The Quota, Not the Ban

Scientific commons eventually run out of moderators before they run out of papers, and arXiv's October 1 policy is the lever that follows. The mechanism is a quota, not a ban: count authors, not ideas, and cap the ones who overwhelm the volunteer queue.

Wires will read this as "AI floods arXiv." arXiv's post says something narrower. The cap is per submitter, applied across every category, and the three-active-submission limit predates the AI surge, having existed since 2024. Rejected submissions count against the monthly allowance. AI assistance remains permitted when research meets scholarly standards. arXiv is using a quota system borrowed from rate-limited APIs to ration moderator attention, not adjudicating content.

The reusable move is older than software: when a commons cannot expand its labor pool as fast as its input stream, the only scalable lever is per-author throughput. The 40,363 September submissions, almost 4x the 9,869 of September 2016, generated nearly 9,000 support tickets. arXiv's choice is the choice any volunteer-moderated system eventually faces: throttle the contributor, not the contribution.

The honest counterweight: arXiv reports no measured effect on quality, delay, or legitimate heavy submitters. The "commons preservation" frame is stated intent, not demonstrated outcome. What would prove it wrong is the same data set six months from now: if the cap barely moves submission concentration or ticket volume, the lever was a gesture.