The Shape of Science to Come

Imagining the days after tomorrow.

It was a tumultuous year for American science.

Throughout 2025, the headlines in Science and Nature read like a slow-drip obituary. Dreadful developments accumulated, tracked by full-time beat reporters and dedicated “bad times for science” sections on their websites. Each day brought a new layer to the unfolding drama.

I’m concerned, but not surprised.

Years ago, I wrote that a financial shake-up was coming to academia. My prediction was milder than actual events, but pointed in the right direction. The heart of the issue was a long-overdue contractual renegotiation. I used Hollywood’s transformation in the 1950s as an analogy for how science might evolve. But the events of the year revealed two mistakes: I started the story in the wrong place—The Endless Frontier, like nearly everyone else— and I fixated on the wrong outcome.

What follows is an attempt to mend those original mistakes and articulate my current theory of change. A longer view helped surface an obvious point: science makes meaningful advances not just by federal policy and funding, but by widening the circle of participation. That circle is defined by the cooperative interface—a concept I’ll explore below.

The year also reinforced the aspects of my perspective that haven’t changed. I still believe that when institutions struggle to adapt, outsiders often can. The parallel social infrastructure they create can become a lifeline to the future.

This essay is mostly a reminder—to myself, and maybe to you—that the situation requires more than concern. The moment demands a sort of imaginative resilience: the ability to acknowledge real damage while still insisting that better futures can be built.

---

Post-Pandemic Science

The COVID-19 pandemic affected science in ways we’re only beginning to understand.

Early in the pandemic, I sensed something important was happening in science. It wasn’t any specific discovery or tool I was watching, but rather a larger, subtler shift. Many of the scientists I found most interesting were mapping their careers onto a new archetype—a fuzzy outline at best, but certainly a different way of being a scientist in the world.

I started documenting conversations I was having with scientists and science-adjacent entrepreneurs to understand this new terrain. I called the effort “Science Better”. The content struck a chord with a small but committed audience of (mostly) scientists. The project was loosely tied to the philanthropic science funding experiments we were running on Experiment, but the conversation broadened. I found like-minded people who were charting a similar path through the uncharted territory between science, tech startups, and philanthropy.

I wasn’t alone, of course. Others were writing and documenting, too—a scene was emerging. Sam Arbesman created the Overedge Catalog as emerging and alternative research organizations began to crop up, which gave visibility and a loose sense of connectedness to the efforts. Adam Marblestone and Sam Rodriques published a white paper articulating their vision for more Focused Research Organizations. Long-time open science advocates, realizing that another generation of reformers was revving up to improve the institutions of research and discovery, added their informed perspective. The whole discussion got tagged under the emerging genre of metascience.

Soon, books appeared, think tanks took up the issue, and international conferences convened around the topic. Notably, Michael Nielsen and Kanjun Qiu wrote the pièce de résistance of the emerging scene: A Vision of Metascience. They made a powerful case that science could evolve—should evolve—through a process of imaginative design and evaluation. Improving science, they argued, would require the exact form of experimentation that scientists espouse, and would likely involve more than any one specific fix:

Rather, the point is that a flourishing ecosystem would rapidly generate and seriously trial an enormous profusion of ideas… The best of those ideas would be rigorously tested, iterated on, debugged, and scaled out to improve the entire discovery ecosystem.

Beyond the thinking and writing, things were actually happening. COVID threw science and medical research into the center of the civic arena. The use of preprints for quickly sharing research outcomes skyrocketed in both use and legitimacy. And initiatives like Fast Grants and Operation Warp Speed showed that science could be done at a different pace and with novel organizational arrangements. The rigid boundaries of science’s ivory tower seemed malleable and movable. Science, the grand civilizational pursuit, was evolving.

Thanks to the booming metascience corpus, I felt like I had gotten the picture—a good-enough mental model of what would happen next. I’d seen the shape of what science was becoming, or so I thought.

