The Convergence Counter-Argument

A companion to last week's essay. Read both. Decide which world you actually live in.

TL;DR — Last week's essay argued that VC tooling is structurally broken for anyone operating outside the US, and that whoever rebuilds it for the rest of the world captures the next decade of global venture. That argument stands. This piece is the opposite side of the same table, written by the same author. It steel-mans the position a thoughtful skeptic would take, namely that the world's startup economy is not just locally diverging from US norms but actively converging toward them, because investors and founders both find that standardization lowers the cost of capital. Both views can be defended seriously. The two essays together are the real argument. The reader is the judge.

Read the original here: The Rest of the World Is Building Companies Too. It is left exactly as published. Editing it inside its own pages to soften the thesis would dilute the debate. The whole point of this follow-up is that strong essays should be allowed to stand, and that the right way to test them is to argue with them honestly in a separate piece, not to revise them into safety.

What follows is that argument.

A note on how to read this

This is not a retraction. It is a deliberate exercise in writing the best version of the case against the original essay, by the person who wrote the original essay. Both pieces are sincere. Both reflect things I think are true. They sit at different points on the same spectrum, and the truth, as is usually the case with infrastructure arguments, is probably somewhere along that spectrum rather than at either end.

If you want my own conclusion, it is at the bottom of this piece. If you want to form your own, read the original, then read this, then ignore the conclusion and trust your gut.

The original argument, briefly

The first essay said that fund management platforms, cap table tools, readiness checkers, deck builders, and LP reporting systems are not jurisdiction-neutral. They are US-shaped by default, with non-US options appearing as cosmetic dropdowns that do not actually change the underlying model. It walked through five products and showed where the assumptions sit: USD-only formatters, "American versus European" waterfall toggles, 409A valuations on diligence checklists, SAFE notes as the default convertible, Delaware C-Corp as the implicit entity type. It concluded that whoever rebuilds this infrastructure for the rest of the world captures the next decade of global venture.

I still believe that piece. Hold it in your head as the thesis. What follows is the antithesis.

Counter-argument one: the world is converging, not diverging

The original essay frames the problem as tools forcing everyone into Delaware-shaped boxes. A thoughtful skeptic would point out that in many cases, it is the investors and founders themselves who are choosing the Delaware-shaped box, and the tooling is following the money rather than imposing the shape.

A Kenyan fintech raising a serious round is often advised to do a Delaware flip before institutional money comes in. A Nigerian SaaS company targeting US investors will frequently incorporate as a Delaware C-Corp first and then own the Nigerian operating entity as a subsidiary. Latin American startups regularly use Cayman holding companies with Delaware operating entities below them. Indian startups have for years used the Mauritius and Singapore stack, and more recently many have flipped to Delaware to access US capital markets. Southeast Asian companies frequently choose Singapore as a holding jurisdiction precisely because it is legible to global capital.

This is not the result of cultural surrender. It is an economic equilibrium. Investors prefer standardized structures because standardization lowers transaction costs. It makes diligence faster, makes term sheets more comparable, makes secondaries easier, and makes exits cleaner. A fund that has done a hundred Delaware deals can do the hundred-and-first in days. A fund that has to learn local share-class mechanics for one investment may decide the friction is not worth it for a check at that size.

If you take this view seriously, the tooling is not merely culturally biased. It is also economically optimized around dominant capital flows. The market is choosing Delaware not because the tooling forces it to, but because Delaware is the lingua franca of cross-border venture, and the tooling reflects where the deals actually clear.

The original essay can answer this, and probably should: the set of founders who will never seek US capital is large and growing, the local-capital-into-local-companies layer is genuinely underserved, and the convergence story understates what gets lost in translation when a Tunisian SARL pretends to be a C-Corp on paper for ten years. Both can be true. But the convergence point is real, and it changes the market sizing in ways the original piece did not fully confront.

Counter-argument two: incumbents are not negligent

The original essay said that Carta, AngelList, and Pulley have shown little interest in non-US jurisdictions. The skeptic's reading is that this language is too soft on the incumbents in one direction and too harsh in another. The incumbents are not negligent. They are responding to structural constraints that any new entrant will face too.

