[bglazer]

First the search space is way too large for brute force enumeration. We’re talking like 10^300 combinations. Also the hard part isn’t just listing amino acid sequences, its finding ones that do what you want them to. The only way to figure that out is by testing them, which is difficult and expensive. So you need an algorithm that is good at only listing sequences that are likely to work. That’s precisely what LLM’s are good at: finding patterns and sequences that are correlated in a useful way