I wrote up that vision in an essay called “The Hollywood Analogy”. It seemed to me that science in 2020 was in a similar predicament to the filmmaking industry in the 1950s. The old order—the impossibly powerful studio system, which controlled all the financing, production, and distribution of films at the time—was suddenly shaken and weakened, owing to the rise of television and a damaging antitrust lawsuit. It opened the door for the next generation to forge ahead. The newly established talent agents renegotiated the deals for their biggest stars, and the balance of power tipped dramatically from studio executives to talent. The industry evolved quickly and irrevocably.

The Hollywood transition was uncomfortable for anyone with a financial interest in the old way of doing business, but a liberating opportunity for the writers, actors, and agents willing to bet on themselves. A New Hollywood emerged, and the movies got much better as a result.

From my vantage point, the same general trend was happening in science. The best scientists had options. They were no longer constrained by the academic career path of old. They were starting companies, securing grants, and building labs on their own terms, rather than those dictated to them by their universities.

I stand by that general prediction. But as the actual events play out, I’ve realized some important aspects I got wrong. My first error was when I started the story.

Like everyone who writes about the American research enterprise, I began the story with Vannevar Bush’s Science: The Endless Frontier as the Big Bang of Big Science—the birthdate of the modern research system. There’s a reason everyone starts here. It is the beginning of the current arrangement, as we’ve come to know it. The major components were cemented into place through Bush’s recommendation to President Roosevelt about a post-war blueprint for American research. Bush laid out how the wartime machine could be pointed at noble causes like curing disease and inventing futuristic technologies, and it happened just so.

But I should have gone back further. There was as much to learn in the decades before WWII—through the failed attempts to launch a national research program—as from those critical years after.

The Endless Frontier was more than a beginning. It was also an ending.

---

Interwar Science

Just like COVID-19, World War I was a shock to the scientific system in the United States.

Prior to the war, scientists enjoyed a precarious social position. The prestige of the profession was steadily increasing with awards like the Nobel Prize, established in 1901, and new relationships with captains of industry. However, the general popularity of science was dwindling after the turn of the century. The great age of science popularization, which had followed the boisterous debates around the theory of evolution, was fading away and people were losing interest. It wasn’t a purposeful retreat but rather a result of advancing frontiers. As science professionalized and specialized, it became harder to maintain an ongoing conversation with other scientists in distant fields, let alone a general, more public dialogue. Science was firmly decentralized and heading further in that direction.

The Great War changed all that. The historian, Ronald Tobey, documented the period in his book, The American Ideology of National Science, 1919-1930:

“During the First World War, nongovernmental scientists in universities and research institutions were recruited to work on defense problems. Their scientific activity in the war was distinctly different from their earlier work. Before the war, they had done research on problems whose solutions were of interest mainly to men of their own specialties. These researches had been individual enterprises in which they had worked without supervision. Their professional activities had been conducted on the local or regional levels except for the annual or semi-annual national conference in their fields. In contrast, during the war many men left their homes for research centers like the New London Experimental Station or Washington D.C. Their research was a team effort, supervised and coordinated with that of other teams by a central agency. And these scientists had the deep satisfaction of knowing that they contributed directly to America’s survival.”

Scientists had tasted Big Science—larger budgets, bolder projects, and deeper coordination—and they wanted more. When the war concluded, discussions immediately continued about how to maintain the momentum of cooperation as well as the generous financial appropriations. According to Tobey, the situation “impelled the scientists to find substitutes if they wished the accelerated scientific progress of the war to be continued in the peace.”

The National Research Council (NRC), which was created as the National Academy of Sciences’ coordinating apparatus for wartime research, was poised to lead the effort. But debates about exactly how a national strategy should proceed kept scientists tied up for years. The scope of disagreement: how to engage industry, how centralized the planning and coordination should be, and how much should be spent on federal laboratories as opposed to building capacity at the universities. The NRC argued and lobbied for dramatic forms of centralization. Everyone agreed on a bigger pie, but couldn’t agree on how to slice it.

Beyond the mechanics, there were philosophical debates about what science should be aimed at—about who science was for. Was it a tool of democracy? An extension of American individualism? Was science an inevitable form of cosmic progress?

