The 1945 Fellowship · anotherroof

A friend of mine — I’ll call him D. — is a very smart person with a very stupid problem. A few months ago he finished a graduate degree at one of the best programs in the country. He is precisely the sort of person you would want building things in your city, starting companies, teaching, inventing. And a ninety-day clock is quietly counting down to the day the United States asks him to leave.

The mechanics are dull and merciless. When your student status ends, you get a short grace period to find an employer willing to sponsor a work visa, win a lottery for one of a capped number of those visas, and begin the long climb toward permanent residence — or you go home. Miss any step and the country that spent years training you shows you the door. D. is doing everything right, against a calendar, and it still might not be enough.

This is not a sad story about one person. It is a thing we do thousands of times a year, and we mostly do it by accident — not because anyone decided it was wise to train the world’s most capable people and then expel them, but because no one ever decided it wasn’t. So here is the thought experiment I can’t stop turning over.

The experiment

Imagine a fellowship. Not a scholarship, not a staffing agency — a fellowship whose entire purpose is to keep exceptional non-citizens in the country and pointed at ambitious work. It would give a fellow four things at once: room, board, and a stipend, so that a proven person can keep contributing instead of leaving; someone whose only job is to manage the visa clock, so the bureaucracy becomes the fellowship’s problem rather than the fellow’s; a circle of mentors and experts who know the work and will vouch for it; and a path into the labs and companies doing the most consequential things.

Put plainly: a machine for keeping people. The questions that follow are the obvious ones. Is it legal? Who pays for it? Does it help anyone other than D.? And — the one I kept flinching from — is this a genuinely good idea, or an elaborate way for me to help a friend and call it policy?

We have run this experiment before

Here is the part that surprised me. The idea sounds novel, even a little utopian. It isn’t. The United States ran a maximal version of it in the 1940s, and it may be the best investment the country has ever made.

The people who built the atomic bomb were, to a striking degree, refugees. Enrico Fermi fled fascist Italy. Einstein, Hans Bethe, and James Franck fled Nazi Germany. Leó Szilárd, Edward Teller, and John von Neumann came from Hungary. It was Einstein’s 1939 letter to Roosevelt that set the Manhattan Project in motion, and it was scientists who had only just reached American shores who delivered the thing that ended the war. Around the same time, the Institute for Advanced Study in Princeton became, by deliberate design, a refuge for the greatest minds in the world. And in the summer of 1945, Vannevar Bush handed the President a report that became the blueprint for half a century of American scientific supremacy. He called it Science, The Endless Frontier.

That is the working name of the thing: the 1945 Fellowship. Because the move that reads as radical today was, within living memory, simply national strategy. Your grandparents did this, and it won them the war, the space race, and the century that followed.

There’s an obvious retort, and it’s worth meeting head-on: those programs were born of catastrophe. Tyranny did the recruiting; America just had the sense to open the door. The objection says we’re at peace now, so this would be a thing we did in cold blood, on purpose, with no one being chased out of anywhere.

I don’t think we’re at peace. We’re funding and fighting proxy wars on more than one continent, locked in a technology race with an authoritarian rival that treats talent as a strategic weapon, and tearing at each other at home hard enough that “cold civil war” has stopped sounding like hyperbole. These are not placid times. They’re frightening ones. Which makes the 1940s a closer analogy than the comfortable version of this objection allows — not because we should wait for a catastrophe to act wisely, but because the catastrophe-shaped pressure is already here. The only real choice is whether we do on purpose what we once did out of necessity.

The part I find genuinely clever

The non-obvious mechanism is this. American immigration law already contains lanes built for exactly these people: the O-1 visa and the EB-1A green card, both for “extraordinary ability,” and the EB-2 National Interest Waiver, for work in the national interest. The lanes exist. The problem is that they demand an evidentiary record — awards, recognition, letters from established experts, evidence of a public platform — that an individual sprinting against a ninety-day clock cannot realistically assemble alone.

A fellowship, almost as a byproduct, manufactures that record. Being selected is an award. Your mentors are your expert letters. The platform is your platform. So the fellowship doesn’t merely buy time against the clock; it builds the case for staying permanently. It isn’t a loophole or a workaround. It’s using the machine exactly the way the machine was designed to be used, for exactly the people it was designed for.

It doesn’t just help someone stay. It assembles the evidence that they deserve to.

