RDF documents in addition to regular web pages. Suddenly documents aren't just a list of boring tags or text, but a pathway of clickable links you can follow to your heart's content. (And, with later versions, you can even edit some of the fields.)
One can imagine tools like cwm and Tabulator sitting behind the applications we use every day, enhancing them with knowledge drawn from the wider Web.
For that's the real idea behind the Semantic Web: letting software use the vast collective genius embedded in its published pages. Think of all the places software uses APIs or databases: your spellchecker queries a website to find the definition of a word, your addressbook does a search to see if your friends are online, your calendar downloads a page to keep you posted on upcoming events. Now, imagine these programs weren't limited to one particular site, but could draw on the intelligence of the Internet at large.
Your spellchecker can suggest related or alternate words, or just keep up to date with the latest slang. Your address book can tell you where your friends are right now and what they've been up to lately. Your calendar can keep an eye out for events you might be interested in.
It's easy to make fun of these kinds of visions. My father, upon seeing such demos, always used to ask, "But why does your toaster need to know about stock prices?" And perhaps, ultimately, they're not worth all the effort. But the Semantic Web is based on bet, a bet that giving the world tools to easily collaborate and communicate will lead to possibilities so wonderful we can scarcely even imagine them right now.
Sure, it sounds a little bit crazy. But it paid off the last time they made that gamble: we ended up with a little thing called the World Wide Web. Let's see if they can do it again.
