>>521126722
All of what you just wrote is bullshit because it's not how these LLMs work. They probably tried some continuous learning technique on a giant baked model and it didn't work (because none of them do).