arXiv submissions jumped 96% in two years. Its volunteer moderators say they can't keep up.

arXiv, the free, public preprint server where researchers post papers before peer review, saw submissions jump 96% in two years, to 40,363 in September 2026 from 20,569 in September 2024. The working mathematicians behind it, from Terence Tao to Kevin Buzzard at Imperial College London, are now watching a community-run research commons face a new kind of load as the cost of producing paper-shaped text falls toward zero.

Submissions went from 20,569 in September 2024 to 26,646 in September 2025, then climbed to 40,363 a year later, about 52% in the most recent year and 96% over two (New Scientist). The growth is not coming mostly from working mathematicians. It is coming from outside the field, from people who now have access to AI tools that can solve reasonably difficult problems and auto-generate drafts shaped like the ones arXiv was built to host. That changes who arXiv is serving: a system designed for professional researchers, now flooded with work from anyone with a model and a problem statement.

Kevin Buzzard, who has moderated arXiv for years, says the volunteer system is no longer keeping up. The review queue now runs weeks. He used to read submissions manually each day; he now runs them through AI first, to figure out which are worth his time. "ArXiv is a serious problem, and it might not survive the onslaught," he told New Scientist. "Journals are also under more and more pressure. The issue is that slop papers do not explain stuff well. It's not just about correctness; it's about human understanding, and AI is sometimes very poor at explaining."

Chaim Goodman-Strauss said he read a paper this week, one that included a valid solution to a long-standing problem, and described it as "basically a piece of slop." The result was real; the explanation was not.

arXiv is a free, voluntary system where researchers post drafts before peer review. It runs on a small staff and a larger pool of volunteer moderators. The unit cost of producing something that looks like a paper has collapsed. The unit cost of telling whether a paper is worth a researcher's morning has not.

The question is what gets built next. arXiv says it is responding to the surge, though the details of that response are not yet public. The old design assumed that anyone who could write a paper could be trusted to post one, and that moderators could keep up. Both assumptions are under load at the same time. The harder question is governance: who sets the policy, and how the community that depends on arXiv gets a say when the volunteers who run it are stretched thin. What arXiv becomes over the next year is a build decision, not a surrender.