The Association of American Universities' April 24 analysis put the National Institutes of Health grantmaking rate at less than half of last fiscal year's level. [1] U.S. News confirmed the structural number: the National Cancer Institute earmarked roughly $72 million for new and competitive grants by late March, against a typical $250 million in the same window. [2] One NIH institute warned internally that $500 million may go unobligated by the end of the fiscal year. The agency is now seven weeks into a quiet decommissioning of its own research pipeline.

The mechanism the AAU and U.S. News both named is a computational text-analysis tool. The tool scans proposals for terms including "racism," "gender," and "vaccination refusal," and routes flagged applications out of the normal review queue. [1][2] Inside Higher Ed reported in February that 15 of 27 institute directors are still in acting capacity. [3] The grant rate is a downstream symptom; the staffing and the keyword filter are the upstream causes.

The Grant-Witness citizen-science tracker — at grant-witness.us — has been collecting individual termination notices from researchers since February, with NIH, NSF, EPA and SAMHSA awards each tagged by agency, date and amount. [4] No major science outlet has yet cited the tracker. The week's silence on it is itself the artifact: a structured dataset on the actual cuts, built by the people receiving the letters, sitting outside the press the cuts are designed to reach. The 5- and 10-year output curves the rate decides do not show up for years. The decision week is now.

-- KENJI NAKAMURA, Tokyo