Issue 17  November 23rd 2025

The theme of this issue is Experimentation.

For years, the conversation about improving science has been building—quietly at first, and now unmistakable. Across disciplines, funders, research communities, and institutions, you can feel the same conclusion forming: the system isn’t keeping pace with the opportunities in front of it. What matters now is recognizing that there is no single blueprint for what comes next—and acting accordingly.

Experimentation is how scientific fields advance. It should also be how scientific systems advance. Yet somewhere along the way, we internalized the opposite idea: that the structures surrounding science must be stable, uniform, and slow-changing. Funding should follow fixed templates. Careers should follow predictable ladders. Institutions should look roughly the same as they did decades ago. The irony is obvious—scientific discovery depends on variation and exploration, while the system that governs it actively suppresses both.

Today, the pressure to experiment is coming from multiple directions. The data on productivity slowdowns is now hard to ignore. Replication rates remain uneven. Younger researchers struggle to find space for ambitious ideas. And on the other side, new technologies—AI, automation, computational biology, open research tools—are making it easier than ever to imagine alternative models. The system is asking the same people to do more with less, while simultaneously generating more possibilities than ever before.

This gap between potential and reality is precisely where experimentation becomes necessary. And not just one grand experiment—many experiments, across different layers of the ecosystem:

Experiments in funding models, from long-horizon grants to portfolio-based support to regranting networks. Experiments in institution design, from independent research organizations to hybrid labs to distributed teams. Experiments in career pathways, giving early-career scientists real autonomy instead of years of incremental credentialing. Experiments in translation and integration, building shared infrastructure that helps ideas move outward instead of dying in isolation.

We don’t yet know which of these interventions will work. That’s the point. No one has sufficient visibility across the whole system to predict a priori which experiments will succeed. And we should expect some to fail—just as we expect failed experiments in science itself. A scientific ecosystem that cannot tolerate institutional failure is one that cannot tolerate institutional learning.

Project Aeon was designed with this reality in mind. We aren’t claiming to know the model for the future of science—we’re running a model, one experiment among many that the moment demands. Our approach tests a set of hypotheses about where the system bottlenecks actually are and how they can be removed. Five interventions summarize our approach:

1. A different organizational model. Traditional funding separates “basic science,” “translational work,” and “startup formation” into different institutions, each with its own incentives. We’re experimenting with a hybrid structure that runs all three as a single pipeline. Early, high-variance ideas receive flexible, non-dilutive support; mature, de-risked work can receive return-seeking capital to scale. Proceeds recycle into the system, letting long arcs compound instead of dead-ending at grant cycles.

2. A different investment thesis. We treat scientific work as portfolios, not projects. Long horizons, open-ended inquiry, and the ability to pivot are built into the design, not negotiated as exceptions. The hypothesis is simple: if you give talented scientists room and time, the tail outcomes dominate.

3. A different way of selecting scientists. We evaluate the researcher—scientific rigor and capability, taste, curiosity with open‑mindedness, and pro‑social behavior matter more than pedigree. We deliberately create space for younger and unconventional scientists, and we score “overlookedness” directly—who has been systematically under-backed compared to their demonstrated potential. The goal is not to predict winners perfectly; it’s to surface variance that conventional filters suppress.

4. A different translation pathway. Instead of treating practical application as a handoff point, we embed translation as an engineering function from day one. A cross-functional systems layer—part Bell Labs, part startup studio—helps promising science become working prototypes, validated tools, or new companies when the timing is right.

5. A different kind of community. We’re creating a movement around ambitious science. Norms matter more than rules. Shared infrastructure, collaboration spaces, and community-wide incentives for helping others are all part of the experiment.

These are not claims of certainty. They are hypotheses under test—experiments about how to widen the funnel of talent, increase learning throughput, and give paradigm-shifting ideas enough space to grow. Some elements will work. Some won’t. What matters is that new institutional possibilities are being tried, measured, and refined rather than assumed into existence.

The real danger is not trying too many experiments. It’s trying too few. Scientific progress has always come from variation, iteration, and exploration. The system that supports science should, at minimum, live by the same principles. We don’t need more certainty—we need more experiments with the courage to test new assumptions, expose new truths, and help build an ecosystem capable of meeting the challenges of the century ahead.