This interview https://youtu.be/oWOz2htozfI?si=qdQ0uZRoZOYeThOn from 2 days ago with a top researcher from OpenAI directly addresses the bitter lesson argument and the importance of scaling for the history of their models.