Regarding obsolete models, that's only partially true. There's loads of content that are effectively "finished" and won't be changing, and will grow obsolete at a fairly slow pace. Meaning they'll be useful in the models once trained for years.
Obviously new technology and similar ideas/content that didn't exist when the model was created won't be there, but the amount that changes and or is new is relatively small each year compared to all the historical content.
This looks really good. Thanks, I had no idea it was out there.