+1, this is a good thesis. It's plausible and if it's true, then that explains the opaqueness of this whole ordeal and why it's been applied to both OpenAI and Anthropic models now.