As models get better, fine-tuning is going to become less important for the vast majority of use-cases. But there will continue to be use-cases where it unlocks massive value, at least for a couple years.

→ View original post on X — @mattshumer_