It may not suit your goals, but one approach is to enroll in a masters program before entering a doctoral program. This could help you get back into the groove of academic life,
and also give you a chance to meet new professors who could write letters for your application to a more high-powered doctoral program. (I once advised a student who had spent quite a long time, maybe 8 years, in the software industry before returning to academia, and
this is the route she took. I think it served her well; because of the masters, which involved a mixture of coursework and a small thesis, she was very solidly prepared for her doctoral work, and was one of the strongest students in her cohort.)
I am going to give a non-serious answer, but there is a serious point behind. Lots of current and former mathematicians and theoretical computer scientists excelled in other fields. Here is my top 10 list:
1) You can change your name and start selling puzzles.
2) You can drop out and start a company (this or that, whatever).
3) You can start a hugely successful hedge fund which will in turn employ over a hundred other Ph.D's.
4) You can write a popular book explaining why people can't count.
5) You can write a comic book, a very good one.
6) You can write three volumes of a proposed seven volume monograph, get upset over its print quality, invent a new way, write a manual on it, and sell these and other books in dozens of languages.
7) You can design a bomb.
8) Why stop on a bomb? You can move to NJ and design a computer.
9) You can start a company selling in bulk a number-theoretic algorithm accessible to undergraduates.
10) In the good old days you could become French Minister of the Interior, but it helps to be friends with an Emperor.
UPDATE: while writing I discarded a few other career choices which I felt were somehow "less relevant", such as Iraqi Oil Minister, Russian oligarch, or even World Scrabble Champion.
Best Answer
I've found JabRef, http://jabref.sourceforge.net/ , to be the best tool (for me) for organising a BibTex database, and linking / organising all of my downloaded pdfs.
The biggest advantages of JabRef are that it is genuinely cross-platform, being written purely in Java, and that the native format for the database is BibTex, which means that if I decide to change my methodology then I can straightforwardly use my database with other programs. Storing the database as a BibTex file also means that I can directly edit the database, e.g. doing search and replace with an editor.
It has all of the functionality I need, and in particular it is possible to apply labels to each paper for categorisation, either manually or via keyword searches.