Mercor: The Hidden AI Hiring Platform for Engineers

A woman named Katya is unemployed, scrolling job listings, and she clicks through to a page that tells her to do a video interview with an AI named Melvin. Her words to The Verge: "It just seemed like the sketchiest thing in the world." She closed the tab.

Melvin. That's the name they went with.

A few weeks later, still no job, she went back. And the company behind Melvin, Mercor, is now valued at ten billion dollars, and its three founders are reportedly the youngest self-made billionaires in the world.

Okay, so what are they actually selling? Because "AI interviews candidates" sounds like a dime-a-dozen recruiting startup.

It was. That's the thing. Founded in twenty twenty-three by three nineteen-year-olds from the Bay Area... Brendan Foody, Adarsh Hiremath, Surya Midha. All Thiel Fellows, all school dropouts. The product matched overseas engineers with tech companies by having an AI read your resume and your interview transcript.

And by twenty twenty-four that's worth what?

Two hundred and fifty million. Forbes reported in September twenty twenty-four they'd interviewed over a hundred thousand people. Jack Dorsey invested. Peter Thiel invested. Benchmark led it, and two OpenAI board members, Adam D'Angelo and Larry Summers, put money in personally.

Hold on. OpenAI board members funding a recruiting startup. That's a tell.

It is. Because what kept happening was AI labs calling Mercor and saying: we don't want engineers for our office. We want doctors. Lawyers. Accountants. We want to pay them to teach our model. And in February twenty twenty-five, Mercor pivots. By October twenty twenty-five it closes a three hundred and fifty million dollar Series C led by Felicis, at ten billion. Five times its valuation eight months earlier.

Five x in eight months. What changed in the market that fast?

Two things. One, Meta took a stake in Scale AI, and cofounder Adarsh Hiremath told CNBC that move "torpedoed" a competitor overnight, because every lab that used Scale suddenly didn't want Meta near their data pipeline.

And two?

Two is the interesting one. Labs ran out of internet. You can't scrape your way to a model that does investment banking. So they need the humans who did the banking.

Why not just buy the data from Goldman Sachs?

Foody answered that onstage at TechCrunch Disrupt in October twenty twenty-five, almost too honestly. He said, quote, "There is an argument that Goldman Sachs doesn't love the idea of having models that are able to automate their value chain." So, he said, the labs "need to hire contractors who previously worked at those companies, understand those workflows, and are willing to train models to automate them."

Wait. So the business model is: the institution won't help you replace it, so you go hire the people it already let go.

Or people who are simply willing. And they pay. As of that Series C, Mercor said it manages over thirty thousand contractors, collectively paid more than one and a half million dollars every single day.

Per day?

Per day. And by the time 60 Minutes went inside the company this month, October fifth, twenty twenty-six, that network was described as more than a hundred thousand freelancers. A winemaker. A music producer. A physician training AI to read lab results. Which brings us to the part your title's really asking about. Mid-career engineers leaving big tech for this.

Do we actually have one? A real person, not a press quote.

We do. A software engineer named Ersin Akinci wrote it up publicly in August twenty twenty-five. Eight months as an AI trainer and reviewer with Mercor. His words: it had been almost a year since he'd written code for money, yet he was paying his bills reviewing code.

That's a bleak little sentence.

He didn't frame it that way. He said it paid down his mortgage and his startup debt, it's a hundred percent remote, the demand's been steady, and the pay is... his word... phenomenal. He described the job as part project designer, part technical writer, part, quote, "paid AI philosopher."

But he's a contractor. No equity, no benefits, project ends whenever the lab says it ends.

Correct, and he says so himself. The Verge put it harder in March this year: these workers are "competing for precarious gigs pantomiming the careers they'd hoped to have."

Okay. And what about the engineers inside Mercor? The actual staff?

Ten days ago. October fifteenth, twenty twenty-six. The Inference reported, citing Gazetteer SF, that Mercor had terminated at least fifty full-time employees the prior week. Engineering, product, cybersecurity. US and India. The notice was a two-sentence email with no reason given. Many were escorted out by security.

Two sentences.

One of them was considered a top performer by colleagues. Another had just gotten a raise.

And the company says what?

Foody posted on X: "It wasn't a layoff and had nothing to do with costs. We're profitable, have one billion plus in cash... The word layoff is inaccurate." A spokesperson said less than five percent of the company left, and they're still hiring. Insiders put headcount near six hundred and fifty, which Mercor wouldn't confirm. On those numbers, fifty is closer to eight percent.

And the hiring claim?

Verifiably true. Over a hundred open roles on the careers page that same day.

There's a detail in the timing that's bugging me.

You caught it. October thirteenth, the day before the cuts surfaced, Mercor and 1Password announced an AI agent that automates site reliability work. Diagnosing outages. Built on Codex from more than three hundred historical incidents. 1Password's CTO Nancy Wang reported ninety-five percent accuracy and fifty-seven percent faster median time to remediate.

So the week they cut engineers, they shipped the thing that does engineering.

Mercor says the two aren't connected, and nobody has shown that they are. But the loop closes anyway. The company that pays experts to train their replacements... employs experts.

So what do you tell a thirty-five-year-old at a FAANG company looking at this?

That the money is real, the work is genuinely interesting, and the job description is honest if you read it closely. Mercor's own slogan is "organizing human intelligence." Nobody said the organizing was for you.