Core Automation

Core Automation is an American artificial intelligence research lab founded in 2026 by Jerry Tworek, the former vice president of research at OpenAI where he led the development of the lab's reasoning-model program through 2025. The company has been characterized in industry coverage as positioning itself to build "the world's most automated AI lab," with an explicit focus on automating frontier-AI research workflows and on developing learning algorithms designed to replace large-scale pretraining. As of May 2026, Core Automation is reportedly seeking between $500 million and $1 billion in fresh funding at a target post-money valuation exceeding $5 billion, just weeks after its formal launch, with a senior research team that has been characterized as "nerdsniped" from OpenAI, Anthropic, and Google DeepMind.

At a glance

• Founded: 2026 in San Francisco. Public coverage emerged in April 2026 following Tworek's January 2026 departure from OpenAI.
• Status: Private. Reportedly fundraising at a target post-money valuation exceeding $5 billion as of April 2026.
• Funding: Reportedly seeking $500 million to $1 billion in fresh capital. Specific lead investors and round-close dates have not been publicly confirmed at the time of writing.
• CEO: Jerry Tworek, Co-Founder and Chief Executive Officer. Former vice president of research at OpenAI; led the development of the o1 and o3 reasoning-model lines.
• Other notable leadership: Senior researchers recruited 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.
• Open weights: Not applicable at the pre-product stage.
• Flagship outputs: Pre-product. Industry coverage has reported a flagship model concept named "Ceres" oriented around continual learning in production environments.

Origins

Core Automation emerged in 2026 following Jerry Tworek's January 2026 departure from OpenAI after approximately seven years at the lab. Tworek had served as vice president of research at OpenAI through 2025, with industry coverage characterizing him as one of the principal architects of the lab's reasoning-model program (the o1 and o3 model lines). His prior contributions across the OpenAI tenure included work on Codex (the code-generation model line that anchored GitHub Copilot), early reinforcement-learning research applied to robotics, and additional reasoning and capability-research lines.

The founding thesis combines two structural commitments. First, that current frontier-AI research workflows have a persistent automation gap: large frontier labs run experimental campaigns that depend heavily on senior-researcher judgment rather than automated workflow execution, and that gap represents both a productivity opportunity and a competitive advantage if a new lab can systematically close it. Second, that the dominant pretraining-then-reinforcement-learning paradigm has known scaling limits, and that learning algorithms enabling continual model updates from production-environment experience can produce capability gains that conventional pretraining-scaling can no longer match.

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 Tworek's founding team. The company's founding-period public framing explicitly described the lab as targeting "the world's most automated AI lab," with research-workflow automation as the central operating commitment.

The company's fundraising activity began essentially at the moment of formation, with industry coverage reporting target cumulative-round size between $500 million and $1 billion at a post-money valuation exceeding $5 billion. Specific lead investors and round-close dates have not been confirmed in public reporting at the time of writing.

Mission and strategy

Core Automation's stated mission is to build "the world's most automated AI lab" and to develop learning algorithms that go beyond pretraining. The strategic premise reflects the conviction that conventional pretraining-scaling has approached diminishing-returns territory, and that the next-generation capability advantage requires architectural and algorithmic innovation beyond the transformer-and-RLHF paradigm that has anchored frontier-AI research through 2025.

The strategy combines three threads. First, foundational research on continual learning, with the reported "Ceres" model concept oriented around models that can update their weights in production from real-world experience rather than depending on infrequent training-cycle updates. Second, research-workflow automation, with the explicit operating commitment that senior-researcher time should be redirected from experimental-execution toward higher-leverage research-direction-setting work. Third, exploration of architectures designed to scale better than transformers, with the explicit framing that architectural diversity beyond transformers may be necessary for the continual-learning research direction.

The competitive premise reflects Tworek's prior track record as the principal architect of OpenAI's reasoning-model program, the depth of senior-researcher hires from OpenAI, Anthropic, and Google DeepMind, and the structural availability of capital at frontier-lab scales for new entrants in the post-2025 environment. Industry coverage has characterized Core Automation as one of the principal next-wave frontier-research startups alongside the broader insurgent-lab cohort founded by senior frontier-lab departures (the Thinking Machines Lab of Mira Murati and others).

Distribution channels are not publicly defined given the company's pre-product status.

Models and products

• Ceres. Reported model concept oriented around continual learning in production environments. Industry coverage has framed Ceres as a single model capable of weight-updates during operation, with the goal of reducing training-data volume by approximately 100x relative to current state-of-the-art models. Specific architecture, training-data composition, and timeline disclosures have not been confirmed.
• Research-workflow automation tooling. Reported internal commitment to automating frontier-AI research workflows. Specific tooling and product positioning have not been publicly disclosed.

Distribution channels are not publicly defined given the company's pre-product status.

Benchmarks and standing

Core Automation's eventual evaluation framework will likely combine standard frontier-AI benchmarks (Artificial Analysis Intelligence Index, GPQA Diamond, SWE-bench Verified, ARC-AGI Challenge) with continual-learning-specific metrics not yet established as industry standards. The company's pre-product status means that public benchmark results have not been reported as of May 2026.

