It's also slightly wrong, and needs some proofreading.

Example:

  > Get at.

That's not what he says at the 24 second mark. If you can't get the first 30 seconds of an hour-long transcript right, why should I read on?

(btw, I get that it was an automated transcript [http://temi.com], and has some OCR-like errors, which is apropos, considering the context of the speech)

Yes, the transcription errors convey a vaguely machine-generated feel, and together with the weird typography (contrast insufficient, background too bright, line spacing too tight, timing marks every 30 seconds) you ask yourself if the whole speech wasn't a machine-generated prank. Considering the subject of that speech, that's ironic.

The best part is that this "Audio to Text" transcription service says on its web page:

Proprietary algorithm

Built by our machine learning and speech recognition experts.

Who are experts in learning new machines and can nearly always recognise speech, but generally prefer email.

As someone who has really enjoyed reading Mickens' highly-caffeinated essays, I was pleased to see that his speaking style is even better!