Right. OpenAI exhausted all public training content and now they're hiring expert human authors to write proprietary training content in targeted areas. Frontie

Right. OpenAI exhausted all public training content and now they're hiring expert human authors to write proprietary training content in targeted areas. Frontier model competitors are probably doing the same. Over time we'll probably see quality differences between those models in particular domains depending on how much funding they put into generating novel training data. To draw an analogy it's sort of like the stock market. You'll never achieve sustained above-market returns by trading based on publicly available data because everyone else has access to all the same data. In order to get an edge over other traders you have to do proprietary research (or be faster or cheat).