Sanctions compliance is a real cost center. Once you accept users from a wide set of jurisdictions, you take on OFAC obligations, EU sanctions list obligations, and an ongoing watchlist screening operation. KYC and AML rules vary by country, by user type, and by transaction size. A cap table platform that issues actual securities, even just digitally, sits closer to broker-dealer regulation than most readers realize.

Local securities law variance is the next problem. Issuing shares to non-US persons triggers different rules in the issuer's home jurisdiction and in each investor's home jurisdiction. No software platform can answer with confidence whether a particular share issuance in Kenya, sold to an Armenian investor, in a fund domiciled in Mauritius, requires a particular local filing. Getting that wrong has consequences that compound across the cap table.

Tax reporting is similarly thorny. K-1s are a US partnership concept. The equivalent reporting in other jurisdictions is not just a different form, it is sometimes a different tax base, computed on a different fiscal year, with different treatment of unrealized gains, with different attribution rules across investor types. A platform that emits the wrong tax artifact is not just unhelpful, it is a liability.

Then there is unit economics. Average contract values in many emerging markets are lower than in the US, and customer acquisition is harder because the buyer base is fragmented across jurisdictions, sectors, and stages. The marginal cost of supporting Tunisia is not just adding a template. It is taking on local counsel, jurisdictional liability exposure, exchange-rate operations, and a long tail of edge cases on revenue that may not justify the burden.

The skeptic concludes that the original essay treats these mostly as solvable engineering problems with clean architectural answers, when they are institutional problems with messy answers. The moat is operational and legal, not architectural.

The original essay can answer this too: someone is going to build this, and being early to the operational and legal moat is exactly the kind of work that creates a defensible business. But the skeptic is right that the architecture section in the first piece sounds cleaner than reality permits.

Counter-argument three: AI is a thinner layer than the original implied

The original essay said AI could power contextual adaptation without requiring a team of lawyers in every jurisdiction. A skeptic would point at that sentence and say it does too much work.

Large language models are genuinely useful for some of this. They can translate between vocabularies, surface relevant questions during onboarding, summarize jurisdiction-specific articles, and reduce the cognitive load of working across legal systems. Those are real wins.

What LLMs cannot do, today or in the foreseeable near term, is replace jurisdictional legal expertise in a way that holds up under audit and enforcement. A hallucinated compliance mapping is not an awkward UX moment. It is the kind of mistake that ends up in litigation. The hard part of building global VC infrastructure is not "understanding concepts." It is correctness, defensibility, and the ability to stand behind the output of the platform when an investor, a regulator, or an opposing counsel pushes back.

In the skeptic's framing, the durable moat for a global venture infrastructure company is almost certainly the operational and legal layer underneath the product. Relationships with local counsel in dozens of markets. A maintained library of jurisdiction-specific opinions and templates that get updated as laws change. An audit trail that someone can defend in front of a regulator. AI sits on top of that. It does not replace it.

The original essay can fairly say that the AI layer was offered as a translator, not a replacement, and that this is exactly what the piece argued for. But the skeptic is right that the prose around AI was the most confident part of the original piece, and confidence about AI on the boundary of legal and regulatory work is usually where smart people get themselves into trouble.

Counter-argument four: founders may not actually want localization

The original essay treats localization as inherently desirable, as if every founder in a non-US market is straining against tools that do not fit. The skeptic would point out that the revealed preference in many of these markets is for tooling that helps a founder translate upward.

The Kenyan founder preparing to flip to Delaware before her Series A may not want a beautifully localized Kenyan cap table tool. She may want a tool that handles the flip cleanly and then operates the resulting Delaware structure as if she had always been a Delaware company. Same for the Bogotá founder targeting a US Series B, or the Tbilisi founder raising from a Cayman-domiciled fund.

In this reading, the market prefers convergence because capital markets reward standardization more than they reward local fidelity. A founder who wants to be legible to global pools of capital may not want their software to celebrate their jurisdiction. They may want their software to make their jurisdiction invisible.

