An Alien Mind
Plus: Tesla De-Weaponized Rare Earths, a16z Called the Underclass a Fantasy
Read the storyChief Scientist, OpenAI
Jakub Pachocki is OpenAI’s chief scientist. He joined the organization’s research work in 2017 and previously served as director of research, leading projects including GPT-4 and OpenAI Five. His published work spans reinforcement learning, multi-agent competition and robotic manipulation. He holds a theoretical computer-science PhD from Carnegie Mellon. His 2026 writing discusses reasoning models, alignment, monitoring and the pace of increasingly capable AI systems.
Get AI Secret’s free daily briefing, including Jakub Pachocki and the wider AI industry.

Jakub Pachocki is OpenAI’s chief scientist. He joined the organization’s research work in 2017 and previously served as director of research, leading projects including GPT-4 and OpenAI Five. His published work spans reinforcement learning, multi-agent competition and robotic manipulation. He holds a theoretical computer-science PhD from Carnegie Mellon. His 2026 writing discusses reasoning models, alignment, monitoring and the pace of increasingly capable AI systems.
Sources checked . Selected career background and milestones.
Selected research, leadership and career milestones, newest first.
Writing as OpenAI’s chief scientist, Pachocki discussed reasoning-model progress and concerns around alignment, monitoring and recursive improvement. His forecasts are presented as his expectations, not established outcomes.
Source: Jakub Pachocki: An Alien Mind ↗OpenAI announced Pachocki as its new chief scientist following Ilya Sutskever’s departure, citing his leadership of GPT-4, OpenAI Five and research on large-scale learning.
Source: OpenAI: Jakub Pachocki appointed Chief Scientist ↗Pachocki co-authored the technical account of OpenAI Five, describing distributed reinforcement learning and the system’s April 2019 defeat of Dota 2 world champions Team OG.
Source: OpenAI: Dota 2 with large scale deep reinforcement learning ↗Pachocki co-authored OpenAI’s work on dexterous in-hand object manipulation, using reinforcement learning and simulation randomization to transfer learned control to a physical robotic hand.
Source: OpenAI et al.: Learning Dexterous In-Hand Manipulation ↗Pachocki co-authored research showing that competitive self-play in simulated environments can generate complex behaviors and a natural learning curriculum without hand-designing every skill.
Source: Bansal et al.: Emergent Complexity via Multi-Agent Competition ↗OpenAI’s appointment biography dates Pachocki’s research initiatives at the organization to 2017, before his later work as director of research and then chief scientist.
Source: OpenAI: Jakub Pachocki appointed Chief Scientist ↗Briefs, articles and relevant sections, newest first. Explore older coverage below.
More coverage
Plus: Tesla De-Weaponized Rare Earths, a16z Called the Underclass a Fantasy
Read the story