I think what you're saying is consistent with the quote: human curation of SFT data is indeed not complex. There might be extra work on top of that RL, but it's the same work that's been done throughout LLM development.
Additionally, in the following days, I've seen evidence suggesting that the SFT part might not even be necessary. I'd argue that work wouldn't have happened if R1 wasn't released in the open.
