General Intuition is raising approximately $300 million at a valuation of just over $2 billion, according to The Next Web. The New York-based lab spun out of gaming-clip platform Medal after founder Pim de Witte reportedly rejected a $500 million acquisition offer from OpenAI in late 2024, The Next Web reports. Reported backers for the raise include Jeff Bezos and Eric Schmidt, with existing investors Khosla Ventures and General Catalyst also named, per The Next Web. The outlet reports the company trains agents using the Medal dataset, which it describes as roughly two billion video clips per year from more than 10 million monthly active users.
What happened
General Intuition is raising approximately $300 million at a valuation of just over $2 billion, according to The Next Web. The Next Web reports the startup spun out of gaming-clip platform Medal after founder Pim de Witte reportedly turned down a $500 million acquisition offer from OpenAI in late 2024. The Next Web names Jeff Bezos and Eric Schmidt among reported backers for the current round and lists Khosla Ventures and General Catalyst as existing investors. The Next Web reports the company trains agents on gameplay footage and describes the Medal dataset as roughly two billion video clips per year generated by over 10 million monthly active users, with clips characterized as first-person, interactive gameplay.
Editorial analysis - technical context
Access to large volumes of first-person, interactive gameplay footage provides dense trajectory-and-observation pairs that are unusually rich for training embodied or world-model-based agents. For practitioners, such data can reduce reliance on engineered simulators by supplying real-action sequences and temporal dynamics useful for imitation learning, model-based planning, and training perception modules under interactive conditions. Models trained on first-person gameplay can surface different failure modes, including overfitting to game-specific physics and shortcut strategies, compared with models trained on passive spectator video.
Industry context
Companies that secure large, proprietary datasets often draw strategic investor interest because data can accelerate capabilities that are costly to reproduce. Industry reporting frames General Intuition's origin from Medal and the reported investor list as part of a broader pattern where specialized data assets become the basis for new ML labs experimenting with agentic or world-model architectures.
What to watch
Observers should look for:
- •published benchmarks or demonstrations that quantify how the Medal dataset improves agent behavior versus synthetic or spectator-video training sets
- •any licensing or access details for the dataset that affect reproducibility
- •whether the lab releases model artifacts, evaluation suites, or tooling that the practitioner community can inspect and benchmark against existing world-model and agent frameworks
Scoring Rationale #
The raise and the reported size of the Medal dataset are notable for researchers building embodied and world-model agents, and the investor roster signals strong market interest. The story is important but not a frontier-shifting model release.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.