It wasn’t until the early 1930s, when the National Academy of Science attempted to fundraise for a National Research Fund, that the failure to unify behind a common vision came to a head. Tobey again:

“The failure of the fund was ideological. The original goal of their campaign was to cultivate public awareness of the values of the pure scientists. This goal was part of the general effort of the national scientists to convince the public of their ideology and thereby to promote cultural unity. The scientists thought that public recognition would bring financial support from the corporations. But the scientists never were able to establish a connection between the public acceptance of their ideology and the corporate donations. Consequently, as the campaign for funds progressed from 1926 to 1929, the scientists’ attention shifted away from constructing a primary relationship between themselves and the public to constructing such a relationship with the corporations. By 1929, their central concern was no longer to relate the method of pure science to the progressive values of individualism and democracy, but to demonstrate to industrialists that pure science, rather than engineering or applied science, was the basis for industrial profits.”

This is the essential bargain of American science. Scientists, in order to justify their budgets and fix their social position, realized they must align with both the military and corporate goals of the nation. Big Science—despite the lofty ideals of the individual participants—could only be achieved through national science. This is the concrete foundation laid beneath the ivory tower.

The failure of the National Research Fund was both lesson and prelude. As soon as “World War II began, the new generation of scientists who had been doing graduate study in the 1920s, would undergo an experience similar to that of the previous generation.”

The Second World War was another chance for scientists to lock in their aspirational social arrangement.

Vannevar Bush got the call to lead the National Defense Research Committee (NDRC), which then became part of the larger Office of Scientific Research and Development. The committees and discussions that stalled out the National Research Fund gave way to a streamlined organizational structure and the decision-making centered around Bush. He employed a famously flat reporting structure, with division leads across different fields, like the Radio Research Lab at Harvard and the Radiation Laboratory at MIT. They funded quickly and the results of their investments enabled important developments like radar and the atomic bomb.

Bush’s lasting impact on American science was not just what he funded, but also how. The NDRC set an important precedent when, as Bush described in his memoir Pieces of the Action, they “decided that we would make contracts for research directly with universities, not individuals therein.”

The decision was important for the war effort and beyond. They were underwriting the cost of research by paying overhead rates to the universities that maintained the infrastructure. Bush credited the post-war outcome as “literally a lifesaver for the universities.”

This simple fact—a contracting decision made in the fog of war, mostly without consultation—has accounted for trillions in research funding. It’s no wonder the current debate and gnashing over science budgets has centered on this contractual detail—the indirect research costs—and the administration has focused its pain-inflicting there.

But that’s not where the solution lies. Mending the contractual details isn’t enough. The situation requires a deeper fix. It’s time to examine and fix the foundation. We need to update the essential bargain between science and society, which is a decidedly larger conversation.

We’re in the same situation as the interwar period, or should be, at least; new ideas, big dreams, and fierce discussion. Ours is a time for Popper-esque philosophizing, not just Bush-like dealmaking (push that off, if possible, until we learn a few things).

Echoing Nielsen and Qiu’s call to action: now is the moment for bold imagination and experimentation.

---

Improving the Scientific Project

The second big problem with my Hollywood analogy: I focused on the wrong outcome. I was following the money, which is a common refrain in the metascience scene.

I wrote about the stars of science and how they were pioneering a new type of career by diversifying their identity outside of academia, and moving into industry and private research organizations. Founder culture, made famous in startup land, was bleeding into science. The repercussions, I thought, would produce similar results to Hollywood: scientists would get wise to their true value, universities would evolve quickly or get left behind, and the new arrangement would settle into an updated, talent-centered equilibrium.

Although this prediction has proven accurate, it was limited, plucking only the most obvious of conclusions. Of course, the star scientists are going to be fine and full of good options for continuing their work. Duh.

But it wasn’t the stars that needed attention; it was everyone else.

