Ndea

Ndea is an American artificial intelligence research lab founded in 2025 in San Francisco by François Chollet, the creator of the Keras deep learning framework and the ARC-AGI benchmark, and Mike Knoop, a co-founder of Zapier. The company pursues artificial general intelligence through deep learning-guided program synthesis, an alternative paradigm to the scale-driven foundation-model approach that has dominated the frontier AI cohort. As of May 2026, Ndea has reportedly raised approximately $43.5 million in disclosed funding, with Y Combinator, Coatue Management, Factorial Capital, and Quiet Capital among its investors.

At a glance

• Founded: January 2025 in San Francisco by François Chollet and Mike Knoop. The name draws from the Greek terms ennoia (intuitive understanding) and dianoia (logical reasoning), reflecting the founding thesis that AGI emerges from combining deep learning intuition with explicit symbolic reasoning.
• Status: Private. Funding rounds disclosed through PitchBook and Y Combinator; no formal Series A announcement as of May 2026.
• Funding: Approximately $43.5 million reported. Investors include Y Combinator, Coatue Management, Factorial Capital, and Quiet Capital. Specific round structure and valuation have not been publicly broken out.
• CEO: Mike Knoop, co-founder. Former co-founder and head of AI at Zapier.
• Other notable leadership: François Chollet, co-founder. Creator of Keras (now part of TensorFlow) and the ARC-AGI benchmark. Former senior software engineer at Google.
• Open weights: None disclosed as of May 2026. Ndea's commercial roadmap and licensing posture have not been publicly specified.
• Flagship outputs: No commercial product as of May 2026. The company is hiring research staff and has positioned its work as a multi-year basic-research program rather than a near-term product launch.

Origins

Ndea was announced in January 2025 by François Chollet and Mike Knoop. Chollet had been a senior software engineer at Google for nearly a decade, where he created Keras, the deep learning framework that became one of the most widely used Python AI libraries. Chollet is also the author of the ARC-AGI benchmark, an evaluation designed to measure abstract reasoning and pattern induction capabilities that scale-trained large-language models have historically struggled to match. The benchmark has been characterized as a stress-test of generalization rather than memorization.

Knoop is a co-founder of Zapier, the workflow-automation company that exited to a private valuation north of $5 billion. He served as Zapier's head of AI before transitioning to the Ndea co-founder role. Both founders are also co-founders of the ARC Prize Foundation, a nonprofit that runs the annual ARC-AGI competition.

The founding thesis is that progress toward AGI requires combining the pattern-recognition strengths of deep learning with the systematic-reasoning strengths of program synthesis, rather than relying on continued scaling of language model parameter counts. The framing is a deliberate counter-position to the scaling-laws thesis that has driven OpenAI, Anthropic, and Google DeepMind capital expenditure on increasingly large foundation-model training runs.

The 2025 launch announcement specified that Ndea would be a research lab, with hiring focused on research scientists and research engineers and remote work supported. PitchBook data through May 2026 indicates approximately $43.5 million raised across one or more rounds, with Y Combinator, Coatue Management, Factorial Capital, and Quiet Capital among the disclosed investors. The company's headcount has been reported at approximately 15 employees.

Mission and strategy

Ndea's stated mission is to develop AGI capable of "true invention, adaptation, and innovation" through deep learning-guided program synthesis. The strategic premise reflects Chollet's longstanding public position that pure scaling of next-token prediction has structural limitations on out-of-distribution generalization, and that program synthesis (the search for explicit programs that produce desired input-output behavior) is a complementary approach that addresses those limitations.

The strategy combines two threads. The first is foundational research on deep learning-guided program synthesis methods, where neural networks guide the search through program space rather than directly produce outputs. The second is the construction of evaluation frameworks (centered on ARC-AGI) that measure progress on the kinds of generalization that the founder team considers central to AGI. The strategic framing positions Ndea as a research-first organization with an explicit basic-research orientation rather than a near-term commercial output.

The competitive premise is that the scaling-laws path will encounter diminishing returns on capabilities that require systematic reasoning, abstraction, and out-of-distribution generalization, and that a program-synthesis-augmented architecture will outperform pure scale on those capabilities. The premise is contested within the broader AI research community; the empirical resolution depends on whether scale-trained models continue to close the ARC-AGI gap or whether the gap proves structurally durable.

Models and products

• Research output. Ndea has not published model checkpoints or commercial products as of May 2026. The company's research output is positioned as the unit of progress.
• ARC-AGI involvement. Both founders are co-founders of the ARC Prize Foundation. The ARC-AGI benchmark is the principal evaluation harness through which Ndea measures progress.
• Hiring and team. The company has been hiring research scientists and research engineers since launch, with remote work supported.

