{"slug": "the-overlooked-memo-that-could-reshape-federal-science", "title": "The overlooked memo that could reshape federal science", "summary": "The Office of Inquiry's Stuart Buck, Aishwarya Khanduja, and Hiya Jain argue that a joint OSTP-OMB memorandum on FY2028 R&D priorities, released in late July by Michael Kratsios and Russ Vought, carries more practical force than OSTP's broader 'Science: A New Golden Age' report. Because OMB controls federal funding, the memo directs agencies to build 2028 budgets around earlier-stage basic science and requires agencies spending over $3 billion on R&D to submit action plans within 90 days.", "body_md": "By **Stuart Buck**, **Aishwarya Khanduja**, and **Hiya Jain**\n\nOver the summer, the White House Office of Science and Technology Policy, led by Michael Kratsios, released “[Science: A New Golden Age](https://www.whitehouse.gov/science/).” Expressly modeled on Vannevar Bush’s classic *Science: The Endless Frontier,* this major report lays out a number of ideas on how U.S. government funding of science should work in order to  “match the achievements of our past and the ambitions of our future.”\n\nSuch statements were greeted with a fair bit of acclaim — and even more widespread suspicion, given that the administration has spent over a year canceling research grants and proposing to cut science budgets.\n\nBut the administration is by no means a monolith. We, at the Office of Inquiry, believe that the administration includes people who have a sincere interest in issues like innovation, scientific progress, reproducibility, AI, and better data.\n\nThus, we want to focus on what is actually possible in the near future — and the key to understanding this lies not with the OSTP report but in another, less-read document.\n\nOSTP, despite being a White House office, is relatively powerless. It does not control any agency’s budget, has a relatively small budget of its own, and does not issue regulations or have oversight authority. Its main authority is through more indirect means, such as coordinating access to the President.\n\nIn fact, OSTP folks are often *not* in the room when important White House meetings are being held. That is partly because, despite the fact that many White House communications are subject to executive privilege or the presidential communications privilege[1](#footnote-1), OSTP’s emails and documents are [generally subject](https://www.congress.gov/crs_external_products/R/PDF/R46238/R46238.6.pdf) to the Freedom of Information Act. The D.C. Circuit [has held](https://law.justia.com/cases/federal/appellate-courts/cadc/15-5128/15-5128-2016-07-05.html) that even private emails sent and received by OSTP employees can be subject to FOIA; this means that anyone in the world can submit a request and then (eventually) get access to those emails or documents, and then publish them for the entire world to see.\n\nAs a result, more important than the Kratsios report is the Joint Memo that Kratsios (OSTP) and Russ Vought (Office of Management and Budget) simultaneously [released in late July](https://www.whitehouse.gov/wp-content/uploads/2026/07/White-House-Fiscal-Year-2028-RD-Priorities-Memorandum.pdf).\n\nRather than being an inspirational and thoughtful 120-page essay on everything from implicit knowledge to manufacturing policies to the nature of scientific creativity, the Joint Memo issued by OSTP and OMB is a shorter and more concise directive to all federal agencies. It has much more direct authority than the Kratsios report, because OMB has immense authority over the flow of federal dollars.\n\nThe Joint Memo is the place where rhetoric is more likely to become reality. Let’s break it down.\n\n### What’s inside the Joint Memo?\n\nThe joint memo has two main sections: R&D Priority *Areas* and R&D Priority *Practices*. For the Areas section, all relevant federal agencies are required to come up with new budgets as of 2028 that follow the guidance there. For the Priority Practices sections, things are far more urgent: Agencies that spend more than $3 billion on R&D are required to come up with an action plan *within 90 days*[2](#footnote-2).\n\nOK, so let’s dig a bit deeper into these two sections.\n\nAs for R&D Priority Areas, there are two big themes. The first is that basic science is more important for the federal government to fund than translational research or commercialization, which the private market may be better suited to support.\n\nIndeed, the memo adds that in 2028 budgets, agencies should “identify the specific programs through which the agency proposes to shift its portfolio toward earlier-stage work,” and “where agencies propose to significantly expand later-stage development activities, they should justify why such activities would not occur absent federal support.” Note, however, that it isn’t yet clear what this requirement means, given that other administration priorities have involved greater (not less) support for commercializable research.