If you haven’t tried it yet, Cuil is supposed to be the next hot search engine. So I tried it while fixing a problem with a Domino server, expecting a much quicker search experience than was the case with dear old Google. Here, “Quicker” means less junk results and more useful technical snippets – something I find one can get, generally, by using more than three search terms.

So imagine my surprise at following cuil’s first match result to my search : A page solely concerned with showing you that the guy writing it can’t diagnose the source of his problems, no matter which bit of software he’s looking at. It has some random and unsubstantiated comments about mail rules: it contains no examples, no code, no fixes – just a raft of inaccurate statements about fossil versions of the product that nobody uses any more.

Why, one wonders, does Cuil rank this page higher than the tens of thousands of pages retreivable from IBM’s documentation and support forums? On one machine I tried this search with, all I got back from Cuil was 2 results – the Computer Gripes one, and another one from openntf (which is at least apposite, if not well chosen). Only by turning off the safe-surf filter, did I get any of the other common resources for Notes agents, rules & security information.

Credibility in search technology is a perennial problem, as the googlewatch people will tell you: but this strikes me as a repeatable example of downright odd results.