Jerry Tworek

Jerry Tworek is a Polish AI researcher, co-founder and chief executive officer of Core Automation, and the prior vice president of research at OpenAI where he led the development of the o1 and o3 reasoning models.
Nextomoro
6 min read
Jerry Tworek

Bio

Jerry Tworek is a Polish AI researcher, co-founder and chief executive officer of Core Automation, the San Francisco AI research lab he established in 2026 following his January 2026 departure from OpenAI after approximately seven years. At OpenAI he served as vice president of research and led the development of the o1 and o3 reasoning models, with prior contributions across Codex, early reinforcement-learning research applied to robotics, and the GPT-3 and GPT-4 series. As of May 2026, he leads Core Automation in San Francisco, with industry coverage reporting a target post-money valuation exceeding $5 billion and target round size between $500 million and $1 billion.

At a glance

Origins

Tworek attended the University of Warsaw for both undergraduate and graduate study in mathematics, completing the Bachelor of Science in applied mathematics from 2008 to 2012 and the Master of Science in mathematics. The mathematical-foundations training informed his subsequent work on reinforcement learning and the reasoning-model program at OpenAI.

He spent the first five years of his career in quantitative finance, including roles at JP Morgan Chase and as a senior quantitative researcher in the hedge-fund industry. The transition from quantitative finance into AI research is characteristic of the late-2010s migration of mathematical-and-statistical talent into machine-learning research roles, with the OpenAI hire in 2019 marking the inflection point.

Career

Tworek's professional career organizes around a quantitative-finance period followed by approximately seven years at OpenAI and the founding of Core Automation in 2026.

He held quantitative-finance roles in the hedge-fund industry before joining OpenAI in 2019, contributing to scaling research and the technical foundations of large language models during the early GPT-3 and pre-GPT-4 period. He was promoted to vice president of research at OpenAI in 2022, with responsibility expanding to cover ChatGPT and the GPT-4 series.

Tworek's most-cited technical contributions span four research lines at OpenAI:

  • Codex (2021 to 2022). Principal contributor to the OpenAI code-generation model line that anchored GitHub Copilot, with subsequent extensions across the OpenAI coding-model series.
  • Reinforcement learning applied to robotics (early OpenAI period). Led research on large-scale reinforcement-learning approaches for robotic-manipulation tasks at a time when the methodology had not yet become widespread in commercial AI research.
  • GPT-3 and GPT-4 series (2020 to 2024). Contributed to scaling research and the technical foundations of the OpenAI flagship language-model series.
  • Reasoning models o1 and o3 (2024 to 2025). Led the development of the OpenAI reasoning-model program, the family of models that achieved PhD-level performance in mathematics, physics, and coding through inference-time scaling and a new training paradigm built around reinforcement learning over chain-of-thought traces. The o1 and o3 lines have been characterized in industry coverage as one of the principal capability inflection points for frontier AI in 2024 and 2025.

Tworek announced his departure from OpenAI on January 5, 2026. He cited a desire to pursue research directions that were difficult to pursue within OpenAI's current operating environment, with subsequent reporting framing the departure around limitations on speculative-research scope at the company.

He founded Core Automation in 2026 in San Francisco. The company has been characterized in industry coverage as positioning itself to build "the world's most automated AI lab," with explicit focus on automating frontier-AI research workflows and on developing learning algorithms designed to replace large-scale pretraining. The reported flagship-model concept is named "Ceres," targeting a 100-fold reduction in training-data volume relative to current state-of-the-art models through continual learning in production environments. The April 2026 industry coverage characterized Core Automation's research team as having been "nerdsniped" from OpenAI, Anthropic, and Google DeepMind, with senior research-and-engineering hires across the three frontier labs joining the founding team. Industry coverage reported a target post-money valuation exceeding $5 billion and target round size between $500 million and $1 billion, with specific lead investors and round-close dates not yet publicly confirmed at the time of writing.

