I think more can be done because of all the browsing history etc, letting people refine what they are looking for when they open a new tab on the blank page itself, based on the places they have visited etc. I guess the additional challenge is how to do it in a slick way so that it is not seen as ugly and intrusive.

I have been trying a few just now, and in a word: no. They aren’t smart enough and,mainly, they require a lot of pre-configuration.

Instead, form filling should take a few lessons from the awesomebar:

1. Use data from forms on all visited websites to give suggestions. So if I have ever entered “David Naylor” into a box it will be suggested when I want to enter it (and start typing “D…”) on a website I never visited before.

Using all the saved entries as suggestions will give us a problem of very/too many suggestions. This can be countered by the following improvements:

2. a) Count form entries per site: When visiting Gmail, keep track of the fact that I’m entering my own user name almost every time and put it at the top of the suggested list (even before I type anything). Then stick my girlfriend as number two, etc. (Currently entries are listed in the order they were first entered, right?)

2. b) Count form entries across sites: When I visit a new webshop, understand that I’m most likely to type “David Naylor” when I start typing “D…” into a form.

3. I guess, ideally, Firefox should understand whether I’m about to type into a “name” box or an “e-mail” box and use that knowledge to bring up the relevant suggestions quicker/further up. This function would have to be implemented with non-english websites in mind though.

4. When asking whether to save a password, give the user the option to “log in automatically” (fill login + password and submitting) in the future.

This would be like Speed Dial minus the tedious manual configuration. And you could even start to pre-load the most frequently visited pages with a clear progress indicator (if the users’ line to the internet is not used (closest approximation is if the browser itself is not downloading anything).

Other use for this would be to pre-fill those annoying ‘certify your age’ inputs. But they’re usually found inside Flash video players anyway, so though luck with this one.

Another thing, that would be great to see is contextual history searches. Imagine this scenario:
* (optional step) the user (me) has the web history set to 365 days ;)
* the user visits a page and asks the browser if he doesn’t know about any already visited pages, that match information about the topic the currently displayed page presents …
This could be done by indexing some content on the page and storing that in the browser history along with the URL.
You could take your imagination for a spin and project this concept one step further and you get:
– the user visits a page and asks the browser if some of his favorite (=frequently visited or bookmarked or starred or whatever) sites contain some information the currently displayed site does.

@Thomas, Mike: The constant fight for bubbling up more information is information overload. Showing the “flows” of the web are interesting, but doubly so if it can help reduce the information overload problem. Any thoughts on how to do that? My hunch is to use that data to augment what I’m already seeing&mash;like making links that my friends have clicked on look more bold, or brighter, etc.