\n\nThe second big theme is AI for science. The Joint Memo argues that federal agencies should prioritize AI in a number of ways: align with the Genesis Mission (a large effort at the Department of Energy and elsewhere to expand the AI capabilities supported by the federal government), use AI as an “instrument of scientific discovery” (focusing on “work that industry is unlikely to pursue on its own”), develop area-specific AI models in fields like quantum information science, developing and maintaining “ambitious new datasets that could open entirely new fields of inquiry” given the power of AI, and integrate AI with scientific instruments and robotics. Moreover, agencies should look for ways to support infrastructure and advanced compute.\n\nA third theme (albeit a lesser one) is that agencies should figure out how to fund R&D in ways that build “domestic manufacturing capacity and supply chains” in areas like semiconductors, advanced materials, and robotics.\n\nWhat about the Practices section, where agencies over $3 billion have to come up with an action plan within 90 days?\n\nThis is where the more interesting metascience ideas come in. Some highlights include:\n\n- Agencies should “support a diverse portfolio of institutions” such as the NSF’s X Labs initiative is now doing, and should work to identify the gaps “that remain unaddressed because no existing institution is well-suited to pursue them.” \n  - There is a deeper implication here. Institutional form shapes what kinds of scientific questions become possible to pursue. For example, focused research organizations open up the possibility of tackling problems that require a 20-person team and five years. More broadly, there are a wide range of scientific problems that are best tackled by particular combinations of time horizons, disciplines, technical infrastructure, incentives, or researcher autonomy, and those requirements may point toward an organization that does not yet exist.\n  - This also creates a selection problem for funders. Funding systems primarily encounter scientific proposals that the existing institutions (such as universities) are already capable of tackling. Entire areas of inquiry may remain invisible because the right institution hasn’t been built or even conceived of yet. Agencies should therefore treat organizational design as part of the scientific process. That is, they should identify a range of important questions that existing structures struggle to address, then experiment with new institutional forms that might be better suited to tackle them.\n\n- Agencies should increase the number of long-term grants of at least five years or more, which “give our best researchers the time and autonomy to pursue bold, ambitious projects whose most important results may take years to emerge.” Yes, yes, and yes. Note that while Pierre Azoulay’s [famous paper](https://www.nber.org/papers/w15466) comparing HHMI to NIH is often seen as a justification for funding “people not projects,” he told Stuart once that he thinks the actual secret behind HHMI’s funding is just the longer time frame so that people can explore bigger questions without feeling the need to publish*now* .\n\n- Agencies should create (or expand) mechanisms to hand out “flexible, low-friction ‘fast grants,’” including “simplified applications” of a few pages while promising to review applications in under a month. The inspiration here is obviously the [Fast Grants initiative](https://future.com/what-we-learned-doing-fast-grants/) that launched during the COVID-19 pandemic.\n\n- Agencies should look for places to use prizes and challenges for important national missions, and should try to “maximize participation from nontraditional teams.”\n\n- Agencies should look for ways to pilot and evaluate new funding mechanisms, including the “golden ticket” model of peer review, advance market commitments, regranting models that “delegate funding authority to working scientists” (!), and even “more speculative approaches such as quadratic funding or eigenfunding.” \n  - The opportunity here is significant, because each such pilot will become a source of evidence about how scientific funding works. Agencies could define in advance what a new mechanism is intended to change, collect comparable data across pilots, track both awardees and near-miss applicants over time (in a regression discontinuity approach), and make the results available for future program design. Over time, this would allow funding experiments to accumulate into an empirical knowledge base about which mechanisms work best for which kinds of scientific problems.