single guy can pull off against a system this size. The honest things I told him were small ones. It's not you. You didn't get worse. The maze changed and nobody rang a bell. And the things that make you good at the actual job, the patience, the quiet, the years, are the exact things the machine is built to throw out. Which means the rejection isn't a verdict on your work. It's a verdict on the machine.

He nodded. Didn't make him feel much better. Hard to, when you've sent two hundred résumés into a robot that was never going to read a one of them.

But he should know it's real, and it's not him. Somebody moved the cheese. Least we can do is say so out loud.

Sources, so you can check me: