It's not every day you get a 100-page-plus document that summarizes the Administration's thinking on an area like science and technology policy. But this month, we've got one.
See the source document PDF here. From the OSTP - Office of Science and Technology Policy. Which is headed by Michael Kratsios (formerly with investor Peter Thiel).
See a news article at WSJ by Palantir's Mike Gallagher - here.
###
See a new white paper by Chat GPT 5.6 that assesses the OSTP report. This is an AI-generated white paper, so it can be a benchmark for AI abilities to read, assess, and write-up. It should not be taken as a truth standard.
Find the 11-page white paper >> here.<<
The White House Office of Science and Technology Policy’s Science: A New Golden Age is an unusually ambitious attempt to redesign the American scientific enterprise. Written by OSTP Director Michael Kratsios, whose background spans Peter Thiel, technology policy, defense R&D, and Scale AI, the report reflects a distinctly Silicon Valley and national-security view of science.
Its central argument is that the postwar research system has become too bureaucratic, institutionally entrenched, slow, and dependent on consensus-driven peer review. Kratsios proposes funding individual scientists more directly, experimenting with long-term grants, fast grants, prizes, “golden tickets,” ARPA-like organizations, and new mission-driven research institutions. NIH and biomedical research are particular targets for reform, while foundational biology, genomics, synthetic biology, neuroscience, and biomanufacturing remain strategic priorities.
The report also advocates “permissionless innovation,” faster translation of discoveries into products, increased use of real-world evidence, tighter integration of academia with industry, and stronger domestic manufacturing.
Its most futuristic section concerns AI. The proposed Genesis Mission would combine national datasets, supercomputing, AI models, scientific instruments, and autonomous laboratories. At the same time, Kratsios warns that AI could industrialize bad science, making reproducibility, machine-verifiable research, and automated scientific validation increasingly essential.
##
What are six surprises in the report?
##
It is much more radical than the title suggests. Rather than a celebratory science-policy essay, it argues that the postwar U.S. research system itself is outdated and needs redesign—from grantmaking and peer review to institutional structure, regulation, publication, and scientific verification.
It treats NIH-style peer review as a structural problem, not just a process problem. The report explicitly favors alternatives such as long-duration grants, fast grants, prizes, ARPA-like models, and even “golden tickets” that let one reviewer rescue a bold proposal rejected by consensus.
It is pro-science but skeptical of the biomedical research establishment. The report argues that large increases in biomedical funding have not produced commensurate gains in breakthroughs, and it implicitly challenges the assumption that “more NIH” is automatically the right answer.
It elevates foundational biology while downgrading the broader category of “life sciences.” Genomics, synthetic biology, neuroscience, molecular biology, and biomanufacturing remain strategic priorities, but the FY2028 guidance explicitly pushes agencies toward foundational biological science rather than simply expanding downstream life-science spending.
Its most futuristic proposal is not AI-assisted science, but AI-native science. The Genesis Mission envisions scientific foundation models, shared national datasets, autonomous laboratories, instrument interoperability, and closed-loop systems in which AI generates hypotheses, runs experiments, interprets results, and iterates.
It worries that AI could make science worse even while making every scientist more productive. One of the report’s sharpest insights is that AI may flood the system with papers, grants, and plausible findings faster than humans can verify them, so it calls for machine-auditable replication, automated verification, and continuous reproducibility infrastructure.
For a Natera, Tempus, or Freenome, my message would be: this OSTP report is not merely “science policy.” It is an early map of where federal research, regulatory, data, AI, and industrial-policy preferences may move over the next several years—and genomics sits unusually close to the center of that map.
To the CEO and Board
This is potentially favorable policy terrain for genomics—but not automatically favorable to every genomics business model. The report explicitly treats genetics, genomics, synthetic biology, quantitative biology, biotechnology, and biomanufacturing as strategically important foundational capabilities. The board-level question is therefore not “Will Washington still care about genomics?” but which parts of genomics will be treated as national infrastructure, which as commercial products, and which as legacy biomedical spending.
The administration appears to value platforms more than isolated tests. Shared datasets, scientific foundation models, automated laboratories, interoperable instruments, large-scale data generation, and precompetitive infrastructure are repeatedly favored. A genomics company should ask whether it can credibly present itself not merely as a seller of assays, but as an AI/data/biology platform capable of contributing to national-scale scientific infrastructure.
