Posted by josh at February 2nd, 2008

Last November Bruce Eckel (the “Thinking in $FOO” guy) gave a speech claiming that 5% of programmers are 20x more productive than the other 95%. Some people flat out denied that was possible. I think it’s a mistake to think of “20x more productive” as “well, Billy wrote class Foo in 3 hours so I guess a superstar programmer would have written the same class in 9 minutes.” It’s not like that at all.

To me, what makes that 5% so bad-ass is that they write code that the mediocre programmer will never write, not if you let them work until the heat death of all the stars in the sky.

I want to take a moment and recognize a few of the people I consider to be the most bad-ass programmers around. This will necessarily be people whose bad-ass work is open source.

  • Fabrice Bellard. This guy is an animal. He wrote QEMU, an hardware emulator that achieves near-native performance by using dynamic translation (translating code from one instruction set to another on-the-fly). He’s won the obfuscated C code contest twice, once with a C compiler capable of compiling itself (in less than 4k, according to the rules of the contest). He later adapted this compiler to a boot disk in which the boot loader loads and compiles the entire Linux kernel (in approximately 15 seconds on modern hardware) and proceeds to boot from it. Also, though its not exactly computer programming, in 1997 he discovered the fastest known algorithm for computing an arbitrary digit of pi.
  • Julian Seward. He wrote Valgrind, which started as the best memory debugger money can buy, and has grown into far more. He also wrote bzip2, as well as GHC, the most popular native code compiler for Haskell.
  • Mike Pall. You want this guy working on your programming language. He created LuaJIT, one of the best and lightweight JITs available for any dynamic language today. And his plans for where he’s going to take LuaJIT in the next year are going to put Lua so far ahead of other dynamic language’s implementations that it hurts.
  • Linus Torvalds: this one is almost too obvious to include, but I have to mention it because with Git he showed that Linux wasn’t just a fluke. He didn’t just get lucky by being in the right place at the right time. This guy’s the real deal.
  • anyone who has ever won the Obfuscated C Code Contest. Maybe not the must useful code ever, but what these people manage to pull off in less than 4k of code impresses me beyond measure. Special mention to Brian Westley, who has won it nine times! My favorite entry: A letter from Charlie to Charlotte.

Guys like this singlehandedly turn out code that 99.9% of programmers will simply never write. More time won’t help, and putting them into teams of people all working on the problem certainly won’t help. There’s just a very small subset of people who will ever create such brilliant solutions to such hard problems, and I have nothing but respect and awe for people of this caliber.