The history of artificial intelligence is a history of constraints.

For years the limit was data, until the internet and a wave of data providers supplied enough labeled examples to learn from. Then it was compute, until GPUs and the cloud made it easier to train the models at scale.

Today, we believe the constraint in AI is the pace of research itself. The industry can only move as fast as a finite number of researchers can dream up experiments, run them, and read the results manually.

It's only a matter of time until artificial intelligence becomes the primary driver improving itself. And when a machine can do research as well as the people building it, the domain experts stop being a bottleneck.

Today, we're talking about:

Most of the progress in AI over the last decade can be traced back to a few dozen people. Eight of them started Recursive.

The founding team has laid the groundwork for tools you already use. One co-invented retrieval-augmented generation (RAG), the technique that lets virtually every modern AI system look something up before it answers. Another co-authored the Vision Transformer, the 2020 paper that sits underneath today's image models. Another ran Meta's Fundamental AI Research lab. Another spent his career on AI that rewrites and improves its own code, which is a comparable problem to the one Recursive is now trying to solve at scale. The rest came from OpenAI, Google DeepMind, and Salesforce.

It's one of the strongest concentrations of AI research talent we've seen assembled outside the leading labs. And the researcher-entrepreneur who pulled them together, Richard Socher, has the standing to lead a team this dialed.

Back in 2013, the most advanced systems still couldn't reliably tell you whether a movie review was positive or negative. Socher, a Stanford PhD student at the time, thought the whole approach was wrong. His paper that year, with Christopher Manning and Andrew Ng, taught machines to read a sentence that way. It was a foundational step toward models understanding meaning.

He went on to found MetaMind (acquired by Salesforce in 2016), serve six years as Salesforce's Chief Scientist, and build You.com, an AI-first search engine, now valued at $1.5 billion, before "AI search" was a category every major company was racing to build.

Recursive's thesis is that the next leap in AI happens when the research loop can run without a human at every step. Their first target is automating AI research itself.

Rather than waiting for a researcher to design the next training method, their systems generate hypotheses, run experiments, and build on their own results.

The first goal is a system with the research capability of "50,000 PhDs," pointed at the science of AI. Once that engine runs, they intend to aim what they call a "Eureka machine" at the harder quantitative problems behind it like drug discovery, battery chemistry, fusion physics.

If AI building better AI sounds circular, one helpful precedent is the compiler. Before compilers, programmers wrote machine code manually. Then they built tools that wrote the low-level code for them, and the result wasn't fewer programmers. It was faster development, more ambitious software, and whole categories of programs that weren't worth attempting before. The tool didn't replace the work; it removed the ceiling. Recursive is making the same bet for science.

This is the bet behind our participation in Recursive’s recently announced $650 million round at a $4.65 billion valuation, led by GV and Greycroft, with participation from NVIDIA and AMD Ventures.

To be frank, this appears to be an extraordinary price for a company founded a year ago with no publicly accessible product, and it deserves to be addressed directly. We don't think it is appropriate to underwrite Recursive with the same methodology used to evaluate a traditional SaaS business. We underwrote this investment primarily to the potential outcome if its thesis is right.

If the Recursive team can build AI that meaningfully accelerates its own research, the outcome may be more than a successful startup. It could become the foundation for future discovery in most fields. At those stakes, the question stops being whether $4.65 billion is expensive and starts being whether this is one of the few teams capable of getting there. We think it is possible.

USVC's investment thesis starts with founders. We believe that, at the earliest stages, the best businesses are often indistinguishable from mediocre ones on paper. What separates them is who is building, and whether that person has the specific depth required to keep solving the problem long after the initial insight stops feeling exciting.

The risks are real, and worth stating plainly. Recursive is early-stage and working in one of the most technically demanding and competitively intense corners of technology. There is no guarantee the research succeeds on the timeline they project, or that they outpace labs chasing the same goal with larger teams and deeper pockets.

Today, we are announcing that USVC has invested directly in Recursive Superintelligence, alongside GV, Greycroft, NVIDIA, and AMD.

A company with this much investor demand normally has its pick of capital and keeps its cap table short and institutional. We negotiated this allocation to make it accessible to our investors.

It's a direct position with no underlying management fee and no carried interest, so when this investment creates value, it flows to USVC without an extra layer of cost in between.

To be precise about what you own: you do not hold Recursive shares directly. You own USVC shares, a regulated fund that holds Recursive as one investment alongside other private companies and emerging managers.

What that gives you is economic exposure, through your USVC shares, to one of the most ambitious bets in AI, as part of a broader portfolio.

Getting into a company like this before it's the obvious winner is exactly what USVC was built for.

U.S. investors can get started with as little as $500. No accreditation required.

Read the full announcement →

Private markets have historically rewarded patient capital. The investors who've built real positions in venture didn't find the perfect moment. They started early and stayed long.

USVC is how you build that position, starting with as little as $500.

Investors should carefully consider the investment objectives, risks, sales charges and expenses of USVC before investing. USVC's prospectus contains this and other information and may be obtained at http://usvc.com/prospectus or by calling +1 (844) 988-1720. Read the prospectus carefully before investing.

This communication is for informational purposes only, is not intended to be a recommendation for any investment or other advice of any kind and shall not constitute or imply any offer to purchase, sell or hold any security or to enter into or engage in any type of transaction. Any such offers will only be made pursuant to USVC's prospectus, which should be carefully reviewed before investing.

Investing in the USVC Venture Capital Access Fund involves significant risk, including the possible loss of principal. Venture capital investments are speculative, illiquid, and subject to a high degree of risk. Past performance does not guarantee future results.

USVC Venture Capital Access Fund is distributed by North Capital Private Securities (NCPS), member FINRA/SIPC. NCPS is not affiliated with USVC’s adviser or its affiliates.

Investing in USVC's shares involves substantial risk, including the potential loss of your entire investment. Shares are not listed on any exchange, are illiquid, and liquidity is limited to periodic repurchases at the discretion of the Board, which are not guaranteed. This investment is speculative and suitable only for long-term investors who can bear the risks of limited liquidity. Certain conflicts of interest involving USVC and its affiliates could impact USVC's investment returns and limit the flexibility of its investment policies. Past performance does not guarantee future results. Fees, expenses, and conflicts of interest may reduce returns.

USVC's shares have no history of public trading. You should not expect to be able to sell your shares other than through USVC's repurchase policy, regardless of how USVC performs. USVC does not intend to list its shares on any securities exchange during the continuous offering, and it does not expect a secondary market in the shares to develop.

USVC invests in private funds which are subject to certain risks including those related to illiquidity, indirect fees, valuation, limited operating histories and limited information regarding underlying investments. As a result of the foregoing, an investment in USVC's shares is not suitable for investors that require liquidity, other than liquidity provided through USVC's repurchase policy. The amount of distributions that USVC may pay, if any, is uncertain.