The real story here is this wonderful exposition in applying diffusion models to a time series data that is neither discrete nor continuous. It’s always fascinating to see diffusion models applied in different scenarios, same with diffusion language models.
I think you’re the only other person that read the article.