Wang

New capabilities seem to appear out of nowhere as models scale. Whether that's real — or just how we measure — is still an open fight.

Hand-editing a model's output gives three training signals at once. Writing notes gives one — and a weaker one at that.