But does it help America, or just D.?

This is the question that actually matters, and it’s the one most likely to be answered with a bumper sticker in either direction. So here are some numbers, and then an honest caveat about what they do and don’t prove.

Figure	What it measures
59%	of America’s billion-dollar startups have an immigrant founder — 455 companies worth roughly $5 trillion, averaging about 833 U.S. jobs each.
~25%	of U.S. patents come from immigrant inventors, who are about 16% of inventors — they patent at well above their share.
7–8pp	faster wage growth for native college graduates from a one-point rise in the local foreign STEM share. They tend to lift wages, not depress them.
most	of the top-tier AI researchers working in the U.S. did their training abroad — the exact workforce that decides who leads the field.

Sources: NFAP (2026); Cato / NBER; MacroPolo. Full links below.

The data points, consistently, in one direction. But I don’t want to oversell it, because being honest about this is the whole reason it’s filed under thought experiments and not pitches. Those figures describe immigrants broadly; a boutique fellowship of a few dozen people is never going to move a national statistic. The real argument is narrower and, I think, stronger: the per-person expected value of the exact individuals this thing would target — the top fraction of a percent, the people who already qualify for “extraordinary ability” — is extraordinarily, almost absurdly high. You are not betting on immigration in the aggregate. You are betting on specific people who have already cleared a very high bar.

The objections, taken seriously

A thought experiment earns its keep by arguing against itself. These are the strongest versions of the objections I’ve heard, including the one people usually aren’t polite enough to say out loud.

“We can’t even take care of our own people. Why are we spending anything on this?”

The weak version of this is xenophobia. The strong version is a reasonable zero-sum intuition: resources are finite, and attention spent here is attention not spent on citizens who are struggling. I take that seriously. My honest answer is in two parts. First, high-skilled immigration is close to the clearest case economists have of a policy that isn’t zero-sum — it creates jobs and raises wages rather than dividing a fixed pie. Second, the fair response to “why not help citizens instead” is “do both,” and the hybrid funding model below is specifically designed so this doesn’t compete for the same charitable dollars that feed people.

“This is a scam for cheap labor dressed up in patriotism — another Valley play to import workers on the cheap.”

The steelman here is real: high-volume visa programs genuinely have been used to suppress wages in some corners of the tech economy. But O-1 and EB-1 are, by law, the opposite of that. They are for people at the top of a field, who command top wages, and the entire goal here is a green card — permanence. A permanent resident answers to no single employer and can’t be held hostage by one. If anything, this is the antidote to the body-shop model, not an instance of it.

“Those WWII programs were different. That was wartime. You can’t compare this to beating Hitler.”

I used to think this was the strongest objection. I’ve stopped, because its premise is that we’re in peacetime and they weren’t — and I don’t believe we are. We’re bankrolling proxy wars on multiple fronts, racing an authoritarian rival for control of the technologies that decide the next century, and divided enough at home that it doesn’t feel stable or safe. If talent decided the last existential contest, it will decide this one. And even setting the present moment aside: the keeping was never only wartime. Bush’s report built the National Science Foundation and the modern research university, which kept America ahead for fifty years after the shooting stopped. We didn’t stay great because there was a war. We stayed great because we gathered the best people and kept them — in war and out of it.

“Forget all of it. Send them back where they came from.”

I’m not going to pretend this sentiment doesn’t exist, and a statistic is the wrong reply to it. The honest reply is that this position, taken to its conclusion, means choosing to export your own advantages to your competitors. The brilliant person you expel does not stop being brilliant. They go and build somewhere else — often somewhere that would very much like to beat you. You don’t have to feel warmly about immigration to notice that a genius on your side is better than the same genius on someone else’s. Keeping them isn’t generosity. It’s self-interest.

Where the idea actually breaks

I’d be doing the thing I dislike in other people’s thought experiments if I stopped at the objections I can answer. Here are the ones I can’t, or can’t yet. These are the reasons it might not work.

