It was particularly timely for me to attend Peter Morville's ETech talk, Ambient Findability, as I'm about halfway through his book of the same subject and title. It was hot, standing room only, and well worth the discomfort. Many good points were raised including:

The Google Problem...
Site search remains the major shortcoming of most sites on the Web. It's a real challenge on its own that's further amplified by the fact that everyone's expectancies are raised by Google.

The Internet has made us all librarians.
Along the lines of search, he brought up how complex Information Retrieval gets with more data and encouraged the removal of "rot" from websites along with proper representation of the content that remains. I felt like I heard this echoed indirectly by John Udell the next morning in his keynote, in which he reminded us that representations of content (think search results, archives, index listings, etc) should be "attention-friendly"—descriptive yet scan-able enough to give people (and their aggregators) the ability to know what your stuff is about in advance. Like the challenges for designers of Operating Systems, Udell points out that context switches are expensive for people on the web too.

Berrypicking, evolving search...
Knowledge workers learn via search. After the homepage, a site's search results page is the second most used. It's a big user experience factor, and often not enough gets done beyond simply displaying search results. Consider next steps for your searching user. Flickr clusters were brought up to illustrate a closing point (that they sort of turn folksonomies back into taxonomies— only better, user-created ones), but it made me think that they're a a good example of processing a query in a more helpful way.

Usability is a loaded term.
"Usable" has become a word people use to mean "better" or "more useful than before" without really articulating clear objectives. Morville walked through the User Experience Honeycomb, a model of the elements user experience. It's well examined here.

I could keep going, but I won't right now. A lot of great ideas were brought up in a mere 45 minutes. Other O'Reilly folks were scattered around the room including Bruce Stewart, who I just noticed posted his great take on the talk here. Hopefully these ideas will get to bubble around OPG . They're all relevant to what we do.