The 100 + page document is here. I’m still reading the report and will eventually react to it here on the blog (and not on my Substack). There are plenty of quick takes (no doubt aided by AI) out there.
July 23, 2026 After 24 hours to digest….
Every science policy document since 1945 has to decide what to do with Vannevar Bush. This one, Michael Kratsios’s new report to the President, pays Bush constant tribute while quietly taking aim at the institutions his 1945 report built. Strip away the “endless frontier” homage and what’s left is a fairly direct hit on Research 1 universities and academic medical centers — the two pillars of the postwar system.
The report doesn’t say this outright. It doesn’t have to. The R01, the workhorse grant of American biomedical research, is described as a source of “systematic blind spots.” NSF is dinged for organizing itself “around academic disciplines, much as it did in the 1950s,” as though disciplinary depth were the disease rather than one input among several. And indirect cost recovery — the mechanism that keeps AMCs and R1 physical plants standing — gets reduced almost entirely to administrative bloat, with barely a nod to the clinical and research infrastructure it actually funds. That’s not a subtle target. Indirect costs are the financial bloodstream of every academic medical center in this country, and this report treats them as a tax to be cut rather than infrastructure to be understood.
None of this is argued against directly. Instead, the report builds a parallel architecture — X-Labs, FROs, ARPA-style program managers, golden tickets — and lets that architecture do the work of making the R1/AMC model look slow, captured, and unaccountable by comparison. It’s redirection of gravity, not a demolition charge. But gravity redirected long enough is still demolition.
I’m not reflexively defensive of the status quo here. I’ve run a research institute, sat on both sides of the review table for decades, and I know the diagnosis is largely right — the twenty-month grant timelines, the graying PI pool, the incentive to armor good science in whatever language currently survives review. That’s real.
What bothers me is the sleight of hand in the prescription. The report wants to restore “merit” against the politicization of peer review, while simultaneously proposing to reorganize the whole federal portfolio around a different set of political priorities — national security, reshoring, great-power competition. It never explains why one externally imposed priority corrupts science and the other doesn’t. That’s not an oversight. It’s the whole argument, left conveniently unresolved.
The one piece I’d take without hesitation is the golden ticket: let a single reviewer stake real conviction on an unconventional proposal, blind the review to institutional pedigree, and you might actually rescue the ideas that consensus panels are built to kill. I’ve watched strange, brilliant proposals die by committee for exactly this reason. Of everything in this report, that’s the one line item I’d install tomorrow.
So: a genuinely interesting vision of AI-catalyzed, mechanism-diverse science funding, built by quietly hollowing out the R1 and AMC model that has carried American research since Bush. Whether that’s a feature or a bug probably depends on which side of the review table you’ve spent your career sitting on.