AI researcher François Chollet and Zapier co-founder Mike Knoop launch Ndea, an AI research and science lab focused on “developing and operationalizing AGI”
François Chollet, an influential AI researcher, is launching a new startup that aims to build frontier AI systems with novel designs.
Context & Ripple Effects
Ndea combines François Chollet’s critique of how AI intelligence is measured, including his work on ARC-AGI, with Mike Knoop’s experience as a Zapier co-founder. The lab is positioned around both frontier-system research and the practical challenge of deploying it efficiently.
The launch sits within a growing cluster of independent AI labs founded by senior researchers and operators, including Core Automation’s push for an automated AI lab. It also contrasts with later efforts to productize agents through managed-agent deployment tooling.
First-order effects
- Chollet leaves Google to build Ndea with Knoop, shifting his primary work from a large-platform research role to an independent lab while remaining involved with Keras externally.
- Ndea’s early technical and deployment choices will be shaped by Chollet’s emphasis on better intelligence evaluation and cost efficiency, rather than treating model scale alone as the objective.
Second-order effects
- Ndea adds another competitor for frontier-AI research talent and compute, while its GPU-supply warning makes efficient architectures and deployment economics a more immediate design constraint.
- If Ndea can translate alternative evaluation methods into usable systems, developers and enterprises gain another potential path between research benchmarks and operational AI products.
Third-order effects
- The launch points toward a more institutionalized frontier-AI market in which independent labs pair research theses with explicit commercialization and infrastructure strategies.
- As labs compete on both capability and operating cost, evaluation methods, hardware access, and deployment discipline may become as strategically important as raw model scale.
The trend: Frontier AI is moving from research inside major platforms toward independent, operationally focused labs that must prove both novel capabilities and viable deployment economics.