\n\nIn addition to all the above, there are many ideas about expanding the search for scientific talent — such as “expand advanced K-12 STEM enrichment opportunities” and “expand hands-on technical learning” — that aren’t as pertinent to what NIH and NSF normally do, but are nonetheless good ideas.\n\nOne thing that particularly caught our eye: a recommendation that agencies should “support individuals agnostic of institutional affiliations,” similar to the the NSF’s Graduate Research Fellowship Program that allows the grantees to carry their funding to the institution or lab of their choice, rather than being pinned down to one particular lab’s grant. This seems like a great idea. Too many graduate students and post-docs are tied to one lab for their employment and salary, which means they are constrained both from pursuing their scientific curiosity and from reporting on abusive working conditions that include the demand to produce positive results (which is a huge incentive to commit fraud).\n\nThe next part of the R&D Practices section notes that the federal government spends around $200 billion a year on R&D, but next to nothing on studying “funding mechanisms, institutional models, workforce development programs, and research practices.” This section asks federal agencies to do things like:\n\n- Create a metascience office that will “evaluate what programs actually work and drive organization-wide reforms”\n- Try to map out the gaps across an entire landscape. That is, agencies should systematically compare their existing grantmaking to the “landscape of unsolved scientific and technical challenges in their domains, rather than relying primarily on historical funding patterns.”\n- Build more databases for metascience, including “longitudinal tracking for both awardees and near-miss applicants.” Great idea: See Stuart’s piece with Chris Marcum, “ [Let Unfunded Grant Applications See the Light of Day](https://issues.org/unfunded-grant-applications-open-science-buck-marcum/) ,” and another piece with Sandeep Patel, Amit Gupta, and Ian Banks, “[America Already Built the World’s Largest Science Vetting Machine](https://newsletter.goodscienceproject.org/p/america-already-built-the-worlds) .”\n- Give program officers greater discretion to “build a research portfolio” and “manage toward ambitious outcomes,” and try to recruit more top talent from outside government to serve as program officers.\n- Reduce the burden of bureaucracy. An obvious yes.\n\nA final set of recommendations is about integrating federally-sponsored R&D into the “far larger national S&T enterprise.” This is interesting, because it is exactly what Jeffrey Tsao (recently retired after a long career at Sandia National Lab) has been [proposing](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5600892) with regard to a new form of Bell Labs: “Agencies should propose funding consortia and partnerships that integrate basic research with manufacturing R&D, reflecting the multidisciplinary, engineering-intensive way science is conducted today.” Tsao’s idea was also, among others, partly an inspiration for the recent NSF [X-Labs Initiative](https://www.nsf.gov/funding/initiatives/nsf-x-labs).\n\n### Where the memos contradict\n\nFor all of the good ideas in the Joint Memo, it doesn’t always align with the longer OSTP report. In these cases, the memo takes precedence.\n\nFor example, the OSTP report firmly recommends expanding the National Science Foundation’s Graduate Research Fellowship Program. But the administration’s [budget for fiscal year 2026](https://nsf-gov-resources.nsf.gov/files/00-NSF-FY26-CJ-Entire-Rollup.pdf#page=28) proposed *cutting* that program by 55% and its [proposed budget for fiscal year 2027](https://nsf-gov-resources.nsf.gov/files/FY-2027-NSF-Budget-Request-to-Congress.pdf#page=29) proposed cutting it by 43%.\n\nIs the administration position to expand this program or to radically cut it? The numbers tell an interesting story. In 2025, the NSF awarded merely 1,000 of these prestigious fellowships, but earlier this year, it [handed out 2,599](https://www.nature.com/articles/d41586-026-01119-1) — which was an all-time record! But the [most recent NSF announcement](https://www.nsf.gov/funding/opportunities/grfp-nsf-graduate-research-fellowship-program/nsf26-526/solicitation) contemplates merely 1,500 awards for 2027. Just by the public numbers, there must be some internal debates as to the funding level for this program.\n\nAnother discrepancy is that the OSTP report has a long section on the reproducibility crisis across many fields of inquiry, and points out that AI could make the problem worse. The OSTP report calls for more investment in AI tools to check for replicability, and praises the NIH’s current efforts to fund more replication studies and to create more infrastructure to address this problem.\n\nThe number of times that the words “replication” or “reproducibility” appear in the Joint Memo? Zero.