There is no commercial API, consumer-facing product, or open-weights release as of May 2026. Distribution and commercialization are deferred until the research program produces results that the founders deem ready for product translation.

Benchmarks and standing

Ndea has not published benchmark results against the standard frontier-AI evaluation suite (Artificial Analysis Intelligence Index, LMArena, GPQA Diamond, SWE-bench Verified, AIME 2025, HumanEval+) because the company has not released a model. The principal benchmark associated with the founder team is ARC-AGI, where progress has been driven by external research groups including Poetiq, the labs participating in the annual ARC Prize, and the frontier foundation-model labs whose newest releases have produced state-of-the-art ARC-AGI results.

Industry coverage has characterized Ndea as a high-profile research bet on a non-consensus paradigm. The standing within the broader AI research community draws on Chollet's long track record (Keras, the original ARC-AGI paper, public commentary on AGI capability) and Knoop's commercial credibility. Whether the bet produces measurable research output that converts into capability gains is the open question.

Leadership

As of May 2026, Ndea's senior leadership includes:

• Mike Knoop, co-founder and Chief Executive Officer. Former co-founder and head of AI at Zapier.
• François Chollet, co-founder. Creator of Keras and ARC-AGI. Former senior staff software engineer at Google.

The senior research bench has not been broadly publicly disclosed. The company has been hiring research scientists, research engineers, and operations staff through its launch and subsequent recruiting.

Funding and backers

Ndea's funding history through May 2026:

• Approximately $43.5 million reported across disclosed rounds, per PitchBook data.
• Investors include Y Combinator, Coatue Management, Factorial Capital, and Quiet Capital.
• The specific seed-versus-Series A structure, valuation, and round dates have not been publicly announced.

The investor base spans top-tier institutional VCs (Coatue), early-stage specialists (Y Combinator, Factorial), and Quiet Capital. The disclosed funding is modest relative to the foundation-model labs (OpenAI, Anthropic, xAI) that operate at multi-billion-dollar capital scales but is consistent with a research-first basic-research lab.

Industry position

Ndea occupies an unusual position in the broader AI lab cohort. The combination of the founder team's research credibility, the explicit non-consensus paradigm bet on program synthesis, and the modest disclosed capital base produces a profile that the foundation-model frontier labs do not match. Industry coverage has characterized the company as a "research bet" rather than a commercial competitor, with the principal value-add being the long-term resolution of the scaling-versus-program-synthesis debate.

The position carries structural risk. If scale-trained foundation models continue to close the gap on ARC-AGI and similar reasoning evaluations (as some recent releases suggest), the program-synthesis paradigm may prove unnecessary. Conversely, if the gap proves durable, Ndea's positioning as the principal commercial-research lab pursuing the alternative paradigm becomes strategically central.

Competitive landscape

Ndea's competitive landscape spans the AGI-research category and the broader frontier AI cohort:

• OpenAI, Anthropic, Google DeepMind. The principal scaling-laws-based foundation-model labs. The strategic disagreement on AGI architecture is the principal axis of competition.
• Safe Superintelligence, Thinking Machines Lab, Reflection AI. Other research-first AI labs founded by senior frontier-lab alumni, with varying paradigm positions.
• Poetiq. The closest direct peer on ARC-AGI-focused research, with a meta-system approach that combines existing foundation models in self-improving agentic configurations.
• University research groups. Academic AI research on program synthesis, neuro-symbolic AI, and abstract reasoning. Ndea draws talent from and contributes to this academic ecosystem.

Outlook

Several open questions affect Ndea's trajectory in 2026 and 2027:

• Initial research-output milestones, including any publications or model releases that demonstrate the program-synthesis-guided approach in practice.
• Senior research-and-engineering hiring depth, given the specialized talent pool for program synthesis and neuro-symbolic AI.
• Subsequent fundraising, with a Series A as the next likely pricing event.
• ARC-AGI progression and whether scaling-laws-based foundation models continue to close the benchmark gap.
• Commercial product translation timing, with the trade-off between basic-research depth and commercial-revenue urgency.
• The broader research-community reception of Ndea's published work as the methodology becomes externally evaluable.

Sources

• TechCrunch: AI researcher François Chollet founds a new AI lab focused on AGI. Launch announcement.
• VentureBeat: Forget Nvidia: Ndea wants to build AI that keeps improving on its own. Strategy and paradigm coverage.
• The Decoder: Ndea's deep learning-guided program synthesis. Methodology framing.
• Y Combinator: Ndea profile. Investor-side reference.
• PitchBook: Ndea profile. Funding aggregation.
• ARC Prize Foundation. Benchmark and prize program founded by both Ndea co-founders.