valid point. its more of a stepping stone towards larger models. we're figuring out what the best way to do this is before scaling up.

If there's very little text before the internet, what would scaling up look like?