Faster Coding Makes Understanding Users the Bottleneck, So Mercor Wants More Product Managers
Mercor's product chief says coding is no longer the constraint, so his team fights to cut features and spends more time learning what users need.
20VC with Harry Stebbings · Mercor Head of Product on Revenue Concentration from Frontier LabsOswald Nitski, head of product at Mercor, an AI data and evaluation company, was asked whether faster building means smaller product teams. He argued the opposite, because engineering velocity is rising with better coding agents, while figuring out what to build has become the hard part.
AI, he said, makes product management harder, because his team is “trying not to build 10x more product surface area.”
By surface area, he means every feature, screen, and workflow a company has to maintain. The temptation to expand it is constant, Nitski said, because an engineer might think, “Oh, I can make all these features really quickly,” and then push changes that run to thousands of lines of code, which makes things incredibly chaotic.
So Nitski described his job as a battle to simplify. Instead of saying yes to everything the technology makes possible, his product team tries to trim what it offers and focus on the uses that can scale best, and to cut the rest.
That shift, in his view, changes who is the bottleneck. Coding is less of a constraint than it used to be, so the hard work is understanding what users actually do, what they need, and which products drive revenue. For that reason, he said Mercor now has a higher ratio of product managers to engineers.
He expects that trend to continue. As coding agents get better and engineering speed rises further, he predicts fewer engineers per product manager, because the limit will be understanding business needs and user needs. He added that a fast growing company still has to grow the two groups in lockstep, so one does not outrun the other.
“We’re constantly in this battle to try to simplify our product surface area,” he said.
Questions people ask
What does a product manager actually do?
A product manager decides what a team should build and why. The job usually involves talking to customers, studying how people use the product, weighing business results, and writing clear requirements so engineers build the most valuable thing next.
What is an AI coding agent?
A coding agent is AI software that writes, edits, and tests computer code from a plain-language request. Engineers still review the work, but the agents increase engineering velocity and let teams produce more code with less manual effort.
Why is a simpler product better?
Simpler products are usually easier to learn, less likely to break, and cheaper to maintain. Every extra feature adds design, testing, support, and future repair work, so companies that keep adding features can end up slower and more confusing to customers.