After FFmpeg (used in all your TVs and gadgets, very likely), QEmu (used also by Xen and VBox and other), tcc and his IOCC entries, the DVB-T emission with an ATI card, Fabrice comes, once again with something crazy...
Don't forget that he discovered the fastest known algorithm for computing an arbitrary digit of pi. If you ever are tempted to get egotistical, just think of this guy. If you are prone to low self esteem, avoid reading about him.
It isn't so much the "normal" company thing, I think, as having a good work/playthings/life balance. The main reason none of my personal/toy/experimental projects are languishing on note paper or electronic equivalent is that my balance is very wrong and getting worse!
He is presumably working for a good company (or at least one that is good for him) and has time management skills himself.
That, and a brain to be envious of plus a supply of inspiration!
OTCC is an Obfuscated Tiny C Compiler for i386-linux. It generates FAST! i386 32 bit code (no bytecode) and it is powerful enough to compile itself. OTCC supports a strict subset of C. This subset is compilable by a standard ANSI C compiler. OTCC compiles, assembles, links and runs C code without the need of any other program.
Great hackers can use whatever languages/tools they like. The fact that they're being great is not the consequence of the tools they chose, but of the masterpiece work they achieved.
I am always fond of such an analogy: in terms of efficiency, the greatest hackers has an algorithmic complexity of O(1); meanwhile, the majority of us may be O(n), if you can manage to get a O(log n) you can make into the club of good hackers. The tools, be it OS, programming languages, etc, are only a constant coefficient to that complexity, i.e., they can make you noticeably more efficient, but they won't improve your "greatness".
Close but I think you have your orders understated. A normal programmer faces O(n^2), while a great programmer achieves O(n log n). Really bad programmers do O(k ^n) amount of work to achieve the same, with k>2.
In the days before KVM, he wrote a paravirtualization kernel module for QEMU (kqemu), gave away binaries and tried to sell the source. I don't know what happened with that business.
According to the Wikipedia page, you need to be a US citizen or resident to be eligible. He seems to be residing in France and I doubt he has a US citizenship. Otherwise, he definitely fits the bill.
He is really impressive...