The Two People Shaping the Future of OpenAI’s Research
As the AI landscape continues to evolve at a rapid pace, much of the spotlight at OpenAI has traditionally centered on its charismatic CEO, Sam Altman. Known for his showmanship and fundraising prowess, Altman has undeniably been the face of the company, even weathering a dramatic ouster and triumphant return that captured global attention. However, behind the scenes, the true engine driving OpenAI’s groundbreaking technological advances lies with its twin heads of research: Chief Research Officer Mark Chen and Chief Scientist Jakub Pachocki.
In an exclusive conversation during a recent visit to OpenAI’s London outpost—the firm’s first international office established in 2023—Chen and Pachocki offered insight into the research strategies guiding OpenAI’s future, particularly as the company prepares for the highly anticipated launch of GPT-5. ### Twin Peaks of Innovation
Mark Chen and Jakub Pachocki share and complement responsibility for steering OpenAI’s research division, ensuring the company remains ahead in the fiercely competitive AI race against giants like Google. Chen, who joined OpenAI in 2018 after a stint as a quantitative trader developing machine-learning models at Jane Street Capital, spearheaded landmark projects including DALL-E, the generative image model, and Codex, which powers GitHub Copilot.
Pachocki, joining OpenAI in 2017 from an academic career in theoretical computer science, took over as chief scientist in 2024. He is recognized as the architect behind OpenAI’s reasoning models, such as o1 and o3, which tackle complex tasks in coding, math, and science.
“We both wear multiple hats,” Chen explained. “My core responsibility is setting the research roadmap and long-term vision, but we also dive into specific technical challenges whenever they arise.” Pachocki added, “We’re both researchers at heart, pulling on whatever threads can help us advance the models.”
Breaking New Ground in Math and Coding
The pair arrived in London energized by recent milestone wins for OpenAI models. In July 2025, their large language model secured second place at the AtCoder World Tour Finals—one of the world’s most demanding programming competitions—and shortly thereafter, OpenAI announced its model had earned gold-medal-level scores at the International Math Olympiad (IMO), placing it among the world’s top 20 to 50 problem solvers.
While the math result attracted headlines, partly due to a coincidental tie with Google DeepMind’s similar achievement, Chen emphasized that the programming competition was an even bigger breakthrough: “Breaking into the very top echelon of human performance in a real-time, high-pressure coding contest—that’s unprecedented.”
Chen, who coaches the USA Computing Olympiad team, and Pachocki, both seasoned competitive coders, take particular pride in this validation of OpenAI’s reasoning models. “Competitive coding is intense and requires deep problem-solving skills. To see our AI perform at near the highest human level is a strong sign of where we’re heading,” Chen noted.
Balancing Research and Product
Since the launch of ChatGPT three years ago, OpenAI has transformed from a research lab into a fast-moving company, valued at $300 billion and competing directly with tech behemoths. While maintaining its research roots, the company now prioritizes shipping products to hundreds of millions worldwide. Release after release—ranging from GPT-4 series updates, new generative video and image models, to innovations like interacting with ChatGPT via voice—OpenAI has kept up relentless momentum.
The challenge of juggling open-ended research with urgent product development is one Chen and Pachocki are acutely aware of. According to Pachocki, “If we’re serious about artificial general intelligence (AGI), there will be countless paths and applications to explore. We keep shaking the tree and harvesting what we can.”
Chen stressed the role of releasing experimental models as part of research itself. “Deploying technology helps educate the public and fuels societal conversations about AI’s impact. It’s also invaluable for seeing how people use these tools in the real world,” he said.
Toward GPT-5 and Beyond
OpenAI’s forthcoming GPT-5 release has generated high expectations, with Sam Altman announcing it will arrive “soon.” While rumors of delays circulate, Chen and Pachocki are focused on ensuring the model meets rigorous standards. They view the evolution beyond GPT-4 and the success of reasoning-focused models as laying the groundwork for more capable, multi-purpose AI systems.
Both leaders emphasize that the next wave involves deepening AI’s abilities in coding, math, and logical reasoning—skills they consider central to building robust, general intelligence. “Coding and math are core to how we reason and solve problems,” Chen said. “Models that excel here push us closer to the broader vision of AGI.”
Legacy and Leadership
Beyond overseeing research advancements, Chen and Pachocki represent continuity and technical leadership at OpenAI amid organizational changes. They have notably taken over some responsibilities that were once part of teams led by cofounder and former chief scientist Ilya Sutskever, including the now-disbanded superalignment efforts aimed at preventing hypothetical future AI risks.
When asked about the future, Pachocki reflected, “We are entering an era where the capabilities of models can’t be fully captured by classical benchmarks anymore. It’s all about what they can do in the real world – coding, reasoning, interacting.”
Together, Mark Chen and Jakub Pachocki are quietly steering OpenAI’s research into the next phase of AI innovation—balancing ambitious visions for artificial general intelligence with practical applications that continue to capture the world’s imagination.
This article was based on an exclusive interview conducted by Will Douglas Heaven for MIT Technology Review, exploring the research leadership at OpenAI and the evolving landscape of AI development.
