Jakub Pachocki

Chief Scientist — OpenAI

Profile

Jakub Pachocki is the Chief Scientist at OpenAI, succeeding Ilya Sutskever in May 2024. If you use ChatGPT, GPT-4, or any of the o-series reasoning models, you are using software whose research direction Pachocki sets. That makes him one of the most consequential technical leaders in the field, even if he is nowhere near as famous as the CEO who announced his promotion.

His path is unusual. Pachocki grew up in Poland, studied computer science at the University of Warsaw, and earned a PhD in theoretical computer science at Carnegie Mellon — the kind of background that produces graph algorithm specialists, not frontier AI leaders. He was a world-class competitive programmer (ICPC gold team, 2012) before a brief postdoc at Harvard. He joined OpenAI in 2017 and never left. That theory-to-systems trajectory shows up in the work: big engineering bets grounded in clear algorithmic thinking.

Inside OpenAI, Pachocki led OpenAI Five — the Dota 2 agents that beat the world champions in 2019 using large-scale reinforcement learning — and then led the post-training work on GPT-4. When OpenAI pivoted hard into reasoning with the o1 model in 2024, he was the central technical driver. Sam Altman has publicly called him “easily one of the greatest minds of our generation.” That is the boss talking, but the track record backs it up: GPT-4, OpenAI Five, and the reasoning paradigm are three of the most important bets OpenAI has ever made, and Pachocki’s fingerprints are on all three.

For developers learning AI, Pachocki is worth paying attention to because he is unusually clear about where the research is heading. He talks openly about models reaching “research intern” level capability, about chain-of-thought as a monitorable reasoning trace, and about AI systems becoming persistent entities that do real science. If you want to understand what OpenAI is actually building — not what the marketing says — read and listen to him, not the keynote speakers.