I’ve pondered this a bit. If we lock down akin to china, we’ll have to start using plausible deniability. Our encoding will need to be in safe and clear looking datasets, which localLLM probably could push out