Not that I care if Twitch supports this or not but, as a engineer on a large cloud-based human resources software package, Swagger files are provided for API endpoints for a simple reason: they are accurate. Twitch has been known to not have accurate documentation and return data models that do not match the documentation. Given their track record, having an automated Swagger solution provided would resolve some of those issues. You can use the YAML to validate the returned JSON and know what to expect and can skip potentially faulty documentation and rather have a generic representation of the data model provided that can be easily parsed.
If he were to make the YAML himself and the documentation is wrong or Twitch changes something, what automatically updates his YAML configuration? This is why automation from the source is important and valuable.
Just my two cents.
Cheers!