The product of a large, national research budget was always more than published papers, headline-grabbing breakthroughs, and Nobel Prizes. What’s hidden in those indirect costs being paid to universities are all the tools, laboratories, and people—the scientific proletariat—that make discovery possible. Slashed budgets mean gutted infrastructure, and especially the people.

The headlines tell the story: the NSF and NIH spigot has seized up. The agencies froze new awards and shrunk paylines. The flagship Graduate Research Fellowship pool was reduced to the lowest intake in 15 years. Early-career scientists are the collateral damage. The traditional “apprentice–tenure” pathway is collapsing, and might already be damaged beyond repair.

At this point, the biggest hit to science is invisible. The lost grants are bad, but the missing generation is likely going to be worse—an acute pain that will be noticed in the years and decades to come. In a testimony to the House Committee on Science, Space, and Technology, Dr. Margaret Leinen, the director of the Scripps School of Oceanography, said that the leading oceanographic institutions had to make pre-emptive cuts in admissions, accepting half of the usual amount of graduate students in preparation for reduced budgets.

Some scientists will jump to industry, while others might find homes at the new science organizations or research institutions outside the United States, but an even larger number of young people won’t enter the system. They’ll miss those critical years watching and learning in the laboratory, in the field, or at the bench.

The new science startups are trying to fill the gaps by hosting fellows or creating resident PhD programs, but these one-off solutions won’t fill a gaping generational hole. We need New Deal-sized ideas about how to engage more people in the scientific project.

Another problem with focusing on the money is that the perspective ignores the core truth of the scientific project: an international community of striving researchers who freely share their best ideas and theories in the hopes of contributing to the storehouse of human knowledge, all for the glory and credit amongst their peers, current and future. The funding of science is downstream from this delicate miracle of global cooperation.

If your story starts with The Endless Frontier, you can make the mistake of thinking the federal appropriations created this community. That’s wrong, of course. The Bush decision helped underwrite the momentum, but it didn’t create the core infrastructure, let alone the core ideas. Just like Google Adwords didn’t create Google Search—it was simply the business model wrapped around an already great product.

The science historian Lorraine Daston tells the against-all-odds history of scientific cooperation in her book Rivals: How Scientists Learned to Cooperate. Starting with the early Republic of Letters and continuing through the 1960s, when the term “scientific community” came to resemble the international collaboration we recognize today, Daston’s history lesson makes clear that creating this ever-growing circle of scientists required continual leaps of imagination. Early global projects like the International Cloud Atlas and the World Meteorological Organization created the observation networks and communities, which showed the path others could follow.

Science gets bigger and better when the circle of participation gets wider. And the big jumps in cooperation are not rational or logical. Daston writes:

“But as historians have shown and political theorists have acknowledged, no viable collective is ever just the result of cool cost-benefit calculations. It is a shared vision of what it would mean to be part of a collective that surmounts hesitation and commands allegiance. What that collective vision should be for science already has a 350-odd-year history, and the work of the imagination is still ongoing.”

The problem with many of these metascience schemes and manifestos (which pop up every few months, and I’m guilty of, as well) is that they obsess about some new economic angle. We need to fix the incentives, find more industry funders, create a marketplace, etc.

But true scientific evolution happens when the tools for cooperation change—when the social circle widens to include more minds. The scientific journal, for example, was a new cooperative interface that expanded the concept of who could participate in the scientific project by moving ideas in smaller chunks across time and space, opening them to both critique and contribution beyond the published books and meetings of the learned societies.¹ The open science crusaders (bless them) have been closest so far, but the cooperative interface (the scientific paper) was the same. Preprints—while very useful—are just a paper, sooner.

AI has a real chance to become a new cooperative interface—a machine-mediated guide to the entire written corpus of science. This is no secret. Almost every scientist is busy working AI into their process and workflow, and seemingly every major AI company has zeroed in on science as their next focused effort. AI for Science startups are busy connecting LLMs to lab equipment in the hopes of creating “self-driving labs”. Even the administration, with its newly announced Genesis Mission, is getting in on the action.

