A decade of metascience writing has taught us how to start a healthcare moonshot. It has told us comparatively little about how to finish one.

This is the first in a six-part series on how mission-driven research organizations are structured, what the first generation of metascience experiments has taught us, and what healthcare-specific organizations still need to address. The series is the shorter, faster companion to our open commentary paper, The Helix Framework. Each post stands alone.

The Year 5 cliff

Here is a pattern most people in the space have watched play out at least once: a well-funded mission-driven research organization launches, hires good people, ships ambitious scientific infrastructure, generates press, and then, around Year 5, runs into a wall.

The wall is not usually scientific. The science often works. The wall is structural. The platform the organization built is ready to leave the bench, but the organization itself was not designed to carry it any further. The nonprofit cannot raise venture capital. The time-bounded research program is ending on schedule. The researchers who built the platform are not experts in IND filings, reimbursement codes, or population-scale access. The scientists' advisors are not the people you need for FDA pre-submission meetings. The foundation's board has no members who have run a clinical trial.

Funders notice. "We funded the R-and-D. Someone else is supposed to do the translation." Journalists notice. "The technology exists. Why isn't it in hospitals?" Patients notice. "Great paper. When can I actually use this?"

The uncomfortable answer is that almost no existing institutional structure covers the full arc from "we built the platform" to "people are being treated with it." The ARPA model ends the program. The Focused Research Organization dissolves on delivery. The for-profit spinout loses mission alignment. The hybrid model gets captured by its commercial subsidiary. And so a decade of good work sits one handoff short of reaching the people it was meant to serve.

What the post-COVID reckoning actually revealed

The COVID pandemic stress-tested the global research enterprise. The headline result was that mRNA vaccines went from sequence to injection in under a year. The less-flattering result was that the speed depended almost entirely on platform technologies developed decades earlier through a handful of far-sighted bets (DARPA, BARDA, a few unusual academic programs). There was no systematic capacity to do this. It worked because of specific prior investments, not because the system was designed to respond.

That reckoning produced a wave of writing and institution-building in what came to be called "metascience." Mark Lutter argued that nonprofits should operate like startups. Matt Webb proposed labs organized around "orthogonal technologies" that no single industry would fund. Ben Reinhardt wrote the operational blueprint for adapting DARPA to the philanthropic sector. Adam Marblestone and collaborators formalized the Focused Research Organization. Erika Alden DeBenedictis wrote the practical startup guide for research nonprofits that the more conceptual pieces had skipped over.

All of this work was necessary. Almost all of it focused on the first five years.

Why healthcare is different

Healthcare imposes three constraints that most metascience writing has not fully absorbed.

First, the timeline is long and the phases are distinct. Platform R-and-D is one thing. Clinical validation is a different thing. Regulatory approval is yet another. Population-scale deployment under equity constraints is a fourth. Each has its own expertise, its own capital requirements, and its own timeline. Compressing them into "five years of research, then someone else takes over" is the failure mode.

Second, regulated deployment is not optional. A software research organization can release its model weights and declare victory. A healthcare organization releasing a diagnostic cannot. Somebody has to run the clinical trial, file the submission, design the post-market surveillance, train the clinicians, and field the inevitable adverse event reports. If that work is not inside the organizational boundary, it has to be structurally handed off to an organization whose mission is aligned enough to keep doing it under reimbursement pressure. The handoff is hard.

Third, equity of access is a first-class structural requirement, not a marketing commitment. Any healthcare moonshot that reaches only the richest 10% of its addressable population has failed in a specific, measurable sense. But equity under reimbursement and regulatory pressure is not something the commercial subsidiary will get to on its own. It has to be structurally embedded: in the bylaws, in the licensing terms, in the pricing architecture, in the emergency-activation triggers. Organizations that do not design for equity from day one do not retrofit it later.

The 15-year frame

This is why we keep arguing, and will keep arguing throughout this series, that the right planning horizon for healthcare moonshots is approximately 15 years, not five. We break those 15 years into three phases:

  • Phase I: Research and Development (Years 1-5). Non-dilutive capital; platform build; open-source by default.
  • Phase II: Clinical and Commercialization (Years 5-10). Dual model. Nonprofit continues non-dilutive mission work. A PBC subsidiary raises commercial capital for the product path.
  • Phase III: Globalization and Equitable Access (Years 10-15). Revenue maturity. Geographic expansion. Structural equity enforcement.

Most current mechanisms fund the first five years. A few cover aspects of the second five. Almost nothing is designed to own the full 15 with coherent mission governance across all of them.

What the rest of this series is about

Over the next five posts we work through the four categories of organization that have tried to close this gap, with honest attention to what each has done well and what each cannot do by design.

  • Post 2 covers ARPA-like models: DARPA's philosophy, ARPA-H, Wellcome Leap, ARIA, and the NSF Tech Labs initiative.
  • Post 3 covers Focused Research Organizations: Marblestone's concept, Convergent Research's operationalization, and four FRO case studies.
  • Post 4 covers hybrids: Mozilla, CZI/Biohub, OpenAI, FutureHouse/Edison, Arc Institute, with specific attention to which of their mission-locks are structural and which are principal-dependent.
  • Post 5 covers alternative structures: Astera, Arcadia, OpenBCI, Episteme, and what it looks like when openness moves from "value statement" to "operational discipline."
  • Post 6 synthesizes: twelve structural desiderata that we believe any healthcare-moonshot organization should ship from day one, and a brief introduction to the Helix Framework as one way to ship them.

The goal is not to declare a winner. Each of these structures is working on a piece of the problem and has produced genuine institutional innovation. The goal is to be honest about what each one does not solve, and to ask what a structure designed specifically for the full 15-year healthcare moonshot arc would have to look like.

One more framing note before we move on. Throughout this series we distinguish "capital-restricted" organizations (ones that spend years writing grants to fund the actual research) from "time-restricted" organizations (ones that get substantial capital up front and then race against the clock). The first kind is the old model. The second kind is what metascience is trying to build. The hard question for healthcare is whether a time-restricted organization can also carry itself through the other ten years, where time alone does not do the work.

That is where we are going. Next post: the organizations that, so far, have done the best job with the first five years.

Part 2: Pasteur's Quadrant and the Heilmeier Catechism