Science is humanity’s most powerful collective invention. It has lengthened lives, illuminated the universe, and given us the means to understand ourselves. Every major advance of modern civilization– clean water, electricity, vaccines, computing– flows from the patient accumulation and application of scientific knowledge.
But somewhere along the way, the machinery of science stopped matching the scale of our ambitions and the needs of society. We are surrounded by brilliant scientists and extraordinary technologies, yet the coherence of progress has fractured. We face planetary challenges that require coordination and imagination on the scale of climate, health, and energy, but our system is optimized for scientists to publish incremental papers and chase the next grant.
Today’s scientific architecture is built for an analog era of the past. We must rebuild it for a world that is digital, interconnected, and intelligent. It is time to renew the promise of science as a collective project of human progress in the new era, one that serves not just knowledge itself but the future of the societies that depend on it.
How Science Falls Apart
The dysfunction is visible everywhere. Of more than 18,000 recognized human diseases, fewer than 4,000 have an FDA-approved treatment. We can sequence a genome in hours yet cannot treat the majority of the disorders it reveals. Solar and battery technologies have improved by orders of magnitude, yet global emissions hit a record 37 billion metric tons in 2023. Global deforestation continues to ravage the environment and accelerate loss of biodiversity.
“Misfit” science, once celebrated, has become a reputational liability for most scientists. Barbara McClintock’s discovery of mobile genetic elements, the theory of plate tectonics, and the first laser were all originally dismissed as fringe or impractical. More recent breakthroughs (such as CRISPR, mRNA vaccines, and gravitational-wave detection) survived only through a combination of decades-long persistence and luck.
Science still produces new insights at an extraordinary pace, but it’s often by swimming against the tide. We’ve lost patience, a type of societal attention deficit disorder. The challenge ahead is not only to invent more and faster, but to rebuild the connective tissue between knowledge and public progress, the relationship that once made science the most trusted and transformative institution of modern life.
The Fracturing of Science and Society
The postwar scientific apparatus rested on two assumptions: that governments, industry, and philanthropies would fund science, and that scientists would set the agenda. For much of the twentieth century, that arrangement worked. It produced vaccines, satellites, and semiconductors. But over time, it also created distance. Citizens now see vast sums flow into research yet rarely understand where it goes or what it achieves. Scientists, in turn, are bound by incentives that reward consensus and publication over imagination and accountability. Large federal funding bureaucracies emerged to responsibly manage the arrangement. And we lost the entrepreneurial spirit of science and replaced it with a need to manage risk and commercialize.
The result is mutual alienation. The public sees opacity and responds with mistrust; scientists see skepticism and respond with defensiveness. Vaccine hesitancy, climate denial, the politicization of basic facts, recent attacks on research institutions and their funding are symptoms of a broken feedback loop.
Part of the fracture is that we’ve failed to innovate in how we communicate, both among scientists and with the public. The scientific paper, peer review, and annual conference were brilliant technologies for a slower century. They are now bottlenecks. Between discovery and publication lies a period of silence when scientists hesitate to share for fear of being scooped or compromising future intellectual property.
What we need are new units of communication built for the modern era, living research objects, intelligible to both humans and machines, that capture progress in real time. These should serve not just as scholarly records but as infrastructure for collaboration, replication, and AI-assisted synthesis. A communication layer built for speed and interoperability would make multidisciplinarity natural rather than heroic and would let the public see discovery as an unfolding process rather than a series of verdicts.
The great challenges of this century–climate adaptation, pandemic prevention, sustainable agriculture, AI alignment–span every discipline. Yet, our institutions are built around centuries-old fields, and conventional wisdom dictates that solving problems across them necessitates formal collaboration: consortia, memoranda, and endless coordination meetings.
True coordination at “market” scale does not depend on everyone bearing the cognitive load to work together directly, but can be achieved through an ecosystem where everyone can work within detailed information systems and a shared understanding of the field. Shared data standards, transparent progress tracking, and interoperable systems can align independent actors without forcing them into the same room. This is how markets work, and they’re remarkably efficient.
Science could move more in this direction. The goal is not to multiply inefficient collaborations but to make them massively more effective. With the right infrastructure, research can become self-coordinating: thousands of semi-autonomous efforts moving coherently because they can see where they fit within the larger system. Incentives should reward contribution to collective progress rather than ownership of narrow turf.
As Stuart Buck observes, we have built a system optimized for individual advancement rather than collective problem-solving. The next system should reverse that, using shared visibility, not central planning, to turn fragmentation into alignment.
Science has never really lacked ideas. It has lacked patient, intelligent capital that can enable iteration and constant learning.
