DevonTHINK -- is this useful?

http://www.nytimes.com/2005/12/04/business/yourmoney/04techno.html?th&emc=th

More ink about DevonTHINK today in the NYT. I go back and play with the program some more every time it pops up in the press, as it seems like something that would be right up my alley, but it never seems to make much sense to me. Today, I thought I'd actually try to figure it out.

It's simple enough on the face of it, anyway. DevonTHINK is a keeper for snippets (of all kinds). It allows instant searching, cross-referencing, even a sort of associative clustering, within the collection. Steven Berlin Johnson was eloquent on its powers earlier this year:

http://www.devon-technologies.com/files/reviews/nyt%202005-01.pdf

People report some overlap with EndNote, the bibliographic management tool that humanities grad students use to get their dissertations wrangled. I believe the difference is (correct me here) EndNote is for just one kind of entry: bibliographic. DevonTHINK can do text, PDF, doc, RTF, images, word, etc. It also has a web browser built into it, enabling it to directly reach out and grab things from the web (and save them off into its collection).

Also, EndNote has only a workaday notion of keyword search. DevonTHINK is fancy: its search is informed by semantics, meaning that searching for "dog" will turn up documents that mention "canines", or even documents with words that frequently co-occur with "dog", such as "tired." This sort of information retrieval style is not new, but it is new and cool in the context of the desktop computer and a personal collection of files.

But DevonTHINK carries what looks to me like an albatross. As near as I can tell, the way DevonTHINK works is to keeps its collection of files in one big file, separate from the original versions of things. If I put a text document into DevonTHINK, the program stores a copy of it. And if (nay, when) the original document changes . . . the change is not reflected in DevonTHINK. I must go back and re-add the file to the collection. DevonTHINK does not keep the 2 versions in sync!

If so, there's little point in putting, say, an email archive into DevonTHINK, because an email archive grows and changes all the time, and DevonTHINK remains unaware of growth and change. Likewise for new documents I'm working on; the new IM chats I'm constantly generating; new files I'm downloading; new podcasts; new videocasts; new songs; . . . if the user doesn't manually put each and every item into DevonTHINK, as it changes or appears, the database has holes. So now, to find things, the user has to search in two places now: inside DevonTHINK, and outside. Inside, cool associative semantic search. Outside, none.

And if that's so, then DevonTHINK seems more or less useless for realtime information management -- instead, it's really best suited for managing a specially designated, static library of stuff such as Steven Berlin Johnson's carefully compiled archive of research articles and snippets. (Which he has a research assistant help him with, by the way.) Such an archive must be manually built and pruned, when one has the time, and when one remembers to, and if one has the interest to even get started.

What DevonTHINK sounds like to my ears, but turns out not to be, is EndNote on crack, for the whole hard drive and for all the files. And I would indeed like that. Also, it would always be on, requiring no investment from the user. When I search for "Christmas" amongst my stuff, I'd like pictures from Christmas time, chats that mention "xmas" lists, brand new Christmas songs from my library (even ones that don't explicitly mention the word Christmas), a Christmas slide show from long ago (again, not mentioning Christmas explicitly, but containing pictures that are Christmasy), recipes for Glogg (might not mention Christmas), calendar entries around Christmas time, the text of "A Christmas Carol" as well as the PDF sheet music for some carols, caroling, etc. But not Adam Carolla. This all is pretty low-hanging fruit, but hard stuff to get at with keyword search. Mac OS X Tiger's Spotlight search might get half of these items -- the most literal search hits. But there's no semantics, and thus, no associative power.

Perhaps what's really needed, to scratch the itch that DevonTHINK begins to scratch, is an associative search tool that works on top of Spotlight's API. The democratic approach would, in this case, actually work better: there would be no special database, no walled garden maintained by some proprietary tool. Spotlight is built into the OS, and it already has a complete index of every file; what's really cool is that it updates the index instantly every time any file changes. So Tool X, by invoking Spotlight and re-processing and expanding its output, would work in realtime, and it could provide all kinds of associations by theme or term or similarity. Even throw in some clustering and a little visualization for good measure, so I can drift through my files purely on the basis of semantic connections . . . .now the filesystem could actually start to look a little different than it has since 1985.

For Windows, a similar thing might be done on top of the Google Desktop Search tool, if their API is accessible and has its hooks deep enough into the OS.