Whether AI becomes a truly autonomous discovery agent—running its own experiments, discovering new laws of physics, etc.—remains to be seen, but I think it’s worth taking the “AI as new cooperative interface” idea seriously as its own unique outcome. What will it mean to contribute to science in this form? (Large, unique datasets seem far more important, to point out one obvious thing.) And what will it mean to consume science in this form? If anyone can ask an AI science engine a question—with full command of the literature, evolving data, and experimental infrastructure behind it—what use are professionals for explanation and interpretation? The line demarcating the bounds of the scientific community is about to get a lot blurrier. What happens to the standards, ethics, and status mechanisms that the community has established over the past century? Will they translate to this expanded circle of participants or will new ones emerge?

The interface matters. For decades, the walls of the ivory tower of academia have been shrinking around the scientific community, with anti-science sentiment swelling along with a general revolt against experts and institutions. The phrase “War on Science” is bandied about, referring to everything from the politicization of science to the influence of moneyed interests. The data behind the distrust isn’t as dramatic as the stories would make it seem. Regardless, scientists certainly seem to feel threatened by their drifting societal role, and miffed about how to fix it.²

Martin Rees, former President of the Royal Society and Emeritus Professor of Cosmology and Astrophysics at the University of Cambridge, spent a large portion of his career in the upper echelons of science, thinking and writing about the existential risks to humanity. His most recent book, If Science is to Save Us, describes our cultural conundrum: science is the common denominator to all of the major global risks. Safely navigating the century ahead, he argues—through the challenges and opportunities of climate change, artificial intelligence, and biotechnology, among many others—will require an expanded role for the scientific project. Science is the most useful tool in the civilizational toolbox and, going forward, it will need more participants, advocates, and friends.

The book was published in 2022, in exactly the moment of the COVID-19 pandemic response when science seemed most triumphant. The mRNA vaccines had brought the world back to some semblance of normalcy, and Rees was championing the role of science to address more of our global issues. From his perch, he couldn’t see the swelling anti-vax sentiment or imagine that, just two years later, federal funding for mRNA research would be halted. Rees correctly identified the science-culture fault line, but underestimated the magnitude and nature of the divide.

Rees’ perspective is representative of science as a whole, which seems unable to act on this cultural blind spot.

The shape of science has always had a mirror image: the society within which it operates. And you can’t change one without the other.

---

Parallel Social Infrastructure

Remaking the scientific enterprise will require new cooperative interfaces, which are hard to prototype within strained institutions.

Rather than rebuilding in place, the more reliable approach is to build next to them: salvage what’s useful from the collapse and put it to good reuse. There’s a relevant historical lesson here from the Soviet Union.

Mathematics in the postwar Soviet Union—roughly the 1950s through the 1980s—is a paradox of progress. In many ways, it was a productive period, full of achievement, producing numerous Fields Medalists and breakthroughs in the fields of topology and group theory. The period is often referred to as the “Golden Age” of Soviet Mathematics.

It’s a most unlikely outcome, given the challenges facing Soviet scientists at the time. The editor Ross Andersen used exactly this period in the history of Soviet science as the throughline story and analogy for his Atlantic story on the current situation in American science, “Every Scientific Empire Comes to an End”, which detailed how corruption and cronyism ruined a vibrant ecosystem of ideas and scientific rigor.

The Soviets sabotaged their own success in biomedicine. In the 1920s, the U.S.S.R. had one of the most advanced genetics programs in the world, but that was before Stalin empowered Trofim Lysenko, a political appointee who didn’t believe in Mendelian inheritance. Lysenko would eventually purge thousands of apostate biologists from their jobs, and ban the study of genetics outright. Some of the scientists were tossed into the Gulag; others starved or faced firing squads. As a consequence of all this, the Soviets played no role in the discovery of DNA’s double-helix structure.

It was a textbook fumble. And Soviet mathematics should have been just as hampered—under the same political pressures, after all—but it wasn’t.