The Tworek-team track record from OpenAI's reasoning-model program is the principal credibility signal at the pre-product stage. The o1 and o3 model lines that Tworek led at OpenAI established the reasoning-model paradigm that has shaped frontier-AI capability research through 2025 and 2026, and the senior-researcher hires from OpenAI, Anthropic, and Google DeepMind extend that credibility across the founding team.

Leadership

As of May 2026, Core Automation's senior leadership and team includes:

• Jerry Tworek, Co-Founder and Chief Executive Officer. Former vice president of research at OpenAI (departed January 2026); led the development of the o1 and o3 reasoning-model lines and contributed to Codex, early reinforcement-learning robotics research, and additional capability-research programs across approximately seven years at the lab.
• Senior researchers from OpenAI, Anthropic, and Google DeepMind. Industry coverage characterized the founding team as having been "nerdsniped" from the three frontier labs, with specific identities for many of the senior-researcher hires not yet disclosed in public reporting.
• Co-founders. Tworek's specific co-founders have not been comprehensively reported in industry coverage at the time of writing.

The team's hybrid composition (concentrated senior-researcher recruitment from the three principal frontier labs) has been characterized in industry coverage as one of the more notable founding-team profiles among 2026-vintage AI insurgent labs.

Funding and backers

• Founding-period fundraising (April 2026 reporting): Reportedly seeking $500 million to $1 billion in cumulative capital at a target post-money valuation exceeding $5 billion. Specific lead investors and round-close dates have not been confirmed in public reporting at the time of writing.

The fundraising activity began essentially at the moment of company formation. Industry coverage has characterized the target round-size and valuation as consistent with the broader 2025 to 2026 trend of senior frontier-lab departures attracting frontier-lab-scale founding capitalization.

Industry position

Core Automation occupies a structurally notable position among 2026-vintage AI insurgent labs. The combination of Tworek's prior frontier-lab leadership track record, the depth of senior-researcher recruitment from OpenAI, Anthropic, and Google DeepMind, and the target round-size and valuation positions the company among the most prominent next-wave frontier-research startups.

Strategic risks include the multi-year pre-product timeline characteristic of frontier-AI research, the competitive pressure from established frontier labs with capital-deployment, distribution, and existing-customer-base advantages, the dependence on a single founder's prior credibility for early research-team recruitment and capital-raising, and the open question of whether the continual-learning research direction produces capability advantages that justify frontier-lab-scale capitalization.

Strategic strengths include the Tworek-team prior track record as principal architect of OpenAI's reasoning-model program, the senior-researcher recruitment from three principal frontier labs, the cap-table-formation reporting at frontier-lab-scale round size and valuation, and the architectural-thesis differentiation from peer insurgent labs through the explicit continual-learning and post-transformer research direction.

Competitive landscape

Core Automation competes with several frontier and insurgent peers across research-workflow and capability-research dimensions:

• OpenAI, Anthropic, Google DeepMind. Established frontier-research peers from which the founding team was recruited.
• Thinking Machines Lab. Insurgent peer founded by Mira Murati and other former OpenAI senior leadership. Adjacent positioning as a senior-departure frontier-research lab.
• Safe Superintelligence (SSI). Insurgent peer founded by Ilya Sutskever and others. Adjacent positioning as a senior-departure research lab.
• Reflection AI. Insurgent peer with frontier-research positioning.
• Periodic Labs. Insurgent peer with autonomous-AI-research-laboratory commitment in materials-science domains.
• xAI, Inflection AI. Adjacent insurgent labs with different mission positioning.

Outlook

• The progression of fundraising activity through 2026 toward confirmed lead investors and round-close.
• The cadence of public-product or research-disclosure milestones beyond the founding-period coverage.
• The translation of the continual-learning research direction into reproducible capability benchmarks.
• The competitive dynamics with established frontier labs and with peer insurgent labs across senior-researcher recruitment and capital-deployment.
• The trajectory of senior-talent recruitment from frontier labs and academia leveraging the Tworek-team founding credentials.
• The continued tension between Core Automation's multi-year pre-product timeline and the rapidly evolving frontier-AI capability environment.

Sources

• The Decoder: Ex-OpenAI researcher Jerry Tworek launches Core Automation. Founding coverage.
• Business Insider via Techmeme: Core Automation launches with talent from OpenAI, Anthropic, and DeepMind. Founding-team composition coverage.
• The Information: Ex-OpenAI Researcher's Startup Targets Up to $1 Billion in Funding. Fundraising target context.
• The Decoder: Two startups want to replace how AI learns. Continual-learning research-direction context.
• Let's Data Science: Core Automation Recruits Researchers From Anthropic, DeepMind. Senior-researcher recruitment coverage.
• Core Memory: He Left OpenAI To Think Bigger - EP 53 Jerry Tworek. Founder background context.