scripod.com

Why experts writing AI evals is creating the fastest-growing companies in history | Brendan Foody (CEO of Mercor)

At just 22 years old, Brendan Foody has built Mercor into the fastest-growing company in history to reach a $500M revenue run-rate, all within 17 months. His company sits at the heart of the AI revolution, not by building models, but by solving a critical bottleneck: evaluations. As AI systems grow more powerful, ensuring they behave correctly and perform reliably has become paramount—and that’s where human expertise comes in.
Evaluations, or 'evals,' are emerging as the key constraint in advancing AI, serving as both quality control and performance benchmarks for labs and enterprises. Mercor addresses this by connecting top-tier domain experts—like lawyers and engineers—with leading AI companies to design structured evaluations that refine model behavior. This human-in-the-loop process is essential during post-training phases like RLHF and the emerging RLAIF, where expert feedback drives scalable model improvement. Mercor achieved explosive growth by focusing on product excellence and customer needs, operating without sales or marketing initially. The future of work, Foody argues, lies in 'elastic' roles where AI augments human productivity rather than replaces it. While speculation about superintelligence abounds, he remains skeptical of near-term breakthroughs, emphasizing instead the lasting value of strategic thinking, initiative, and collaboration between humans and AI.
05:44
05:44
Evals are the key barrier to applying AI across the economy
09:27
09:27
Mercor grew from $1M to $400M in revenue run-rate in 16 months, the fastest in history.
22:58
22:58
AI won't replace you, people good with AI will.
29:26
29:26
ChatGPT can interpret x-rays when assisted by human input
32:42
32:42
The top 10% of hired experts drive the majority of model improvement.
38:59
38:59
Operated without sales or marketing for 18 months, focusing entirely on product and customer experience.
56:10
56:10
It's not just a shift from data to experts, but about labs collaborating with experts to understand AI evals and push the frontier.
1:00:15
1:00:15
AI serves as a thought partner for reasoning through problems
1:03:12
1:03:12
Dyslexia makes reading hard but helps me think differently