The traditional NIH-to-publication-to-commercialization pathway is being challenged. OSTP wants faster grants, long-duration awards, prizes, ARPA-like mechanisms, public-private consortia, and stronger industry participation. For a company, that potentially creates new ways to obtain federal partnership or validation that do not look like a conventional academic grant.
Regulatory strategy may become a competitive weapon. The report strongly endorses “permissionless innovation,” real-world evidence, reduced administrative burden, faster translation, and regulatory experimentation. A board should assume that companies capable of generating high-quality evidence rapidly—and of engaging FDA early around new evidentiary models—could gain a meaningful advantage.
Data assets may become more strategically important than the individual assay. OSTP treats curated scientific data as infrastructure for AI and specifically favors capturing negative results, operational laboratory data, and other material that is normally discarded. For Tempus especially, but also Natera or Freenome, the board should be asking: what proprietary dataset do we possess that becomes dramatically more valuable in an AI-for-science world?
Do not underestimate the China and domestic-capacity dimension. The report repeatedly links biotechnology to domestic manufacturing, resilient supply chains, strategic technology leadership, and national security. A company with U.S.-based laboratories, sequencing infrastructure, biobanks, reagents, compute, or manufacturing should think about how to make that part of its policy narrative.
The biggest risk is assuming that “biomedical innovation” itself guarantees favored status. OSTP explicitly argues that life sciences have absorbed a disproportionate share of non-defense R&D while productivity has disappointed. The winning argument will therefore be not “we are health care, please fund us,” but “we are foundational technology that increases scientific productivity, creates strategic capability, and produces measurable downstream value.”
I would phrase it differently to the C-level team as a whole
Yes. With the CEO and Board, I would emphasize strategic positioning, capital allocation, federal-policy risk, and potential competitive advantage.
With the broader C-suite, I would make it operational:
“Assume that Washington may increasingly reward genomics companies that can demonstrate five things: unique data, AI-enabled scientific productivity, rapid real-world validation, domestic strategic capability, and willingness to participate in precompetitive federal infrastructure.”
Then I would assign concrete workstreams.
The Chief Scientific Officer should identify which company assets could fit Genesis Mission-style science: multimodal datasets, longitudinal genomic records, foundation models, autonomous analytics, assay-development automation, or large-scale validation.
The regulatory leader should examine where FDA's evolving use of RWE, one-pivotal-trial concepts, adaptive evidence generation, and other streamlined approaches could change development plans. The report clearly endorses these directions, although it does not specify a new genomics-specific FDA pathway.
The government-affairs team should stop treating OSTP as peripheral. NIH, FDA, HHS, NSF, DOE, OMB, and White House technology policy are becoming more interconnected under this framework. The FY2028 annex is particularly important because OSTP and OMB explicitly tell agencies to use these priorities in budget formation.
The CTO/data team should ask what would be required to make company data AI-ready, interoperable, auditable, and scientifically reusable—because those characteristics are repeatedly privileged in the report.
The business-development team should look for federal consortia, challenge prizes, shared infrastructure, national missions, and cost-shared partnerships, not merely grants. OSTP explicitly encourages those mechanisms.
And the CFO should recognize one important nuance: this philosophy is simultaneously pro-innovation and skeptical of routine federal subsidy. Later-stage projects may increasingly be expected to attract private cost share, while government concentrates on foundational or precompetitive work.
If I had to reduce the CEO/Board message to one slide, I would probably title it:
OSTP Is Redefining What Counts as a Strategically Valuable Genomics Company
And underneath:
The favored company of this policy framework is not simply one with a good molecular test. It owns differentiated biological data, uses AI to accelerate discovery, validates rapidly in the real world, contributes to national scientific infrastructure, and can translate innovation into scalable U.S. capability.
For Natera, that points toward its enormous longitudinal clinical-genomic evidence base and scale of real-world testing. For Tempus, the fit with multimodal data, AI, clinical data infrastructure, and foundation-model thinking is almost explicit. For Freenome, the opportunity is somewhat different: position MCED not simply as another diagnostic product, but as a large-scale biology-and-data platform whose validation could itself become national scientific infrastructure.
That last distinction is where I would spend a surprising amount of CEO time.
#
#
#
A bonus. We closed by asking Chat GPT if the visions and changes forecast by "Science: Gold Age" could be leveraged by the MolDx program. Page 13ff.
