Imagine three witnesses who saw the same event. Two tell the same story, the third tells a different one. Whom do you believe? This is exactly the situation in astronomy when measuring carbon in the star J0815+4729, an ancient object with a record-low iron abundance (carbon is one of the key elements used to infer the properties of the first stars).

For decades, the main witness was molecular CH, the absorption bands in the star's spectrum. Atomic C I lines in such objects are too weak, so they were rarely used. But J0815+4729 is unique: it is so carbon-rich that both atomic lines and molecular C2 bands are visible in its spectrum. This provided a rare opportunity for a cross-examination.

The discrepancy could not be dismissed as measurement errors or the star's temperature. Even accounting for three-dimensional gas motions in the atmosphere only widens the gap. It seems that CH lines in extremely iron-poor stars behave differently than previously assumed.

If CH is systematically wrong, then all previous estimates of carbon in the most primitive stars may be incorrect. This would change our understanding of how the first supernovae enriched the early Universe. The classification of such stars as CEMP-no or CEMP-s may also need revision.

J0815+4729 is the first case where we have three independent witnesses. And two of them say the third is wrong. Now we need to check whether CH also errs in other ancient stars, where it remains the only available witness.