Seventeen years ago, Science rejected Ingber's lung-on-a-chip and asked for mouse data as validation. The FDA has now — as of September 2026 — proposed rulemaking replacing "animal tests" with "nonclinical tests" in drug-development regulations, after a pharma company was denied a trial for *presenting only animal data*. The pendulum has swung entirely around. The technical picture for practitioners: organs-on-chips are microfluidic devices lined with human cells (lung, heart, kidney, placenta; up to 10 linked into multi-organ systems), plus organoids and computational "digital twin" models. Emulate's liver-on-a-chip flagged roughly 7 of 8 drugs that passed animal trials but were toxic in human livers; a heart simulation hit 89% accuracy on arrhythmia prediction versus 75% for animal studies. But validation is the bottleneck: that Emulate liver study took 870 chips, 16 FTEs for 16 weeks, and a separate study would be required for every NAM, every drug class, every use context. The article's real thesis is sociological: FDA Modernization Act 2.0 (2022) and NIH grant requirements changed the formal rules, but reviewers and journals still expect animal data, careers are built on animal models, and the monocyte activation test took 15 years to formally replace rabbit pyrogen testing — with rabbits still in wide use. The technology is nearly ready; the humans are not. Somewhere in Palo Alto, Dario Amodei read "wet lab" in this article and founded three more of them.