Today, most sources of funding–government, philanthropy, and private investment–operate with high latency and low transparency. Each moves on its own timeline, guided by separate reviews and metrics, with no shared view of where discoveries are emerging or where duplication and gaps exist. The result is a fragmented, slow-moving system that cannot adapt to new evidence or redirect resources in real time.
AI can change that. It can integrate information from across funders, research institutions, and public data to build a continuously updated picture of where knowledge is advancing and where support is needed. It can identify early signals of discovery, model dependencies among fields, and recommend how to reallocate capital as opportunities shift.
This would create a new kind of financial infrastructure for science: capital that learns. Instead of isolated grants reviewed on fixed cycles, we could have a dynamic allocation system that evolves with real-time frontiers of knowledge and coordinates across funders.
The old model treats funding as administration. The next model should treat it as intelligence. When capital becomes responsive to discovery, progress accelerates and learning compounds across the entire scientific enterprise.
A generation of scientists have internalized the habits of caution. Across the U.S. and Europe, surveys show that more than 70% of researchers believe the funding system discourages high-risk ideas, and over half report spending at least a quarter of their time on grant writing and administration. The average age at which American scientists receive their first NIH R01 grant has risen from just over 36 in 1980 to above 43 today, signaling a system that increasingly backs proven track records rather than bold potential.
Administrative load, risk aversion, and short grant horizons push creative researchers toward safe, fundable problems. In one study, only 14% of NIH grant reviewers said they prioritize novelty when evaluating proposals, while nearly all emphasized feasibility and preliminary data.
The health of science should be measured by the freedom it gives its best people. Bureaucracy should protect integrity, not define imagination’s boundaries. We need institutions that trust talent, lengthen horizons, and protect unorthodox ideas. Ambition should be treated as an asset, not a liability– because the cost of caution is not safety, but stagnation.
The deeper problem is the lack of a shared system of measurement to tell us when science is underperforming or lacks ambition. We count papers, patents, and prizes as the visible numerator of success but have never defined a denominator. How many plausible hypotheses go untested? How many promising directions die in committee? How much time and talent are consumed by friction instead of discovery?
Without that denominator, we don’t really know the theoretical productivity of science– how much progress the system could make under ideal conditions. Other domains have reference points: economists estimate potential GDP; engineers calculate theoretical efficiency. Science has no equivalent benchmark, no common measure of how quickly we are improving our capacity to discover. As Patrick Collison and Tyler Cowen have argued, progress itself deserves to be studied as a scientific question– a “science of progress” that examines how and why discovery accelerates or slows.
Fundamentally the first step to achieving this is to capture the data that informs this denominator. Which projects were not funded and why; the most ambitious version of a research proposal; failed lines of research. Surprisingly, we have never done this on any reasonable scale.
This does not necessarily mean we should optimize science as if it were an industry. Science does depends on variance– on creative deviation, false starts, and the freedom to chase intuition beyond what seems practical. But the system should be self-responsive: able to sense where progress is accelerating or stagnating and to reallocate attention and capital accordingly. The goal is not for any one person or institution to see the whole playing field, but for the system itself to adapt in near real time.
A self-measuring, self-adjusting scientific ecosystem, one that learns from its own dynamics and shortens the latency between discovery and application, would allow insight to flow faster and compound more effectively. That, more than any single breakthrough, is how science will recover its true pace.
Rebuilding science is not the job of governments alone. It demands a new coalition of scientists, technologists, and investors who see science as shared civic infrastructure.
For half a century, the partnership among government, universities, and industry built the world’s most powerful innovation engine– one that turned public research into startups, industries, and national prosperity. As Steve Blank observes, when we starve public science, we don’t just slow discovery; we cut off the upstream source of innovation itself.
The task now is to renew that partnership for the networked, intelligent age we are about to enter:
Scientists should design research as reusable infrastructure, open, interoperable, and connected across fields.
Funders should build learning capital and shared data systems that make progress visible and cumulative.
Technologists and investors should create the platforms and tools that let knowledge move with the same velocity as money and code.
The tools exist. The talent exists. What’s missing is shared intent. The next century’s breakthroughs will not come from lone geniuses or siloed labs, but from rebuilding the architecture of science itself, an open, intelligent system worthy of the civilization it sustains.
This essay is part of an opening four-part series exploring how science can work better to solve our most pressing problems, through new forms of coordination, capital, and intelligence.
Part 1: Rebuilding the Architecture of Science
Part 2: The AI Program Manager
Part 3: Why Science Needs a Marketplace
Part 4: From Grants to Portfolios: Index Logic for Science