Mercor’s Unique Interview Approach Defies AI Trends

There’s a certain rhythm to Silicon Valley hiring. It’s a beat many know by heart: submit your resume, maybe do a take-home assignment, then face the whiteboard. For years, that final step—standing before a blank slate to diagram systems and reason through algorithms—was the ultimate test of a software engineer’s mettle. It was a ritual, a rite of passage, and for many, a source of profound anxiety. Then, almost overnight, the rhythm changed. AI coding assistants like GitHub Copilot and ChatGPT arrived, promising to offload the grunt work, to be the ultimate pair programmer. The industry’s response, broadly, has been to adapt. If AI is the new calculator, why not let candidates use it? Interviews have evolved into “vibe coding” sessions and work trials, where the goal is to see how someone collaborates with the machine.

But one prominent player is stepping out of sync. Mercor, the $10 billion AI training startup, is doing the unexpected: in its hiring for full-time roles, it’s doubling down on the very thing others are phasing out. In a recent conversation on the ’20VC’ podcast, Mercor’s head of product, Osvald Nitski, laid out a philosophy that feels almost contrarian in 2025. The company has moved away from take-home assignments that are easily completed with AI. Instead, it wants to see raw, unaided cognition. “We’ll do one round where we find out if the person is familiar with AI tools,” Nitski said. But the core of their process remains staunchly analog: whiteboarding sessions focused on systems design, statistical understanding, and crucially, judgment.

“We care a lot about being able to set up good experiments and understanding statistics, having good judgment, and then systems design as well,” Nitski explained. He pointed to the central paradox of the AI era: these tools make it effortless to “offload” decision-making. The risk, as he sees it, is creating a generation of engineers who can prompt but cannot reason. “We want to make sure that people still have the ability to have good judgment and know what they’re doing and not just, like, regurgitate what comes out of Claude,” he stated. This isn’t about being a Luddite; Mercor itself uses AI to screen candidates on its talent marketplace platform. This is a deliberate, philosophical choice for their core team builders. They are betting that foundational problem-solving intelligence—the kind you can’t prompt-engineer—is the ultimate competitive advantage.

This stance creates a fascinating split screen with the rest of the tech world. On one side, companies are racing to integrate AI into every part of the hiring funnel. As reported by Business Insider, giants like Google, LinkedIn and Cisco have revised interview policies to explicitly allow and sometimes even encourage AI use. The logic is pragmatic: this is the tool of the job, so test with the tool. Startups built on AI, like the coding-focused Cognition, have redesigned their interviews around this reality. Emily Cohen, who leads people and operations at Cognition, offered a telling analogy: restricting AI in an interview is like “asking a kid to take a math test without a calculator.” For her, the interview should mirror the work: “For the bulk of building something similar to what you would do on the role, you can and should use AI tools.”

Meanwhile, other innovators are ditching traditional interviews altogether in favor of extended work trials or “auditions.” AI-powered startups like Lovable, Cursor and Kilo advocate for these projects as the most efficient way to gauge both technical skill and soft skills like communication and collaboration. It’s a trend that sidelines the whiteboard in favor of real-world, albeit simulated, output. Mercor’s approach rejects both trends. It says no to the AI-crutch interview and no to the purely output-based trial, opting instead for the pressured, cerebral exercise of the whiteboard session.

So, who’s right? The answer is likely that both approaches are probing different dimensions of a future-proof engineer. The “AI-allowed” camp is testing for a critical modern skill: the ability to leverage advanced tools to augment productivity and solve higher-order problems. It values efficiency and tool literacy. The work-trial model tests for practical execution and team fit in a realistic setting. Mercor’s low-AI strategy is a deep investment in what we might call “first principles thinking.” In an age of instant, AI-generated answers, the ability to deconstruct a problem from the ground up, to reason through trade-offs without an autocomplete, may become the rarest and most valuable skill of all.

It’s a high-stakes bet. By filtering for this specific type of rigorous, independent cognition, Mercor may be building a team with an unusually strong architectural and strategic backbone. They are selecting for the engineers who won’t just implement a solution suggested by an AI, but who can critically evaluate if it’s the right solution, foresee its downstream implications, and understand the statistical validity behind it. This is the kind of judgment that prevents catastrophic system failures and builds truly novel, resilient technology.

The true test of Mercor’s unique interview method won’t be in the hiring headlines, but in the products they ship. Can a team hired for its unaided judgment out-innovate teams hired for their flawless collaboration with AI? The industry will be watching. This isn’t just a debate about hiring tactics; it’s a live experiment on the very definition of expertise in the age of artificial intelligence. Mercor isn’t rejecting the future. It’s making a calculated gamble on which human qualities will matter most within it. In a world filling up with brilliant prompters, they are searching, deliberately and against the grain, for the irreducible thinkers.