Are it actually six Polish geniuses who are in charge of the fate of OpenAI? Jakub Pachocki, chief scientist of OpenAI, succeeded Ilya and achieved a breakthrough that the latter had been dreaming of for many years. Altman even posted a short article mentioning two indispensable figures in OpenAI: Jakub Pachocki and Szymon Sidor. Altman believes that the two of them are a perfect match. "OpenAI has no problems that the two of them cannot solve." Without them, there would be no OpenAI today.
What you may not know is that the two have known each other since they were in high school in Poland, and netizen Teortaxes posted a photo of the OpenAI team, and the concentration of Polish people was overwhelming─
The six top contributors in the OpenAI team are all Polish, from left to right, Jakub Pachocki, Łukasz Kaiser, Łukasz Kondraciuk, Szymon Sidor, Wojciech Zaremba, and Jerry Tworek.
Among them, Wojciech Zaremba is one of the eleven co-founders of OpenAI.
When OpenAI was founded, Polish engineers accounted for a large proportion of the team, including the above-mentioned Jakub Pachocki, Szymon Sidor, Łukasz Kondraciuk and others.
They became the core strength of the original OpenAI team, and the term "Polish Mafia" even appeared within OpenAI.
They have a common channel on Slack, where they occasionally post some AI-related Polish news.
The underrated OpenAI “Polish Legion”
In the early and core research of OpenAI, many researchers or engineers with Polish background made outstanding contributions.
Among them are a co-founder of OpenAI, a former vice president of research, and a co-author of the famous paper "Attention is All You Need."
Jakub Pachocki
Jakub Pachocki
Since joining OpenAI in 2017, Jakub has held important positions such as research director and is a leader in core projects such as GPT-4.
In the early days of OpenAI, he was committed to the research of large-scale reinforcement learning and complex game systems. Through projects such as OpenAI Five (Dota 2), he verified the key understanding of "large-scale training triggers a leap in capabilities" from a practical level.
Since then, as the core person in charge, he has led the research and development of GPT-4, built an "engineering-research" closed-loop system integrating training, optimization and scalability, and made breakthrough progress in using reinforcement learning to improve the model's complex reasoning and coding capabilities.
His core contribution lies in establishing a large-scale methodology combining large models with reinforcement learning (RL), transforming cutting-edge research into a reusable training system, and successfully leading a team to deliver key models with generational significance.
In May 2024, Jakub succeeded Ilya Sutskever as the chief scientist of OpenAI.
Łukasz Kaiser
A Polish computer scientist and machine learning researcher, he has been engaged in basic research on deep learning for a long time.
He received a PhD in computer science from RWTH Aachen University in Germany and worked as a researcher at the French National Center for Scientific Research (CNRS).
Łukasz Kaiser was a research scientist at Google Brain and made great contributions to the invention of the attention mechanism and Transformer architecture. He was one of the co-authors of the paper "Attention is All You Need". This architecture has become the core foundation of modern large language models.
Łukasz Kaiser
Łukasz Kaiser joined OpenAI in 2021 and is engaged in research related to large models and inference; public information shows that he has participated in the research and development of ChatGPT and GPT-4 multi-modal directions, and worked with the team to promote inference models such as o1.
Łukasz Kondraciuk
Łukasz Kondraciuk
Polish engineer/researcher listed as one of the contributors to the OpenAI o1 inference model.
He is one of the early team members and has practical contributions in ChatGPT and AI development.
Szymon Sidor
Szymon Sidor currently serves as a technical researcher at OpenAI. He holds a bachelor's degree in computer science from the University of Cambridge and a master's degree in mechatronics, robotics and automation engineering from the Massachusetts Institute of Technology.
Szymon Sidor
He has joined OpenAI since around 2016 and is one of the early researchers and core technology drivers of OpenAI.
He played an important role in building the GPT-4 model, and was a key figure in introducing reinforcement learning (RL) into the large language model together with Ilya and Łukasz Kaiser, which directly gave birth to the subsequent o1 inference model.
Altman praised him as "tireless," noting his role in solving seemingly impossible problems.
Wojciech Zaremba
A Polish computer scientist, he is one of the co-founders of OpenAI.
Wojciech Zaremba
Zaremba holds master's degrees in mathematics from the University of Warsaw and Ecole Polytechnique in Paris, and a PhD in computer science from New York University, where he studied under Yann LeCun and Rob Fergus.
Before joining OpenAI, he interned at Google Brain and Facebook AI Research, where he conducted research on neural networks, adversarial examples, and distributed training.
In 2015, Zaremba co-founded OpenAI with Altman, Musk, Ilya Sutskever and others.
He initially led the robot team. After 2020, he turned to leading the GPT series model, Codex and code-related teams. He is one of the key figures in the company's strategy and technology development.
Jerry Tworek
Also known as Jarosław Tworek, former Vice President of Research at OpenAI.
Jarosław Tworek
He received a master's degree in applied mathematics from the University of Warsaw. He joined OpenAI in 2019 and led early work such as o1, o3, Codex, and GPT-4. He is hailed by the industry as a key figure in the "development of large language model reasoning capabilities."
According to media reports, in January 2026, he left OpenAI and founded Core Automation to develop new AI models that require less data and calculations.
He said that he left to pursue core research that was difficult to carry out at OpenAI, and believed that AI has entered the "research era" and requires new breakthroughs.
Ilya and Tworek leave
Can the "Polish Legion" create another miracle?
In 2023, Microsoft CTO Kevin Scott's email revealed that OpenAI chief scientist Ilya was dissatisfied with the research breakthroughs and promotions of his disciple Jakub Pachocki, which led to a conflict with CEO Altman, triggering the board of directors' dismissal of Altman.
Ilya's departure is the inevitable result of OpenAI gradually shifting its computing resources to ChatGPT-related application departments during its strategic transformation.
In the process, Ilya, who was interested in pure research, lost to his apprentice Jakub, who was good at producing practical results.
Scott wrote in the email:
Jakub is more able to promote research breakthroughs than Ilya, so Ultraman promoted Jakub to be responsible for the main model direction. Jakub's work has accelerated since then, and he has made amazing progress in recent weeks.
Ilya found it difficult to accept that his mentor status was overturned. The success of the application department (such as ChatGPT) led to a shortage of GPUs and manpower in the research department. Under the blow of these two factors, a frustrated Ilya left OpenAI, which he helped to create.
What is regrettable is that after Ilya left, in January this year, one of the six Polish Legion members, OpenAI Vice President of Research Jerry Tworek, also announced his resignation.
Jerry, who had been at OpenAI for seven full years, finally had a conflict with Jakub Pachocki because he repeatedly applied to the management for more computing power and personnel support to no avail:
Pachocki is more optimistic about the existing large language model architecture that can achieve quick results. What the company needs is a product that can be implemented immediately, rather than a theory that will be implemented in an unknown year and month.
The departure of a series of core scientists and researchers such as Tworek and Ilya is due to the fact that application departments such as ChatGPT are "eating up" key computing resources. The reality that basic and cutting-edge research is being marginalized has also been interpreted by the outside world as a regression of OpenAI's idealism.
Some netizens even think thatAs a result, OpenAI will lose its former leading position in basic research, leaving only its brand value.
As AI has entered the "research era", for the world's top AI laboratories such as OpenAI, Anthropic, and Google DeepMind, basic and cutting-edge research is the root of development.
Without the leadership in research, OpenAI will be far away from replicating the next "ChatGPT" moment. No matter how powerful the "Polish Legion" is, it may not be able to do anything.