The 8-Cofounder Cap Table: What Recursive Superintelligence's $650M Round Says to First-Time Founders

On May 13, 2026 Recursive Superintelligence emerged from stealth with a $650M round at a $4.65B post-money valuation, led by GV and Greycroft with participation from Nvidia and AMD Ventures [1][2]. Richard Socher, previously Salesforce chief scientist and the founder of You.com, is the CEO. He is one of eight named co-founders [3]. The other seven are Tim Rocktaschel, Alexey Dosovitskiy, Josh Tobin, Caiming Xiong, Yuandong Tian, Tim Shi, and Jeff Clune. Four of them came directly from OpenAI. Two carry the most-cited paper credits in transformer history. The roster reads less like a typical seed company and more like a research conference program.

The round set new defaults for what an at-launch frontier-AI company looks like. It also set a new trap for first-time founders trying to extract a lesson from it.

Eight co-founders, at a $4.65B post-money, with roughly $650M of outside capital. That math implies dilution of about 14% in this round. The remaining 86% is split among the eight founders and any pre-stealth angel allocations [1]. Assuming roughly equal splits with a small CEO premium, each individual co-founder is holding the equivalent of around $450M to $500M of value at the post-money mark. None of those numbers are confirmed, but the structure tells the story.

This is not a typical seed cap table. A typical seed company has one or two co-founders holding 40% to 50% each after the first priced round, with employees at 10% to 15% and investors at 20% to 25% [4]. Recursive's cap table is not a startup cap table. It is a research lab cap table dressed up with venture economics. Confusing the two is the most common mistake first-time founders are making this week.

TechCrunch's coverage of the round called out the central reason GV and Greycroft moved at this size and speed [2]. There is no product yet. There is no revenue. There is, however, a roster that any AI partner at a top-five firm can read in 90 seconds and underwrite. Investors paid a legibility premium. The team is the product, in the sense that the team is the only thing the market has to score, and the market knows how to score it.

This is the part first-time founders should understand clearly. The premium is not for having eight co-founders. It is for having eight co-founders whose individual citation pages are public, whose past employers are top labs, and whose prior projects shipped at scale. A first-time founder with eight unknown co-founders does not get the same treatment. They get a confused cap table and a slower first close.

Within hours of the announcement, Hacker News and X were full of first-time founders asking whether they should add more co-founders to look stronger on paper. The answer for almost every first-time founder is no. Three of the largest failure modes in early company building map directly to over-stuffed founding teams.

The first is decision drag. Every additional co-founder roughly doubles the meetings required to make a single product call. Second is equity drag. Adding a fourth founder at 20% means giving up a slice that any early hire with the same skills could have earned for half the share. Third, and the one founders see least often, is the legibility tax in reverse. A team with eight names and no track record reads as confused to a seed investor. The same investor would read a two-person team with one shipping product as competent. The advice first-time founders should take from Recursive is the opposite of what it appears to recommend on the surface.

Y Combinator's library on finding a technical co-founder still holds up in 2026, and the most useful frame in it is staged complementarity [5]. At pre-seed, the highest output pairing is one builder plus one seller, with one of the two also doing design. At seed, the same pairing extends to one builder, one seller, and one operator who owns finance and hiring. At series A, the team chart starts to look like an exec team, not a co-founder list. Beyond three co-founders, additions should almost always join as senior hires with meaningful equity grants rather than as named co-founders.

A planning tool like Foundra walks first-time founders through this staged thinking before incorporation, when the cap table is still cheap to change. The team mix at incorporation is the choice that everything else compounds against. Get the first three names right and the next twelve hires almost design themselves. Get the first three wrong and every later hire has to compensate for the gap.

Run the numbers before adding a fourth name to the cap table. A four-person founding team typically gives each co-founder around 22% to 24% post a typical seed round [4]. A senior engineer hired at the same point in time would receive 1% to 2.5% with a four-year vest and a one-year cliff. The implied premium of bringing someone in as a co-founder versus an early hire is roughly twentyfold. That premium is only worth paying if the new person brings two things that a senior hire cannot. The first is a skill the existing founders do not have and cannot learn in the next 12 months. The second is true ownership of an outcome the existing founders are unwilling or unable to own.

