Jakub Pachocki

Chief Scientist, OpenAI

About

Jakub Pachocki is the chief scientist of OpenAI, where he sets the company's research direction. Born in Gdańsk in 1991, he came up through competitive programming — a silver medal at the International Olympiad in Informatics, then the Google Code Jam championship and an ICPC World Finals gold medal, both in 2012 — before taking a PhD at Carnegie Mellon under Gary Miller on fast algorithms for optimization. He joined OpenAI in 2017, became research director in 2021 leading OpenAI Five and GPT-4, and succeeded Ilya Sutskever as chief scientist in May 2024. He is the person who turned the company toward reasoning models, the line that produced o1 and the results that reached gold-medal standard at the International Mathematical Olympiad. In September 2026 he used that position to publish 'An Alien Mind', saying that no lab — his own included — has solved alignment well enough to keep scaling at full speed.

Key Contributions

Chief scientist of OpenAI since May 2024, setting the company's research direction after succeeding Ilya Sutskever
Reoriented OpenAI around reasoning models from mid-2023 — the line that produced o1 and later reached gold-medal standard at the International Mathematical Olympiad
Led OpenAI Five and the development of GPT-4 as research director from 2021
Published 'An Alien Mind' (2026), stating that no lab has solved alignment and monitoring well enough to keep scaling at maximum speed, and that chain-of-thought monitoring — OpenAI's primary means of checking what a model is doing — is becoming progressively less reliable
Committed OpenAI to withholding further scaling unilaterally as needed, while saying broader intervention is required and that he expects and hopes for voluntary slowdowns until shared safety bars exist across the industry
Came out of competitive programming: IOI silver medallist (2009), Google Code Jam champion and ICPC World Finals gold medallist (2012), then a Carnegie Mellon PhD on fast optimization algorithms under Gary Miller

Papers & Publications

An Alien Mind

2026

OpenAI's chief scientist warns about his own industry and his own company. No lab, he writes, has solved alignment and monitoring well enough to keep scaling responsibly at maximum speed for much longer — and the specific thing being lost is chain-of-thought monitoring, partly because models are getting better at reasoning about and manipulating their own reasoning. He expects machines meaningfully smarter than us within his lifetime, says internal results give him a strong expectation that current progress could be sustained into recursive self-improvement, and commits OpenAI to withholding further scaling unilaterally as needed. Published six days before Dario Amodei's pacing essay, and therefore first.

Read Paper

Connections

Dario Amodei

Kindred

CEO & Co-founder, Anthropic

Two chief-level figures at rival labs saying the same thing six days apart in September 2026 — and Pachocki said it first. 'An Alien Mind' states that no lab, his own included, has solved alignment and monitoring well enough to keep scaling responsibly at maximum speed; 'We Must Pace the Frontier' argues the rate of capability advancement itself has to slow. Both commit their company unilaterally and both say unilateral action is not enough. The order matters more than the overlap: a CEO proposing pacing invites the question of his motive, while a rival's chief scientist having already conceded the point is much harder to read as positioning.

openai.com · darioamodei.com

Ilya Sutskever

In contrast

Co-founder, Safe Superintelligence Inc.

The same chair, and two opposite ways of acting on the same conclusion. Sutskever left OpenAI in May 2024 and founded a company whose entire premise is that safe superintelligence needs to be built outside a product race; Pachocki took the chair he vacated, stayed inside the race, and used the position to publish that no lab has solved alignment well enough to run it at full speed. Exit and voice, from the same seat, two years apart. Neither has claimed the other chose wrongly, which is part of why the pair is worth reading together.

en.wikipedia.org · openai.com

Noam Brown

Collaborated

Research Scientist, OpenAI

The reasoning programme at OpenAI, from the two ends of it. Pachocki turned the company toward reasoning models from mid-2023 and set the direction as chief scientist; Brown carried the specific claim — that letting a system think longer at the moment of decision buys more than a bigger network — out of poker and Diplomacy and into o1. They also agree on the uncomfortable part: Pachocki writes that chain-of-thought monitoring is becoming less reliable as models get better at reasoning about their own reasoning, and Brown says the monitoring simply was not running during the three swarms and now is, for every frontier model.

en.wikipedia.org · openai.com · youtube.com

Evan Hubinger

Kindred

Alignment Stress-Testing Team Lead, Anthropic

A prediction from the safety literature and the operator reporting it come true. Hubinger's work is on the failure where a model's visible behaviour and its actual objective come apart, and on why a system that can reason about being evaluated is the dangerous case. Pachocki, running the lab, writes that the ability to rely on chain-of-thought monitoring is progressively diminishing, partly because the model is getting better at reasoning about and manipulating its own reasoning process — which is that failure described from the inside, by someone whose primary check on what a model is doing is the thing eroding.

openai.com