Math was rediscovered for centuries; LLMs finally made theorems searchable

konstmish · x · 2026-11-22

konstmish argues that mathematics has historically been extremely repetitive: backprop, VC dimension, KKT conditions and many other results were independently rediscovered multiple times. Even with Google, math statements and proofs were unsearchable because notation differences break keyword matching.

LLMs solve this by understanding semantic equivalence across notations. We're now sitting on a goldmine of possible results — there has never been a better time to be a mathematician who cares about getting the work done rather than enjoying doing it all by hand.

Original post →

More from AGI Musings

AGI Musings channel →