What Level Programmer Are You?

What are some highly specialized niches that I could look into for consulting?

I actually don't care that much about money, in absolute terms. I care about maximizing my free time and maintaining a somewhat decent standard of life. Ideal scenario for me: a high hourly wage + the freedom to take on work in whatever chunks I prefer.

More on this: http://esr.ibiblio.org/?p=316

Similarly, if I were at the top of the scale, that would show a lack of imagination on my part, and also make the scale worthless in a "I won at a game where I just wrote the rules" sort of way. Personally, the territory that's most interesting to me is 1.7 to 2.2, because that's what I intend to be covering over the next 5-10 years.

People tend to estimate themselves as average within what they consider their peer group when comparisons are difficult or impossible to make (and the fact is that it's really hard to compare two programmers). So I think the 95 comes from the fact that it's "common knowledge" that 90% of "programmers" are incompetent.

Of course, these percentiles depend on a population and competence distribution that are by no means immutable, which makes them meaningless as technologies evolve and more people come into software, so I needed a scale for professional development of an engineer based on general knowledge and approaches. The actual 95th percentile is probably around 1.5 on that scale, but that's just a wild guess.

I'll still use "5-percenter" (loosely; no clue what the actual percentage is) to describe an elite set of programmers, but that has more to do with attitude, bearing, and trajectory than present-time skill. For example, the guy who picks up SICP because he's topped out and bored at his Java job is probably only 1.2-1.3, which is not yet 95th-percentile, but he's a "5-percenter" because he has the attitude and has found the resources that will make 1.5 likely and 2.0 possible-- which wouldn't be the case if he stuck with his same-every-day enterprise Java job.

I know scientific programmers who still use FORTRAN and write incredibly fast programs in a domain that's much harder to get things correct than you think.

On one level these people are "geniuses" but many of them would be completely lost if it came to object-oriented programming or relational databases.

As for the computer scientist, he's an odd animal.

I remember a time when I got a program that did something almost magical from a guy who's a household name in machine learning and CS instruction. It turned out this program crashed before getting into main()! On close inspection the problem was that he was allocating a huge array statically that he never used! This worked OK on the 64-bit monster he was using, but would kill any 32 bit machine.

My experience is that code written by computer scientists is very rarely suitable for commercial use because computer scientists usually aren't professional programmers. They get rewarded for writing papers instead of writing programs, so a succesful endpoint for them is producing some artifacts that would demonstrate how the program would have worked if it really worked.