That's a good analogy. Makes a lot of sense. The only caveat I see is that there is a lot of context locked up in proprietary data sets (e.g. YT, books, podcasts) and I'm not sure how OSS models get access to that.