Recursive launches agentic AI startup

Recursive Superintelligence emerged from stealth on May 13 with more than $650 million in funding and a plan to build AI systems that improve themselves. The startup has offices in San Francisco and London, fewer than 30 employees, and a founding team drawn from OpenAI, Google DeepMind, Meta and Salesforce, according to the New York Times and other reports. Richard Socher, the former Salesforce chief scientist, is among the co-founders. The company says its research focus includes recursive self-improvement, world models and optimization. ### Who is behind Recursive? Richard Socher and Tim Rocktäschel are the best-known names attached to the company. The New York Times reported that Socher founded the startup with seven other researchers, while outside reports named Rocktäschel, Jeff Clune, Josh Tobin and Tim Shi among the co-founders. Rocktäschel is a professor at University College London and a former Google DeepMind scientist. (nytimes.com) The founding roster reflects the same talent pattern that has defined several recent frontier AI startups: small teams, senior researchers and immediate access to large pools of capital. The New York Times said the company has fewer than 30 employees, while earlier reports described a team of about 20 before the public launch. ### What is the company actually trying to build? (nytimes.com) Recursive says it is working on self-improving AI rather than a conventional chatbot or enterprise assistant. Reports describing the company’s pitch say it wants to automate parts of the AI research process itself, including selecting research directions, running experiments, evaluating results, choosing training data and refining models. (nytimes.com) World models and optimization are central to that pitch. Rocktäschel’s background is tied to world-model research at DeepMind, and outside coverage of the launch said the company is drawing on expertise in open-ended algorithms, quality-diversity methods and self-improving coding agents. Those descriptions line up with the hiring language highlighted in social posts about the launch, which pointed to demand for engineers and researchers with agent, world-model and optimization experience. (siliconangle.com) ### How much money did it raise, and from whom? Recursive disclosed more than $650 million in funding at launch, according to the New York Times and SiliconANGLE. The New York Times said investors included Google Ventures, Greycroft, Nvidia and AMD, and described the company as valued at more than $4 billion. SiliconANGLE separately reported a $650 million raise led by GV and Greycroft. (officechai.com) April reporting had pointed to a smaller figure. The Financial Times, as cited by The Decoder and other outlets, previously reported that the startup had raised at least $500 million at a $4 billion pre-money valuation in a round led by GV with Nvidia participating. The larger May figure suggests the financing expanded by the time of the public launch. (nytimes.com) ### Why are San Francisco and London the two locations to watch? San Francisco and London are the company’s two named hubs. The New York Times said the company has offices in both cities, and launch coverage described the team as split across the two locations. That matters because the hiring push attached to the launch is not broad consumer hiring; it is concentrated around frontier research and engineering. (the-decoder.com) The social post cited in the launch context pointed readers to open jobs in San Francisco and London. While the company’s own careers page was not directly accessible in this search, multiple reports tied the launch to active hiring and described the team as still small. That makes the staffing signal straightforward: Recursive is using its launch and financing to recruit more researchers and engineers in the two main AI talent markets it has chosen. (dnyuz.com) ### What happens next? May 13 was the company’s public debut, but the next concrete step is hiring. Reports around the launch say Recursive is expanding beyond its sub-30-person team, and the open roles flagged in San Francisco and London focus on the research and systems work behind agentic AI. The company has not publicly outlined a product launch date in the sources reviewed here. (siliconangle.com) For now, the milestones investors and recruits can track are the build-out of the San Francisco and London teams, the company’s published job openings, and any future research or product disclosures from Socher, Rocktäschel and the rest of the founding group. (nytimes.com)