Comment by nylonstrung

I've found the same: strong types are disproportionately valuable in LLM coding vs human coding because they essentially function as context and the typechecker actively enforces correctness