The NCSA Mosaic Demo Document, NCSA Starting Points, and CUI W3 Catalog all contain a wide variety of useful pointers. Periodically checking out the NCSA What's New usually proves interesting. For searching the Web, the best thing I've found so far is the EINet Galaxy. It's just way c00l.
Usenet FAQs are available in hypertext form. Everything from sex to cryptography to Star Trek. A database of high energy physics preprints at LANL is available and searchable. RFCs are also available. Various government documents are available including the National Health Security Plan, the National Performance Review and a searchable index of White House press releases. Cornell has a collection of Legal information. There is now a Linux information server. GNN has a wide variety of pointers to recipes. The Project for American and French Research of the Treasury of the French Language is an outstanding collection of French literature and linguistic data.
PERL! 'nuff said. It's The Late Show with David Letterman!
Wired Magazine, a truly outstanding and very cool publication, has put the full text of their magazine on the Web. Lego has a web server with a complete list of lego sets and jpegs of the assembly instructions for lego sets, including the lunar lander (#6985)
The Library of Congress is available online.
Hypertexted Books from Project Gutenberg are being made available by rgs@cs.cmu.edu. Lysator Etexts are very nice as well, including The Terrorist's Handbook.
LANL Physics Papers has a cool forms interface. Nifty.
Other sources include a WAIS index of all the White House Press Briefings. Movie Reviews from rec.arts.movies.reviews are available in a searchable index. While I'm on movies, you should definitely play with Cardiff's Movie Database Browser which has a huge amount of movie data with a spiffy forms interface and some neat features like this day in history.
Animaniacs are on the web. Are you pondering what I'm pondering?