That transcription seems to be really bad. It's full of strange errors. :(
I wonder what they used to generate that transcription?
I was kinda wondering too, and did a (very shallow) dive into the JavaScript on that page. I'm almost positive they are using Deepgram(dot com)'s speech-to-text service.
I ran whisper.cpp on that audio file on my laptop, and it does a reasonably well job too.
They discuss their transcription process at 23:15 in episode #621