Comment by verytrivial

Is there evidence for the claim that the training was that much cheaper? Or is is 'apparent' by inspection of the verified ingestion architecture?

I can't seem to find any evience beyond the team's statements. What have I missed?