But the one constant for the past four books has been an ingenious piece of software called Devonthink, which is basically a free-form database that accepts many different document types (PDFs, text snippets, web pages, images, etc). It has a very elegant semantic algorithm that can detect relationships between short excerpts of text, so you can use the software as a kind of connection machine, a supplement to your own memory. I wrote about this several years ago for the Times Book Review, and I still get emails from people every couple of weeks asking about the software. (The Devonthink guys should put me in an infomercial.)