ParentFull threadamelius·The models are good enough now, so I'm waiting for the day they start selling inference ASICs with 100x the token output speed. See Taalas demo.View on HN