Affiliations

  • JP Morgan Chase: Quantitative analyst, late 2010s.
  • Hedge-fund industry: Senior quantitative researcher, late 2010s.
  • OpenAI: Research staff 2019 to 2022; Vice President of Research 2022 to January 2026.
  • Core Automation: Co-founder and Chief Executive Officer, 2026 to present.

Notable contributions

Tworek's published record runs through approximately seven years at OpenAI, with the o1 and o3 reasoning-model lines as the most consequential technical output and the broader contribution set spanning code generation, reinforcement-learning robotics, and the GPT scaling program.

  • OpenAI o1 and o3 reasoning models (2024 to 2025). Led the development of the OpenAI reasoning-model program; PhD-level performance on mathematics, physics, and coding benchmarks through inference-time scaling and reinforcement-learning over chain-of-thought traces.
  • Codex (2021 to 2022). Principal contributor to the OpenAI code-generation model line that anchored GitHub Copilot.
  • Reinforcement-learning applied to robotics (early OpenAI period). Led research at a stage when large-scale RL for robotic manipulation had not yet become widespread in commercial AI research.
  • GPT-3 and GPT-4 series (2020 to 2024). Contributor to scaling research and technical foundations.
  • Core Automation (2026). Founded the AI research lab; reported target post-money valuation exceeding $5 billion; flagship "Ceres" model concept oriented around continual learning.

Investments and boards

  • Core Automation (AI): Co-founder and Chief Executive Officer, 2026 to present. Privately held; reportedly seeking $500 million to $1 billion at a target post-money valuation exceeding $5 billion as of April 2026.

No public personal angel-investor activity is on record outside the founder-and-operator role at Core Automation as of May 2026.

Network

Tworek's longest-running professional relationships span the OpenAI research-and-engineering team, particularly the senior technical leadership across the GPT, Codex, and reasoning-model programs. The Core Automation founding team draws senior research-and-engineering hires from OpenAI, Anthropic, and Google DeepMind, with industry coverage characterizing the team as one of the most concentrated frontier-AI research groups outside the established frontier labs. The senior recruitment pattern overlaps with the broader 2025 to 2026 trend of senior frontier-lab departures attracting frontier-lab-scale founding capitalization.

Position in the field

As of May 2026, Tworek is one of the principal senior-departure founders to leave OpenAI's research leadership and form a competing frontier-research lab. The closest peer comparators are Mira Murati at Thinking Machines Lab and Ilya Sutskever at Safe Superintelligence, each having departed OpenAI senior leadership roles to found research-first Insurgent labs. Tworek differs in the explicit operating-research framing (Core Automation has been characterized as targeting "the world's most automated AI lab" with explicit research-workflow automation) and the architectural commitment to continual learning and post-transformer architectures.

The o1 and o3 reasoning-model legacy from OpenAI gives Tworek a credentialed track record at a stage in frontier-AI research when reasoning-model capability is the most-watched capability axis. Industry coverage has characterized Core Automation's founding-period capital formation as one of the principal next-wave frontier-research startups, with Tworek's prior architecture-of-reasoning role as the central credibility signal.

Outlook

Open questions over the next 6 to 18 months:

  • Fundraising progression. Confirmed lead investors and round-close at or above the target post-money valuation exceeding $5 billion.
  • Public-product or research-disclosure milestones. Beyond the founding-period coverage.
  • Continual-learning research. Translation of the architectural thesis into reproducible capability benchmarks.
  • Competitive dynamics. Direct competition with OpenAI, Anthropic, and Google DeepMind, and with peer Insurgent labs.
  • Senior-talent recruitment. Continued hiring from frontier labs and academia leveraging the founding-team credentials.
  • Multi-year pre-product timeline. Tension with the rapidly evolving frontier-AI capability environment.

Sources

Person Core Automation
About the author
Nextomoro

Nextomoro

nextomoro tracks progress for AI research labs, models, and what's next.

Read next

Zamba2

Yoshua Bengio

SGLang

Tudor Achim

Rnj-1 (Ramanujan)

AI Research Lab Intelligence

nextomoro tracks progress for AI research labs, models, and what's next.

AI Research Lab Intelligence

Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to AI Research Lab Intelligence.

Your link has expired.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.