In April, a team at University College London and the Research on Research Institute published a study tracking grant applications across seven major funders, including the National Science Foundation, the European Research Council, and the UK Research and Innovation councils. Since the public release of ChatGPT in late 2022, applications had risen 57 percent. Approval rates fell in near perfect inverse proportion. The researchers could not prove that AI wrote the proposals. They did not need to. The pattern was unmistakable: the cost of writing a grant application had collapsed, and every rational actor in the system responded exactly as you would expect. What makes this more than a volume problem is what comes next. The current generation of AI tools does not just help researchers polish their prose. Companies now offer agentic systems that identify funding opportunities, draft full proposals to specification, and submit them, with minimal human involvement. These are not hypothetical products. They are live, they are improving, and their business model depends on generating more applications, not better ones. The peer reviewers absorbing this wave are themselves stretched to a point that the system was not designed to handle. The NSF has awarded just 613 grants in fiscal year 2026, roughly 20 percent of its normal pace, partly because of political disruption and partly because reviewers cannot keep up with the volume. Nature reported that some funders are seeing reviewer pools shrink even as submissions balloon. The math is going in one direction: more proposals per reviewer, less time per proposal, lower signal in the process that is supposed to separate transformative research from noise. And now, from the opposite direction, comes a proposed rule from the Office of Management and Budget that would give political appointees authority to override peer review decisions on federal grants. We wrote about the broader push for political control of research funding in May. The OMB rule makes it structural. Under the proposal, agency heads appointed by the White House could overrule panels of scientists on which research deserves funding and which does not. The comment period has already drawn more than 90,000 responses ahead of its July 13 deadline. The New England Journal of Medicine compared the approach to Lysenkoism, the Soviet era practice of subordinating science to ideology. The comparison was not made lightly. What is striking is how these two forces converge. From below, AI agents are flooding the grant pipeline with applications that may or may not represent real research agendas, eroding the ability of peer review to function as a quality filter. From above, political appointees are claiming the authority to replace that filter with their own judgment. Neither force, on its own, would necessarily break the system. Together, they leave peer review squeezed from both ends, overwhelmed by volume it cannot process and overruled by authority it cannot challenge. The American research university was built on a particular bargain: the government funds the science, scientists decide what is worth funding, and the results belong to everyone. Every piece of that bargain is now in play. The AI tools rewrite who can apply. The OMB rule rewrites who decides. And the researchers in the middle, the ones still writing grants by hand at eleven at night because they believe the work matters, are watching both developments with the same quiet dread.