I've been struggling a bit with Google Instant - not because it's difficult or anything.
More with how it's really different from the way Google's been sort of working for a while, with its predictive suggestions etc (yes, I know ... but really?).
And with what I'll do with the two to five seconds I'll save per search - which could work out at around three minutes per day. Or about 0.75 of a pee (including the traipse to/from).
Flippant, of course. I do know that under the bonnet there's a lot of stuff going on that's mimicking human intuition rather better than it should and that there's also something seriously epistemological about it that's casually and callously redefining what I know and the way I know it.
At the practical end - and it's early days, of course (how many more qualifications?) - there've been a couple of interesting observations; and one scary worry.
The first from that cortex-bulging gang who think about search engine optimisation here at the BBC. The second from an old friend of the College who now works at the World Economic Forum.
As I say, it's early days. But, if you worry about SEO (and the SEO honchos have this insistent way of making you worry about it), you want to know that the world's favourite search engine (other search engines are available) is likely to drive traffic your way if you get your heads right.
I mean, that's the deal, isn't it?
But here's one thing: if you start typing, let's say, someone's nickname instead of their full name, Google Instant (GI from now on) helpfully starts reeling in the results based on the nickname. So does that mean the SEO honchos should start enforcing nicknames in optimized headlines? Too early to tell, I suppose.
Plus, the gaps in GI's knowledge are a bit well ... glaring. Try 'Montenegro'. Bit spare? Yup.
But here's the scary one - a tip from Prof Adrian Monck, formerly of London's City University, now of the WEF.
We all know better than we should, thanks in part to the wonderful worldwide web, a host of rumours about celebs - some so untrue and scurrilous that if you were to publish them the legendary ton of bricks would fall. Rumours at which I can only hint - for obvious reasons.
Now, in old 'Googleland' (other search engines are still available), if you were to type elements of that rumour into the box, you would probably have generated returns that took you to sites publishing, at their risk, that rumour.
But the whole point about GI is that, as you type away, the better matching results are pushed to the top - perhaps to the exclusion of all others and giving them a currency they might not warrant. And just a bit like publishing in its own right.
Plus, the predictive function as you type starts to take you down routes which are, frankly, libellous - again, because that stuff is out there and GI is making its predictions based on that. To the extent that you could, if you wanted to, argue that GI's predictions were to all intents and purposes repeating a libel.
Chuck into this mix the obvious elaboration of real-time search and you're into a world where, potentially, a short-lived, reputation-busting, defamatory bit of bile is, for a time at least, 'published' either as predictives or GI results.
Discuss - as my professor friend suggests.
