AI for science
Seventy years of US national-lab data have never been uploaded to a model. The White House bet is that unlocking the archive doubles American scientific productivity — not building a better chatbot.
Michael Kratsios, who runs the White House Office of Science and Technology Policy, frames the bottleneck as archival, not algorithmic. “The US national labs have a tremendous amount of scientific data across a wide variety of domains that have not been made AI ready, that have not been uploaded to models and have not been part of the scientific process.” The opportunity, he says on Peter Diamandis’s Moonshots podcast in July 2026, is to “bring all that great data from 70 years of scientific discovery in the US into an AI model to be able to accelerate discovery.”
The vehicle is the Genesis Mission — a Manhattan-project-style presidential program signed in December, aimed at applying frontier AI to scientific discovery “across all of government,” not at one agency. Diamandis calls it “a Manhattan Project style effort to accelerate breakthrough discoveries.” The stated goal Kratsios carries publicly: “We essentially want to double the productivity of the entire scientific enterprise in the United States over the next decade.” He notes that 2× target is a number from the prior year when they were building what the president signed; they discussed raising it internally but kept the public figure.
Data, not a new model
The load-bearing mechanism is ingestion, not a bespoke foundation model headline. Genesis is a bet that verification-rich scientific domains — chemistry, physics, biology, math — are where AI floods first, the same split the wiki tracks separately under jagged capability and automated-research boundaries. Unsolved-problem benchmarks now have an administration venue plus a funded upload program.
Funding landed in public view the week of the Genesis Mission summit: Kratsios says they announced “about $5 billion worth of genesis mission grants just last week,” many biotech-nested. That “last week” is a late-July 2026 snapshot from the program’s architect — the wiki flags it for re-validation against a DOE or Treasury grant list. A sibling quantum executive order is a separate national mission; the wiki treats it as not Genesis itself.
The most important thing for the nation is to make sure that we’re applying this unbelievable technology called artificial intelligence to scientific discovery. And it’s not just at one agency, it’s across all of government.
Michael Kratsios, Moonshots, recorded 2026-07-29
What would falsify the chain? If a lab inventory showed the bulk of national-lab archives were already in foundation-model training sets before Genesis. Or if measured scientific-productivity metrics were flat or down three years after the December signing despite material grant awards in FY2026–FY2027 public reports. The 2× line is a political target, not a measured baseline — Kratsios himself treated it as negotiable before keeping the public number.