\n\nTo be sure, the OSTP report was long, and full of many good ideas that didn’t make it into the Joint Memo. The point is that those of us who care about metascience need to 1) focus on what is *actually required* of federal agencies by the Joint Memo (and check on whether the agencies actually follow through with a meaningful action plan) while 2) continuing to lend support to great ideas that as of now are aspirational without budgetary teeth.\n\nFor example, the National Science Foundation recently put out a [10-page memo](https://nsf-gov-resources.nsf.gov/files/Statement-NSF-Leadership-Golden-Age-Science.pdf) detailing its plans to establish a metascience office, use different models of peer review, expand the use of rapid grants with short applications, look to create new institutions altogether, and accelerate AI in science.\n\nThe NSF memo was greeted with much criticism and suspicion, in part based on [a story in](https://www.science.org/content/article/nsf-memo-implementing-trump-s-golden-age-science-unsettles-researchers) *[Science](https://www.science.org/content/article/nsf-memo-implementing-trump-s-golden-age-science-unsettles-researchers)* whose sub-headline announced that NSF “emphasizes support for White House priorities over curiosity-driven ideas.” That is an over-interpretation, as the NSF memo repeatedly points out that NSF’s “unique and indispensable contribution” to federal R&D “is foundational research, spanning both curiosity-driven and use-inspired discovery,” and that NSF “will balance funding opportunities focused on exploration by the scientist (‘investigator-driven inquiry’) and NSF-directed opportunities in areas of special emphasis or need.”\n\nMost recently, the White House [announced](https://www.whitehouse.gov/fact-sheets/2026/10/fact-sheet-trump-administration-announces-the-most-ambitious-set-of-science-initiatives-this-century/) the “most ambitious set of science initiatives this century” (with perhaps a bit of bravado), including a new X-Labs Consortium with outside commitments (e.g., philanthropy and industry) of $310 million. Moreover, the order announced a broader commitment to standing up metascience units not just at NSF but at the Department of Energy, Department of Defense, and NASA, as well as a partnership with the UK government’s own Metascience Unit.\n\nNonetheless, putting out a memo or even an executive order is a long way from actually carrying out any of these initiatives. What is needed now is a Congressional commitment to authorize and fund such programs on a long-term basis, and in particular to give metascience units enough of a budget and authority to implement experiments in the face of likely internal pressure from existing bureaucracies. The challenge is heightened here given that DOGE succeeded in firing or laying off so much of the workforce at NIH and NSF.\n\nRhetoric is just a start. We now need to see whether these many great ideas become reality.\n\n*Stuart Buck is the Executive Director of the Good Science Project, and formerly was Vice President of Research at Arnold Ventures.*\n\n*Aishwarya Khanduja is the founder of the Analogue Group.*\n\n*Hiya Jain is Head of Editorial and Inquiry at the Analogue Group, and a fellow at the Good Science Project.*\n\n*All three collaborate on the Office of Inquiry, which focuses on new scientific institutions, metascience experiments, new tools and financial instruments, and policy work in those domains.*\n\n[1](#footnote-anchor-1)\n\nThe Supreme Court has held that the presidential communications privilege applies to “communications in performance of [a President’s] responsibilities, of his office, and made in the process of shaping policies and making decisions.” Nixon v. Adm’r of Gen. Servs. 433 U.S. 425, 449 (1977).\n\n[2](#footnote-anchor-2)\n\nThat said, agencies often find reasons to delay, and it seems that of yet there is no plan to make these plans public", "url": "https://wpnews.pro/news/the-overlooked-memo-that-could-reshape-federal-science", "canonical_source": "https://asteriskmag.substack.com/p/the-overlooked-memo-that-could-reshape", "published_at": "2026-10-13 12:00:56+00:00", "updated_at": "2026-10-28 08:45:10.449459+00:00", "lang": "en", "topics": ["ai-policy"], "entities": ["Office of Science and Technology Policy", "Office of Management and Budget", "Michael Kratsios", "Russ Vought", "Office of Inquiry", "White House", "Stuart Buck"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/the-overlooked-memo-that-could-reshape-federal-science", "markdown": "https://wpnews.pro/news/the-overlooked-memo-that-could-reshape-federal-science.md", "text": "https://wpnews.pro/news/the-overlooked-memo-that-could-reshape-federal-science.txt", "jsonld": "https://wpnews.pro/news/the-overlooked-memo-that-could-reshape-federal-science.jsonld"}}