Wouldn't it be fun if they were both correct, but with wildly different structures? We could maybe use the proof difference to characterize the prover difference (AI-assisted vs not).
this is actually something i'm trying to study
this is actually something i'm trying to study