General Intuition

General Intuition is an American artificial intelligence research lab headquartered in New York City, focused on building foundation models and AI agents with spatial-temporal reasoning capability. The company was spun out of Medal (a video-game clip-sharing platform) in 2024 by Pim de Witte (Medal's founder), Adam Jelley, Eloi Alonso, and Vincent Micheli, with the strategic premise that gameplay video at scale supplies a previously underexploited training signal for general spatial-reasoning models. As of May 2026, General Intuition is one of the principal world-model and embodied-AI insurgent labs, with a $134 million seed round announced October 2025 led by Khosla Ventures and General Catalyst with participation from Raine.

At a glance

• Founded: 2024 in New York City. Public launch October 2025.
• Status: Private. Seed round of $134 million announced October 2025 with valuation reported in industry coverage at approximately $2 billion at full unicorn-class headline framing.
• Funding: $134 million seed round announced October 2025, led by Khosla Ventures and General Catalyst, with Raine participating. Spun out of Medal, the gameplay-clip sharing platform founded by Pim de Witte.
• CEO: Pim de Witte, Co-Founder and Chief Executive Officer. Founder of Medal.
• Other notable leadership: Adam Jelley, Co-Founder. Co-author of the DIAMOND world-model research line. Eloi Alonso, Co-Founder. Lead author of the DIAMOND paper. Vincent Micheli, Co-Founder. Co-author of DIAMOND.
• Open weights: Limited. The team's prior research (DIAMOND) is open-source; commercial models are closed.
• Flagship outputs: DIAMOND world-model research (2024); spatial-temporal reasoning foundation-model program; gaming and search-and-rescue agent applications.

Origins

General Intuition was founded in 2024 as a research-lab spinout from Medal, the New York-headquartered video-game clip sharing platform that Pim de Witte had founded in 2014. Medal had accumulated, by 2025, a content library characterized in industry coverage as approximately two billion gaming videos per year contributed by a community of approximately 10 million monthly active users across tens of thousands of distinct games.

The founding thesis was that gameplay video at scale supplies a previously underexploited training signal for spatial-temporal reasoning. Game environments compress an enormous diversity of physical and social dynamics into structured, action-conditioned video streams, with explicit player-input data attached to most clips. The premise is that foundation models trained on this corpus can develop spatial-reasoning capabilities applicable beyond games, including robotics, search-and-rescue, and other embodied-AI domains.

The research team brought specific credentials. Adam Jelley, Eloi Alonso, and Vincent Micheli had been the principal authors of DIAMOND (Diffusion as a Model of Environment Dreams), the 2024 research paper that demonstrated diffusion-based world models capable of competitive performance on the Atari 100k benchmark. DIAMOND was characterized in research coverage as the first agent trained entirely within a diffusion-based world model to achieve a mean human-normalized score above 1.0 on the Atari 100k task suite, with a reported score of 1.46.

The October 2025 seed round of $134 million, led by Khosla Ventures and General Catalyst with Raine participating, was the company's most consequential public-facing milestone and was characterized by TechCrunch as one of the largest seed rounds in AI history at the time. Industry coverage noted that OpenAI had reportedly attempted to acquire Medal in the period preceding the spinout, with a reported $500 million offer that General Intuition's founding team declined.

Mission and strategy

General Intuition's stated mission is to build foundation models and general agents for environments that require deep spatial and temporal reasoning. The strategic premise is that current foundation-model architectures, trained predominantly on text and 2D image data, have a structural capability gap in three-dimensional spatial reasoning, and that the Medal gameplay-video corpus supplies a uniquely scaled training resource for closing that gap.

The strategy combines three threads. First, foundation-model research on diffusion-based world models, building on the DIAMOND research lineage and scaling beyond academic-benchmark settings. Second, the Medal data partnership, which provides exclusive training-corpus access at a scale most peer labs cannot match. Third, application focus on gaming (where the data and customer profile align most directly) and search-and-rescue drones (where spatial-reasoning capability translates to high-stakes commercial demand).

The competitive premise reflects General Intuition's positioning at the intersection of frontier research talent (the DIAMOND team), proprietary training-data access (the Medal corpus), and venture-backed scale (the $134 million seed). Industry coverage has consistently characterized the data-asset advantage as the principal structural feature distinguishing the company from peer world-model labs that work from public-domain datasets.

Distribution channels include the eventual rollout of gaming-specific applications, search-and-rescue partnerships, and continued research releases extending the DIAMOND line.

Models and products

• DIAMOND research line. Diffusion-based world-model approach demonstrated on Atari 100k (mean human-normalized score 1.46). Open-source through GitHub. Anchors the company's research credibility.
• Spatial-temporal foundation-model program. Closed-source commercial work scaled beyond DIAMOND on the Medal gaming-video corpus. The program targets foundation-scale models capable of generalizing to novel environments not seen during training.
• Gaming and search-and-rescue applications. Initial application targets characterized in launch coverage. The gaming angle aligns with Medal's existing user community; search-and-rescue addresses an early commercial-customer category in which spatial-reasoning translates to operational drone capability.

