My default keywords for testing searches are, you'll not be surprised to hear, "badger" and "badgers". Now, bear with me, because I know I verge on the obsessive when it comes to the little stripy beggars, but "badgers" is actually a very good keyword to use. A search for "badgers" usually returns enough hits to judge the organisation, ranking and deduplication of results, without returning so many millions of results that you're overwhelmed. When I was working for ProQuest on their Literature Online database, I could have told you all the references to badgers in English Literature from 600AD to the present day (roughly). As an area of expertise it's of limited use, but it breaks the ice at parties.
In a general web search, though, the term "badgers" will return results in a variety of different areas, from useful factual sites about actual real stripy snuffly mustelidae to sports in Wisconsin, from Flash movies (and other silliness) to things that are just named after badgers (thanks to brrm for pointing me at another of these). Then of course there's "badgers" in the sense of "those who badge", which is a whole different kettle of badgers and fish.
And this is where Grokker comes in. dead badgers". Then there's the helpfully-named category "More...", whose subcategories read like the options in the "Dove from Above" round on Shooting Stars: "Contains Information", "Hockey", "Guide", and "Mount". (Miscellanies may be an easy target, but that doesn't mean they don't deserve a bit of a kicking occasionally.) Some of the categories are more useful, though; the "General" section breaks down the, well, general badger-related information into subsections such as British, Austrian, Canadian, Network, Nonprofit... and it's nice to have "American Badger" hived off into its own section so I can safely ignore the impostor (though I have to remember to ignore it under its Latin name, taxidea taxus, as well).
I can't help feeling that it's definitely more of a toy than a tool; the taxonomy (no pun intended) is sufficiently inconsistent that the system works better as an idea-gathering tool, a browsing device, than a way of targeting a search. The bigger the section, the more likely it is that it'll have one or two peculiar anomalies; conversely, though, a set(t) with only one member is likely to be a less-than-useful category which could have been included in another. But then, this sort of grouping is always going to be subjective, and it's always going to work better (on average) with average data. What would be really interesting would be a system that presented the results in this format, allowed you to regroup them according to your own way of looking at the world, and learned from this. Fortunately, such a system has already been developed over several thousands of years; it's called the human brain, and the invention and application of categories is something it's damned good at (almost as good as it is at arguing about the categories, and sending "poke the heretic with sticks" signals to our stick-poking apparatus when other people disagree with it).
The interface is stylish, if you like that sort of thing; all spots and squares of different sizes and colours on a dark background, zooming in and out like Paperchase meets the Matrix. It's not as neatly and cleanly information-packed as the BBC's election map, nor as stubbornly focused on style over content as Warp Records' "ography", but it strikes a good balance between pretty and practical. Loading time is (understandably) noticeable but certainly tolerable; navigation is smooth (both visually and conceptually). The actual visual representation tackles a difficult problem competently; that a larger circle should mean a set with more members and a square within a set should be a single page/site seems (I hesitate to say) intuitive, or at least coherent. It's less immediately obvious to me why some of the spots appear as 3D spheres while others appear as flat circles; but once I realise that it means there are even more sub-categories, it kind of makes sense; each sphere is another whole world (or at least a small satellite) of information. However, in the context of this (fairly loose) metaphor of zooming in on planets and areas on their surfaces, buttons marked "top" and "back" are slightly discordant -- yes, I work out what they do quickly enough, but only because I'm used to "top" being used in the context of directory trees and "back" being used to go back to a previous state as well as a previous page. (I'm also used to relentless mixing of metaphors and cross-pollination of contexts and registers.) Going up a level can also be achieved by clicking on an empty area in the circle you're currently "in", too, which is more consistent with the information space that Grokker seems to be trying to create. To the right of the Grokkerverse, there's a side-panel for previews of pages; this is a great idea, and cleanly presented, though if I wanted to nit-pick (imagine!) I'd note that in terms of look-and-feel it's not very well integrated with the rest of the navigation.
Despite all my criticisms, though, I like what they're doing; attempts to produce visual representations of the way we might organise our minds are always interesting (at least, they're filed under "interesting" in my personal set of mental categories... perhaps we need a meta-category of "disputable categories") and this one's not only interesting but, on the whole, well-implemented. It is "a new way to look at search", appalling tagline though that may be; its mission statement is "Broad exploration, unexpected discovery, deep understanding", and it definitely does its bit on the first two of those. This is still only a demo version; perhaps deeper understanding will come with time. In the meantime, I'll stick my neck out and say that it's not going to revolutionise the way we use the web ... but it may be a boon for bloggers when it comes to fishing quirky links out of the memepool.
Finally, as promised, a cute picture of a fluffy animal. Have a nice day!