The Soviet leadership had placed tight constraints on all the formal and official research environments, like academic institutions and government positions. First, the authorities forced isolation. Soviet scientists couldn’t travel, which meant they were cut off from the international mathematics community. Ideas were also confined; books and journals became scarce, and translations were limited. Faculty was filtered for political and religious purity, and the remaining university administrations were left with strict instructions about curriculum, which failed to adapt and evolve as new fields emerged. To top it off, the research institutions had guards who checked the students’ IDs and limited access—literal fences in addition to the ideological barriers to knowledge. Slava Gerovitch, the MIT historian of Russian sciences, described the situation in a paper on the period, Parallel Worlds: Formal Structures and Informal Mechanisms of Postwar Soviet Mathematics:

All these factors worked against the creation of a fully functional mathematical community. In other words, the conditions in which Soviet mathematics developed in the 1950s-80s looked like a recipe for disaster, not for a “golden age.”

However, despite the restrictions, curiosity found its way through the cracks. Enterprising scientists took it upon themselves to create alternative forms of academic scholarship and havens for intellectual ferocity, which Gerovitch referred to as a “parallel social infrastructure, which existed apart from and in some sense in opposition to the official institutions.”

The scientists went around the obstacles, rather than through them. This parallel social infrastructure wasn’t just one solution, and hardly, if ever, official—it was all makeshift.

The most visible parts of this parallel world were the open seminars, with the “most famous and influential” of those being led by Israel Gelfand, which he ran from his post at Moscow University for more than four decades. Gelfand had done work for the Soviet military, so he enjoyed a modest amount of immunity from administrative backlash, which he flaunted by creating a seminar series to explore emerging and frontier ideas in mathematics.

“These seminars covered a wide range of topics beyond the rigid Mekhmat curriculum. The system of open seminars, which gave instruction in the most recent, booming fields of mathematics, became a key component of the parallel social infrastructure. Since these seminars were offered outside the regular curriculum, attending them did not bring students any credit. In fact, many participants were not university students at all but came from outside the university, figuratively or literally climbing the fence.”

The seminars were only part of the picture. At almost every level—young and old, amateur and professional—a generation of mathematicians took up the cause. They acted for each other, in defense of intellectual freedom and resistance to administrative influence. A network of math circles, or kruzhoks, emerged where students would teach each other concepts and curriculum. They created their own schools, too. Prominent mathematicians parlayed the influence they developed by participating in the nuclear program into lobbying for the creation of specialized schools with rigid adherence to meritocracy and technical ability.

I originally heard about these seminars from a second-generation mathematician, now at Princeton, whose father had grown up in the Soviet Union and attended these seminars. Even a generation removed, with the stories secondhand, I could hear the excitement and purpose that were present in those evening sessions. Mathematics had ceased to be a profession and had instead become a cultural movement, fueled by enthusiasts.

This idea—parallel social infrastructure—is relevant now. It proves that federal budgets aren’t destiny, and ivory tower decrees are not the only voice of response.

The shape of science to come is open for interpretation—and for invention. It could be rebuilt on a currency of enthusiasm, radical participation, and merit. It could break down cultural barriers and rise to meet the issues of our time. What it cannot do is retreat to the arrangements of the last century. There’s no back button here.

If we can act with enough imaginative resilience, the scientific project will continue, remade soundly for the century ahead.

1. I thought this was a novel coinage—it wasn’t. After I wrote down “cooperative interface”, I searched to see if anyone else had used it before and for what purpose. Julian Jonker wrote a paper, “Automation, Alignment, and the Cooperative Interface” in 2023, which presents AI as a new cooperative interface for the workplace, which he defines as a cooperative institution. His definition works perfectly, and he’s owed the citation:

“More abstractly, the cooperative interface is the infrastructure that determines the opportunities for cooperation, the affordances that allow individuals to engage in cooperation, and the styles of cooperation that are salient.”

2. At every scientific conference I’ve attended over the past few years, I’ve heard some version of the same refrain: “We have to get the general public to care about science.”

Every time I hear it, I shake my head. Framing the issue as science vs. “the public” as a distinct and separate group of people is certain to further the divide.

A simple fix: “public” should only be used as an adjective, and never as a noun.