•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
According to 36Kr, in the summer of 1956, a small group of scientists gathered on the Dartmouth campus to lay the first bricks for the concept of artificial intelligence. Today, even leading researchers face the risk of being replaced by the very technology they helped create.
Recently, a startup has emerged from Silicon Valley with a claim that has drawn significant attention: Recursive Superintelligence says it has found a “short cut” toward advanced AI.
In just four months since its founding, Recursive Superintelligence has secured $500 million in funding, bringing its valuation to $4 billion. The company’s pitch centers on a bold promise: AI that is not only smarter, but able to learn and evolve itself without human hands.
The involvement of Google Ventures (GV) and Nvidia among the shareholders suggests the funding is not merely speculative hype, but a strategic bet by major technology investors.
Recursive is founded by Richard Socher, who is described as a less familiar public figure but a highly influential researcher. Socher was born in 1983 in Germany and studied under Andrew Ng at Stanford. He helped develop neural network methods for natural language processing, which the article links to the foundations of BERT and GPT models.
The article notes that Socher has more than 180,000 scientific citations on Google Scholar. Before Recursive, he founded MetaMind (sold to Salesforce) and You.com, an AI-powered search tool valued at $1.5 billion.
It also highlights that Recursive’s core team includes key personnel who previously left Google DeepMind and OpenAI. The article says they departed not primarily for money, but for “freedom,” citing increased rigidity at large labs due to compliance, KPIs, and pressure to commercialize.
The company’s name, “Recursive,” is presented as a technical reference to recursion—where a function calls itself. In the company’s vision, Recursive Superintelligence is a system that studies itself to create a better AI version, repeating the process in a spiral upward.
Recursive aims to build a “self-directed scientific research system.” The article contrasts this with the traditional research workflow—Observe, Hypothesize, Design experiments, Run experiments, Analyze results, Adjust hypotheses—describing it as costly and slow.
Instead, Recursive wants AI to manage the full chain. The article describes an AI system that can identify why a model is underperforming, propose a new algorithm, and coordinate resources to test and fix errors. If successful, the article says the role of human scientists could be displaced from the loop.
The article asks why Google and Nvidia would fund a four-month-old company with no publicly available product. It frames the rationale as strategic competition and supply-chain demand.
For Google, the article says the investment is a bet to compete directly with DeepMind. While DeepMind has achieved milestones such as AlphaFold and AlphaGeometry, the article characterizes these as solving specific scientific problems rather than automating the discovery process itself.
For Nvidia, the logic is described as more direct: the fuel for self-learning AI is computation. If AI begins running recursive experiments, the article argues, GPU demand could rise beyond linear growth. In that view, Nvidia’s investment is portrayed as nurturing a future customer that could consume billions of dollars of chips.
The $4 billion valuation for a startup without a publicly available product is described as likely to raise eyebrows. However, the article argues that in the AI era, valuation metrics based on revenue or user growth are increasingly outdated.
It states that venture capitalists are valuing companies based on “capabilities to own intellectual property” and the fear of missing out. After lessons from OpenAI, the article says investors do not want to be late in the race for superintelligence.
In this framing, the article presents Richard Socher’s name and the roster of engineers from DeepMind as a “fixed asset” supporting the valuation—suggesting investors are buying access to a future direction rather than a current product.
The article describes a scenario in which Recursive achieves its goals, leading to an “Intelligence Explosion.” It argues that if an AI system can upgrade itself faster than humans can upgrade it, the pace of technological evolution could move beyond human control.
On the positive side, the article says it could accelerate discoveries such as cancer cures, new superconducting materials, and climate-change solutions within months rather than decades. On the negative side, it warns that without robust ethical safeguards, a self-evolving system operating without human oversight could produce unpredictable consequences.
Source: 36Kr
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
