Formal verification — (machines constructing and checking mathematical proofs) — keeps getting better. Cobblestone ( @TaliaRinger https://arxiv.org/abs/2410.19940
) and other divide-and-conquer approaches now automate proofs we once thought unreachable.
But something feels different about where this is heading.
As we push automation forward, we’re starting to touch the edges of what a “proof” actually is — not just a verified computation, but a statement about what can’t be opposed.
I’m not sure the field has fully realized how close we’re getting to an ontological shift in what counts as proof itself.
 
      
  
            