You're overstating the situation. The previous top-performing protein folding method is essentially impossible to reproduce from public literature, and availabl

You're overstating the situation. The previous top-performing protein folding method is essentially impossible to reproduce from public literature, and available only via licensing agreement with the lab that created it. More generally, while it's far more common to find open-source implementations of lab research than it used to be, it's still quite common for code to be closed, and important methodological details to be secret. Even in the "open" fields of ML, today, I routinely find papers where essential bits of code, parameter sets, etc. are not made available.