Konstantin Schürholt
Konstantin Schürholt is a researcher at Ndea, the AI lab co-founded by François Chollet. He previously completed a PhD and postdoctoral research in computer science at the University of St. Gallen and was a research intern at Google DeepMind. He is best known for the August 2025 ARC Prize analysis of the Hierarchical Reasoning Model (HRM), in which his ablation studies demonstrated that the outer iterative refinement loop — not the hierarchical H-L architecture that received widespread media attention — was the primary driver of the model's ARC-AGI performance.
“This guy, Constantine at Francis Chalet's company, India, actually did. And it's this amazing breakdown that he posted on YouTube. The main takeaway is that the outer refinement loop is the main reason why these things work so well.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.