>3.of course user can run first exe independent of windows scheduler as option.
>
>this solution is decent and any search is " like instantaneous" one

Yup, that's pretty much what I'm doing with the current version of the screen savior - and the main reason I'm not distributing it.

There's a larger question (as Al wrote) about what you want to scan for, not only in terms of speed, but actually the meaningfulness of results. Imagine searching for words like "setup", "display", "update", "helper", "task", "edition", "starter" etc... There are hundreds of files with those words in the name in windows folders. Include them or not?

I've found even worse sources of false positives. Just this week, in a search inside Jira, I was looking for a case dealing with image thumbnails, to see whether I should create one. Result: yes, I should create one, and yes there are already 150 cases with that word inside... because the idiot who wrote the search didn't exclude tags from the search, so any word mentioned in markup, but not in text, will still be found.

So some selection of what to scan and what to skip is necessary. I prefer it the way Launchy does it - you create a set of folders to scan, and types of files which to scan for. Everything else is skipped. It scans in two occasions: on login and when you open the options dialog and click the button. Which makes sense: you will remember what you did since the last login, the MRU lists will still contain the filenames, so these will be at hand. Anything older will be in the tables.