@gakonst This is why I’m so much more interested in languages like Agda and Idris2 that have intrinsic proofs that compose more nicely with the actual type system. Lean and coq are awesome, but the extrinsic proof approach feels like it inevitably leads to your latter example.
@zachlieberman I love this quote from a recent Laurie Spiegel interview:
> You can’t replace human self-expression or the need for it by simulating their results.