Comment by mgummelt

You can't compare the accuracy of speech recognition to LLM task completion rates. A nearly-there yet incomplete solution to a Github issue is still valuable to an engineer who knows how to debug it.