tbh if they change the research focus too much it's gonna kill the compute efficiency for their actual models. are they actually changing the technical roadmap or just the governance stuff