The original essay does not lose to this. The local-capital-into-local-companies layer is real, and so are the founders who will never want or need to interact with US-shaped capital. But the bulk of cross-border venture is probably not in that layer. It is in the layer of companies straddling two legal worlds and trying to look domesticated to the world that holds the capital. The skeptic would say the first piece reads as if localization is the obvious goal, when convergence is at least as defensible.

Counter-argument five: tooling is not the binding constraint

The original essay implies a causal chain: better tooling produces more institutional capital flowing into emerging markets. The skeptic says this is too clean.

Institutional capital tends to arrive after a set of preconditions are in place. Enforceable contracts. Courts that respect creditor and shareholder rights. Mature exit pathways, whether through acquisition or public markets. Predictable governance. Reliable accounting and audit. Property rights that hold up under political stress. Those are the things that move LP allocation models in pension funds, sovereign wealth funds, and endowments.

Software helps at the margins. It cleans up reporting, reduces administrative friction, makes the GP's job easier. It does not by itself make a market institutional. A best-in-class LP reporting tool deployed into a market with weak shareholder rights produces prettier reports about a market that institutional capital still will not enter at scale.

The skeptic's conclusion is that tooling is necessary but nowhere near sufficient, and that the original essay risks overstating the causal importance of software relative to the much slower, much harder work of institutional development.

The original essay can answer: the tooling and the institutional development happen in parallel, not in sequence, and being early to build the tooling is a way to participate in the institutional development rather than wait for it. Both are reasonable. But the skeptic has a real point about which lever actually moves the most capital.

The synthesis the skeptic would offer

If the original essay had a missing question, this is it. Should global venture infrastructure adapt to local legal systems, or are local legal systems gradually adapting to global venture norms?

The skeptic's answer is that the market is doing both in layers.

Global capital wants standardized abstractions at the top of the stack. It wants a cap table it can read, a waterfall it can model, a set of reports it can compare across funds. It wants Delaware-shaped, Cayman-shaped, or Singapore-shaped abstractions because those abstractions are the lingua franca of the institutional LP world.

Local legal implementation is what actually happens at the bottom of the stack. A company is incorporated somewhere. Employees are hired under local labor law. Taxes are paid to a local revenue authority. Disputes are resolved in local courts. That layer is irreducibly local, and no amount of tooling collapses it into Delaware.

Between those layers sit the translation devices the market has already built. The Delaware flips, the Cayman holdcos, the Mauritius structures, the Singapore parents, the SPVs in Luxembourg. These exist precisely because the top of the stack wants one thing and the bottom of the stack is something else, and the value of the translation layer is the spread between the two.

So the skeptic's punchline is: the real opportunity may not be a localized Carta for every jurisdiction. It may be a translation-first platform that handles the global-to-local mapping and back, with deep operational infrastructure underneath, and that knows when to render a company as Delaware and when to render it as its native form.

That is a different essay than the one I wrote last week. It is also a defensible essay.

Where I actually land

Two essays. Same author. Opposite-leaning conclusions. Here is the honest accounting of how I weight them.

The original essay is the one I wrote because the asymmetry it describes is real and underdiscussed. The tooling is opinionated. The opinions are in the data model, in the defaults, in the document checklists, in the math underneath the waterfall. The asymmetry is large enough that someone is going to capture meaningful value by building for it, and the first piece is my best argument for why that is true.

This essay is the one I wrote because every strong thesis deserves to be tested against the strongest version of the other side, not the weakest. The convergence argument is the strongest version of the other side. It is not a strawman. Smart people hold it sincerely, and they are not wrong to.

Where I actually land is closer to the original than this follow-up, but not all the way at the original's end of the spectrum. I think the translation-layer framing the skeptic offers is probably the more accurate description of how the winning product looks in practice, and I think the original essay was too quick to treat localization as the obvious goal when convergence is genuinely defensible. I also think the original was right that the current tooling embeds opinions it does not acknowledge, that the asymmetry creates a real business opportunity, and that capital is globalizing faster than the infrastructure.

The reader gets both pieces. The reader gets to decide. That is the whole point of writing the second one without rewriting the first.