Note that this is still not the raw chain of thought - OpenAI's goal here is to be more transparent, but not entirely transparent. This, to me, implies that this should resolve NO unless they walk this back.

@MingCat https://x.com/polynoamial/status/1887621287616651429?s=46&t=Ue9JkPtZblyMSUmHw-Ugaw

@MingCat They are 'beautified' raw tokens, because the raw ones are meaningless to the user sometimes, because of multiple languages usage. I think it should resolve to Yes, or sth like 80% if it remains like that by June.

@ikoukas currently would resolve no.

@Ronan Is this decision based on a hard-and-fast rule that any sort of processing in principle resolves NO, or that your impression of the specific way OpenAI is beautifying them is too extensive to count as "raw tokens"? If it were somehow demonstrated that the current processing is very minor would you resolve YES?

@MingCat howdy . Formatting would be fine, but the tokens should be raw (which would just allow for Mark down rendering, for example).

Sam Altman claims that it will very soon

Makes a big difference seeing them in DeepSeek for sure