← Back to context

Comment by NekkoDroid

4 days ago

> Hasn't everything DeepSeek and Alibaba created thus far been distilled from the results of many, many accounts logging into Claude and ChatGPT?

I doubt it is really any different to what the US labs do [1]. I never really bought the "they were basically all just distilling from us" shtick from Anthropic, I just assumed they were either comparing or also creating training data as basically any lab is doing.

[1]: https://www.reddit.com/r/ClaudeCode/comments/1tqaist/opus_48...