Distribution channels include planned gaming-application rollouts, drone-customer partnerships, and continued research releases.

Benchmarks and standing

General Intuition's evaluation framework focuses on world-model and spatial-temporal reasoning benchmarks rather than horizontal-LLM leaderboards (Artificial Analysis Intelligence Index, LMArena, GPQA Diamond), which do not address embodied or spatial-reasoning capability. The DIAMOND research lineage established the team's published-result credibility on the Atari 100k benchmark; subsequent commercial models have not yet entered comparable public benchmarks at scale.

Industry coverage of the $134 million seed has consistently characterized General Intuition as one of the principal commercial world-model labs alongside World Labs (the spatial-intelligence company founded by Fei-Fei Li), Decart (the Israeli real-time interactive world-model company), and Pika and similar generative-video startups. The categorical positioning has emphasized General Intuition's training-data asset and DIAMOND research lineage rather than user-facing product traction.

The eventual benchmarks expected to define the category include open-world agent generalization metrics, where novel-environment transfer is a central capability; gaming-environment performance benchmarks, where the data alignment is closest; and embodied-AI transfer benchmarks, where applicability beyond games is the central commercial question.

Leadership

As of May 2026, General Intuition's senior leadership includes:

• Pim de Witte, Co-Founder and Chief Executive Officer. Founder of Medal.
• Adam Jelley, Co-Founder. Co-author of the DIAMOND world-model research.
• Eloi Alonso, Co-Founder. Lead author of the DIAMOND world-model research.
• Vincent Micheli, Co-Founder. Co-author of the DIAMOND world-model research.
• Senior research and engineering staff across the foundation-model program and product applications.

The team's structural advantage combines the DIAMOND research credentials with de Witte's Medal-platform operating background, supporting both technical-recruiting credibility and exclusive training-data access.

Funding and backers

• Seed (October 2025): $134 million led by Khosla Ventures and General Catalyst, with Raine participating. Spun out of Medal.

The seed round has been characterized in coverage as one of the largest in AI history, reflecting both the founders' research credentials and the strategic value of exclusive Medal-corpus access. Subsequent funding rounds have not been publicly disclosed as of May 2026.

Industry position

General Intuition occupies a structurally distinctive position among 2024 to 2025-vintage world-model insurgent labs. The combination of Medal-corpus exclusive training data, the DIAMOND research lineage, and the size of the seed round produces a profile differentiated from peer labs that work from public-domain datasets or from broader generative-video positioning.

Strategic risks include the open question of whether spatial-temporal reasoning trained on gameplay video transfers cleanly to physical-world embodied tasks, the competitive pressure from World Labs and other commercial spatial-AI labs that approach the problem through different data and modeling choices, and the timeline-uncertainty associated with pre-product foundation-model research at venture-backed scale.

Strategic strengths include the Medal-corpus exclusive training data, the DIAMOND research lineage that anchors public credibility, the senior research-and-engineering team, and the seed-round size that supports multi-year foundation-model research without near-term commercial-revenue pressure.

Competitive landscape

General Intuition competes with several world-model and embodied-AI peers:

• World Labs. The Fei-Fei Li-led spatial-intelligence lab. Direct world-model research peer with the Marble product line in market.
• Decart. Israeli real-time interactive world-model startup with the Oasis demonstration. Adjacent commercial peer.
• Google DeepMind. Genie multimodal foundation models. Frontier-lab peer with comparable research scope but distinct distribution channels.
• Meta AI / FAIR. JEPA research line under Yann LeCun and the broader spatial-AI research at Meta.
• Physical Intelligence. Robotics-foundation-model startup with embodied-AI focus. Adjacent peer with different data sources.
• Wayve. Autonomous-driving foundation-model company. Adjacent embodied-AI peer.
• Pika, Runway. Generative-video startups with adjacent applications in entertainment and creative-industry tools rather than spatial-reasoning research.

Outlook

• The cadence of foundation-model release milestones beyond DIAMOND across 2026.
• The progression of gaming-application and search-and-rescue partnership rollouts.
• The translation of Medal-corpus training to embodied-task generalization beyond gameplay environments.
• The competitive dynamics with World Labs, Google DeepMind, and Meta as the world-model category matures.
• The continued tension between General Intuition's exclusive-data structural advantage and the competitive pressure from peer labs scaling alternative training-data sources.
• The trajectory of follow-on financing and senior-talent recruitment leveraging the seed-round size and DIAMOND research credentials.

Sources

• General Intuition official site. Company reference.
• TechCrunch: General Intuition lands $134M seed to teach agents spatial reasoning using video game clips. Seed-round announcement.
• SiliconANGLE: Spatial-temporal reasoning startup General Intuition closes $133.7M investment. Funding context.
• TheAIInsider: General Intuition Launches with $133.7M Seed Round. Seed-round context.
• DIAMOND research repository. World-model research from the founding team.
• Eloi Alonso research page. Co-founder reference.