I haven't had a chance to try any of this in anger yet, but it looks like a well-thought out effort. The API is built on top of the Guardian's Endeca search engine, so you get full text search as well as filtering by editorially assigned tags. There's a lovely API explorer, and the whole thing "speaks web" convincingly. Interesting to see they're using Mashery for API management.
It's also brave—pretty much all their text content is there. Understandably, they're moving cautiously through the beta phase. You have to get approved for an API key, the terms of use are subject to change, and by default you get your API calls capped at 5k per day, cacheable for up to 24 hours. That last restriction will rule out any personalised news targeting a large user base.
There are no photos or other media covered by the APIs—I expect rights issues relating to these will pose problems with making them available in future either. So you won't be releasing a rival site to the Guardian any time soon ;) But there is a fantastic looking data API, which gives you access to authoritative, curated data sets. This really works for me; the crowdsourced web too often overlooks the value of editorial winnowing and fact checking.
Maybe the most interesting bit (not yet on stream), is talk of opening up a commissioning channel for embedding widgets and apps in the Guardian network itself, with some kind of ad revenue split.
More at www.guardian.co.uk/open-platform