If both conditions are not true, the right move is to make the hire, not the co-founder offer. Altar's 2026 guide to finding a technical co-founder makes the same point in slightly softer language [6]. Co-founder status should be earned, not used as a recruiting incentive.

The strongest 2024 to 2026 venture data on solo versus team founders comes from the Y Combinator outcomes database. Solo founders raise smaller rounds on average and exit at lower aggregate values, but solo founders also fail at lower rates than three-plus founder teams when measured by 24-month survival [5]. Two-person teams remain the most consistent performer. The reason is unglamorous. Two people can agree quickly, disagree productively, and split work cleanly. A two-person team also has the minimum surface area for governance disputes, which Embroker's 2026 founder dispute data flags as the single largest source of legal costs in the first three years [7].

For a first-time founder reading the Recursive announcement, two is still the right answer. One is acceptable if the founder can hire well. Three or more should be treated as an exception, not a goal, and only entered into with full vesting and signed founder agreements in place from day one.

Any founding team with three or more names needs four documents in place before the first dollar of outside money lands. A reverse-vested founder restricted stock agreement with a one-year cliff and a four-year schedule, which gives the company the right to repurchase unvested shares if a co-founder leaves [4]. An invention assignment agreement signed by every co-founder and every early contractor, which prevents the IP-ownership disputes that account for nearly a third of founder lawsuits [7]. A bylaws document that specifies vote thresholds for amending the charter or removing a co-founder. And a one-page written record of original intent, signed by all co-founders, capturing who agreed to do what at the moment of incorporation.

This sounds like overhead. It is the cheapest insurance policy in startup law. The cost to put all four in place with a startup-focused firm is around $5,000 to $8,000. The cost to litigate any one of them after a fundraise begins is closer to $250,000 and 18 months of distraction.

Recursive Superintelligence is a useful data point about how the top of the AI market is being priced, not a template for the bottom 99% of first-time founders [3]. The story to take from it is that named, legible talent is the only asset venture capital is willing to pay a research-conference premium for at the moment. Almost no first-time founder has that asset to sell.

The better lesson is the inverse one. Pick the smallest co-founder team that can do the job. Reverse vest. Sign IP assignments. Write down original intent. Keep the cap table simple. Then build something a customer wants to pay for, which Recursive will also eventually have to do [2]. Talent rounds clear early. Revenue rounds clear forever.

Should I add more co-founders to look stronger on paper? Almost never. Investors at pre-seed and seed read confused cap tables as a negative signal. A two-person team with a clear builder and seller is the stronger default. Only add a third or fourth co-founder if they bring a skill the team cannot learn and an outcome only they can own.

What is the typical co-founder equity split in 2026? For a two-person team, splits of 50-50 or 60-40 are most common, with reverse vesting over four years and a one-year cliff. For a three-person team, splits of 40-30-30 or 50-25-25 are standard. Equal splits at the start are the simplest, but small adjustments for differential commitment are normal.

When does it make sense to have one founder rather than two? When the founder can ship the first version themselves and hire well. Solo founders raise smaller rounds on average, but they also avoid the governance disputes that derail many three-plus teams. If you can ship and hire, solo is a legitimate path in 2026.

What does the Recursive cap table look like in practice? Eight named co-founders sharing roughly 86% post-money, with about 14% sold to GV, Greycroft, Nvidia, and AMD Ventures in this round. None of the founders are first-time founders. The structure is a research lab dressed in venture economics, not a startup template.

What is the single biggest legal risk for a four-plus co-founder team? Missing reverse vesting. Without it, a co-founder who leaves in month four keeps all of their equity and remains a long-term claim on the cap table. Reverse vesting solves this in one document and costs almost nothing to put in place at the time of incorporation.

1. Recursive Superintelligence emerges from stealth with $650M raise (Tech.eu, May 13 2026)
2. What happens when AI starts building itself? (TechCrunch, May 14 2026)
3. Recursive Superintelligence raises $650M to build self-improving AI models (SiliconANGLE, May 13 2026)
4. Founder's Guide to Startup Cap Tables and Equity Splits (Embroker, 2026)
5. How to Find a Technical Co-Founder (Y Combinator Startup Library)
6. How to Find a Technical Co-Founder for Your Startup in 2026 (Altar.io)
7. Founder Disputes: The Cost of Skipping the Paperwork (Embroker)