Founder agreements, without the folklore.

Carta’s data shows equal two-founder splits climbing from 31.5% of teams in 2015 to 45.9% in 2024. For three-founder teams, equal splits rose from 12.1% to 26.9% over the same period. Where splits are unequal, the median has narrowed from 55/45 toward 51/49.

The academic source behind most “don’t split equally” advice is a real, peer-reviewed study — Wasserman and Hellmann, based on 1,476 founders across 511 ventures. It found that roughly a third of founding teams simply default to an equal split. What gets lost in the retelling is the authors’ own conclusion about why equal splits correlate with lower valuations: in their words, it is not a direct causal effect, but a pattern driven by unobserved differences between the kinds of teams that choose to split equally in the first place. The paper itself asks readers to stay cautious about how strong that relationship really is.

Dynamic equity models such as Slicing Pie address a real problem — contribution changes over time in ways a fixed split cannot capture — but we found no independent evidence that they reduce conflict, only the model’s own marketing. Y Combinator’s advice runs the other way entirely: keep the mechanism simple, and treat equity splitting as somewhere you should not be innovating.

Three separate US appeals decisions have turned on the same two words. An agreement that says a founder agrees to assign future intellectual property has assigned nothing — it is a promise to do so later. An agreement that says a founder hereby assigns transfers it immediately. Stanford lost ownership of its own patents over this distinction in a case that reached the US Supreme Court.

One correction worth making explicitly: the famous Winklevoss dispute with Facebook is usually filed under “founders’ agreement gone wrong.” It was not one — there was no agreement between the parties to begin with, only an alleged verbal understanding. What actually went to appeal was whether a later settlement could be enforced, and it was, on the strength of a mutual release clause. The lesson is the same either way: nothing verbal survives a disagreement.

Who owns the code by default

If a contractor writes your software with no signed assignment, ownership depends entirely on which country you are in — and the assumption baked into most US-written advice does not travel.

And one thing no agreement can fix retroactively: you cannot assign intellectual property to a company that does not exist yet. Work done before incorporation belongs to whoever did it, personally, until it is formally assigned afterward — which is why the IP assignment is usually part of what a founder receives their first shares in exchange for, not a separate step.

This is the part almost nothing written about founder equity mentions, and it is the one most likely to matter to a Dutch founder building an international team.

The Netherlands: forward vesting runs into a notary, and leaver clauses run into tax

Dutch law requires a notarial deed — executed in front of a Dutch civil-law notary — for every issuance or transfer of a BV share, with the single exception of shares placed at the company’s founding. Vesting month by month, the way US advice describes it, would mean a notary appointment every month. Dutch practice runs the other way instead: shares are issued up front, with an obligation written into the company’s own articles to hand back the unvested portion if a founder leaves, often routed through a foundation that holds legal title so the economic interest can change hands without a notary visit each time. A second trap the US playbook does not warn about: the Dutch tax authority has named exactly this kind of good-leaver, bad-leaver and drag-along clause as a factor that can push founder shares into a less favourable tax treatment, taxed as income rather than as a capital gain.

The United Kingdom: not the UK’s 83(b), whatever it is called

Shareholders’ agreements in the UK usually stay private — unlike the company’s Articles, which are filed at Companies House and public, and unlike an agreement every single shareholder has signed, which UK law can require to be filed regardless. Founder shares subject to vesting can also trigger a specific election that gets described online as “the UK’s 83(b),” and it is not that. It has to be filed within fourteen days rather than thirty. It has to be signed jointly by the company and the founder, not by the founder alone. And it is never sent to the tax authority at all — both sides simply keep a signed copy in case it is ever questioned later. Three separate ways the American habit fails to carry over, not one.

The United States: the most repeated advice, and the least portable

The 83(b) election — file within 30 days of receiving restricted stock, or risk being taxed on the full value as it vests rather than at grant — is probably the single most repeated piece of founder tax advice on the internet, and it is a creature of US tax law specifically. For a founder holding shares in a Dutch BV, a UK Ltd or most UAE structures, there is nothing to file, because the concept does not exist there.

The UAE: the same choice decides tax and paperwork

DIFC and ADGM run as common-law free zones where a written transfer instrument is enough to move shares — no notary appears anywhere in either zone’s company law. A UAE mainland company runs on the opposite principle: transferring a founder’s stake requires what the law calls an officially authenticated document, and any change to the company’s own constitution is void if it has not been authenticated first by the relevant licensing authority, in person or electronically. A notary public only enters the picture in the specific cases that authority designates. A 2025 update to the law now lets companies write drag-along and tag-along rights directly into that constitution, which onshore founders previously had to rely on a private side agreement for, with far less certainty it would hold up.

Two things worth knowing before you go looking yourself. Y Combinator, despite being cited constantly as a source for founder agreements, publishes none — its public library covers fundraising documents (SAFEs) only. And for the Netherlands specifically, we could not find a free, institutionally published co-founder or shareholders’ agreement template at all. What exists there is commercial template services, not the kind of free, reputable document available in the US, UK or Germany.

A few numbers circulate constantly in founder-agreement content and do not survive a check. “65% of startups fail because of co-founder conflict” traces through a chain of retellings with a sample size that changes at every step and no page number anyone can point to. “Equal splits cause lower valuations” overstates a correlation the underlying study explicitly says is not causal. And the precise equity percentages usually quoted for Facebook’s Eduardo Saverin have no public source — that dispute was settled under a confidentiality agreement, and the numbers in wide circulation come from a book and a film, not a filing.