[serjester]

This is a hand wavy article that dismisses away VLMs without acknowledging the real world performance everyone is seeing. I think it’d be far more useful if you published an eval.