PLoS One's been launched in beta for a while now, but there had been technical problems that seem to have been fixed now. It's a great idea: a much lower barrier of entry to article acceptance, publishing articles of any length, with peer review provided by readers comments on the article after it's published and publication charges paid for by the authors' funding agencies.
Technically, PLoS should be built on solid foundations, as it uses the NLM Journal Publishing DTD to store articles as XML and retrieves all objects (articles, figures, etc) using DOIs. The underlying TOPAZ software is supposed to be released as open source, though there hasn't been anything made available yet. Hopefully this project should cross-pollinate well with OJS, which recently had its own annotation system added thanks to Geof Glass's work on Marginalia. PLoS only producing RSS feeds and still not even getting them right doesn't exactly inspire confidence though.
As far as published articles go, some people publish long, technical papers similar to those found in existing journals; others publish short, one experiment papers (which will hopefully get even shorter, if methods and introductions can be referenced elsewhere).
There was one article that caught my attention because of the wording of the funding section: "The authors claim they did not receive any financial funding." - suggesting, perhaps, that the publishers weren't entirely sure about that claim. This article in particular has quite a few style errors (even one in the title), so while peer review may come afterwards (and hasn't in this case, yet, but maybe someone out there is repeating the experiment themselves), there's still a role for publishers in copy-editing articles for readability. It would be good if authors have enough control over their published papers that corrections can be made at a later date, and with archiving of open access articles using LOCKSS, updated articles could feasibly be distributed to multiple archives.
It's a shame that Chris Surridge is already lamenting the lack of comments on papers, when the infrastructure isn't in place to properly handle discussions at the moment. It's not surprising that people are more willing to comment on papers within their own communities where they can see that discussion threads are treated as important, permanent content and displayed appropriately.