Richard Socher's Recursive, building an AI-research-automation 'Eureka Machine,' described as having raised a $4.65B seed round
2026-09-14
In a Latent Space interview, Richard Socher (co-founder of You.com, AIX Ventures, and now Recursive) discusses Recursive, which he says has raised a $4.65 billion seed round to build systems that automate AI research itself — what he calls the 'Eureka Machine.' Socher says an early Recursive research system outperformed humans and their agents on optimization tasks in under two days, and separately discovered Nvidia GPU kernel improvements without a team of CUDA experts. No primary funding announcement or specific date for the raise is given in this material.
Significance 3: A seed round of this reported size would qualify as a significant new entrant, but the only source is a podcast interview restating the figure in passing rather than a primary funding announcement, so significance is held below the 'major new entrant' ceiling pending corroboration.
Implications · machine-drafted, not owner judgment
If the funding figure holds up, Recursive becomes a well-capitalized entrant explicitly targeting recursive self-improvement of AI research — a more direct assault on the research-automation frontier than general-purpose coding agents, worth tracking alongside frontier labs' own automated-research efforts. The GPU-kernel-optimization claim, if independently verified, would also be a notable data point on how much specialized engineering work agents can already replace without domain experts.
- A primary-source funding announcement (press release, filing, or Recursive's own site) confirming the $4.65B figure and date
- Independent reproduction of the reported optimization-task or CUDA kernel results
Sources
- curator Humanity's Last Invention — Richard Socher of Recursive retrieved 2026-09-14