Think of this as text-to-speech for articles you would normally read, except done by an actual human so it doesn't sound off-putting.
I think it's a fantastic idea, but its success will be determined by the team's ability to scale across a wider gamut of content.
Each station/organization curates information (in a similar way web-based reporters do) and then reports as much of it as their resources allow via someone reading/talking?
I'm not trying to beat up on their idea but I dont understand their competitive angle or what would cause me to use it over the huge amount of content out there.
Moreover, there is simply more written content than there are podcasts or radio. This can help bridge the gap.
Radio news... I haven't really looked into this since moving to the US, but in the UK, no radio news channel covered technology well and it's difficult, given a finite amount of time, to use that time to listen to exactly what you want. e.g. tuning into the World Service here, I get some very random content which is sometimes interesting, sometimes not. Perhaps NPR is different.
I feel Umano addresses both of these - topic freshness/specialisation and good use of finite time - in different ways.
I've also got a few things subscribed that are "genuine" podcasts. My observations mirror your own (a few good ones/almost all are terrible). But if it works, it works really well, I've had podcasts that went on about a topic for 3 hours without getting boring.
But almost none of the stuff I get this way deals with technology per se. It's mostly politics, science, sociology, history -- anything but pure tech.
See API flowchart: http://voicebunny.com/developers/get_read