The legal knot. You cannot simply hand someone room, board, and a stipend if they aren’t authorized to work — “support” quietly becomes unauthorized employment, which is the one mistake that sinks a case. The whole thing only functions if the fellowship itself qualifies as a visa sponsor and times everything to the grace window. That’s a hard, lawyer-shaped constraint at the center of the design, not a footnote.
Who counts as “exceptional.” “High impact” is easy to say and very hard to measure without just reproducing the pedigree bias that already decides who gets an O-1. A fellowship for the exceptional risks quietly becoming a fellowship for the already-anointed — the people who went to the right schools and knew to collect the right letters.
It helps a rounding error. The model is expensive per person. It helps dozens, maybe hundreds — against tens of thousands on the cliff every year. Is a boutique concierge program even the right shape? Or should the same effort go into fixing the lanes for everyone, which is roughly what groups like the Talent Mobility Fund are already attempting?
The framing is a trap. Wrap it in flags and you lose half the room. Wrap it in humanitarianism and you lose the other half. The politics of high-skilled immigration are such that the same idea reads as nativist or globalist depending entirely on who’s describing it. I’m not sure that’s solvable in the current climate.
It might not need to exist. A competent immigration lawyer and some money already do most of this for one person. The entire bet is that productizing it — the brand, the cohort, the network, the proactive management of the clock — is worth more than the sum of those parts. Maybe it is. That’s the actual hypothesis, and I hold it loosely.

The back of the envelope

I did build a small model, mostly to see whether the numbers were insane — you’ll find it among the artifacts below, with every input left for you to change. A pilot of ten fellows runs somewhere around $155,000 fully loaded per fellow in the first year — stipend, housing, legal, mentorship, a share of a lean core team. Call it roughly $3.7 million to run a first cohort for a couple of years. A hybrid structure helps: an earned-revenue arm that sells visa concierge work to companies trying to retain their own talent, plus placement fees, could plausibly cover something like half the cost by year three, with philanthropy carrying the rest.

The figure that actually made me stop was downstream. If even a fraction of a cohort goes on to found something — and the base rates for this population are not low — the implied cost per American job eventually created lands in the low thousands of dollars. I don’t fully believe my own spreadsheet; the founding-rate assumption is doing an enormous amount of work, and I’ve labeled it as such. But even after discounting it hard, the leverage is strange enough to be worth staring at.

The artifacts

I didn’t only think about this — I built the case out, probably further than a thought experiment strictly warrants. If you’d rather argue with the strong form of the idea than this hedged one, here’s the actual work. Fair warning: these were written to persuade, and they’re more certain of themselves than I am.

The one-page case (PDF) — the whole idea laid out: the 1945 precedent, the four things a fellow gets, the evidence flywheel, and the hybrid model, written the way you’d put it in front of someone who could fund it.
Hard questions, straight answers (PDF) — the objections above, including “that was wartime,” answered with the actual numbers in the blunt register you’d want if someone were trying to kill the idea in a room.
The back-of-the-envelope, live (XLSX) — the pilot economics with every assumption left editable; change the blue cells and watch the founding raise and the cost-per-job move.

Back to D.

The thing I keep returning to isn’t the spreadsheet or the history. It’s a sentence I believe more the longer I look at it: “American” was never really a description of where you were born. It’s a description of what you’re willing to build, and who you’re willing to build it alongside. For a stretch of the twentieth century we acted as if we believed that — not out of sentiment, but because it made us formidable. Then we stopped, without ever quite deciding to.

D. wants in. He wants to build here, under the same flag, pointed at the same future. The current answer is a ninety-day clock.

So: is the 1945 Fellowship a good idea, or just an elaborate way to help one friend and dignify it with a name? I honestly don’t know. That’s why it lives here, under thought experiments, and not somewhere more confident. If you can find the hole I can’t — the legal flaw, the economic sleight of hand, the reason this is naive — I genuinely want to hear it. The ways to reach me are on the about page.

Notes & sources

National Foundation for American Policy, “Immigrants and U.S. Billion-Dollar Companies” (2026) — founder share, company value, jobs per company.
Cato Institute / NBER research on high-skilled immigration, patents, and wages.
MacroPolo, The Global AI Talent Tracker — where top AI researchers come from and work.
Atomic Heritage Foundation / National Museum of Nuclear Science, “Scientist Refugees and the Manhattan Project.”
Institute for Advanced Study, “A Refuge for Scholars.”
Vannevar Bush, Science, The Endless Frontier (1945).

This is a thought experiment, not a prospectus. The numbers are illustrative, the design is unbuilt, and “D.” is a stand-in. I’m writing it down to find out whether it survives contact with people smarter than me.