i could definitely be wrong, but my thinking is that models are very good at doing problems that have similar solutions in their training data at the time of release for every new line of GPUs, there is no training data available

→ View original post on X — @jxmnop