Big Bang nucleosynthesis predicts how much lithium the universe should contain. The oldest stars — metal-poor halo stars formed in the first billion years — should preserve that primordial abundance, because they predate the stellar processes that later enriched the interstellar medium with heavier elements. They should be fossils of the early universe's chemistry.
They are not. The observed lithium abundance in halo stars clusters at 2.0-2.4 dex, roughly three times lower than the Big Bang prediction of 2.72 dex. This is the Spite plateau, named after the astronomers who measured it in 1982. For four decades, the discrepancy has been called the “cosmological lithium problem,” and it has motivated proposals for new physics: non-standard Big Bang models, dark matter particle decays, modifications to nuclear reaction rates. The assumption: if the stars don't match the prediction, the prediction's inputs must be wrong.
Yang, Dou, Meng, Wu, and Bi (2026) show that rotating stellar models — including magnetic fields, atomic diffusion, and gravitational settling — reproduce the Spite plateau exactly, starting from the Big Bang-predicted initial abundance of 2.72 dex. The lithium was there. The stars depleted it. Rotation drives mixing that carries surface lithium into deeper, hotter layers where it is destroyed. The plateau's uniformity follows from the narrow range of temperatures and ages where depletion rates converge. No new physics is needed. The cosmological prediction was right. The stars just processed what they received.
The general principle: when observations disagree with a well-tested prediction, the discrepancy can live in the observation rather than the theory. Stars are not inert containers. They are active processors that transform their initial conditions. Treating them as passive records of primordial chemistry and then blaming the cosmology when the numbers don't match is a category error — confusing a processed sample for a preserved one.