Super-star programmers

Agree -- also one of the most under-rated skills is someone who keeps up to date on all the latest frameworks / libraries, etc. More often than not, you're building something that solves a unique problem and you do that by stitching together pieces of other people's engineering.

I would argue that this presents several sub-skills:

Being able to do this quickly b/c you don't have to spend the time learning

Being able to do this effectively -- by not painting yourself into a corner b/c you well versed in best practices and anti-patterns with using said 3rd party technology

Being able to do this smartly -- because you are well-versed in a lot of the different options, you can effectively way their pros/cons, and make smart decisions on which 3rd party stuff to use, and when to build vs. buy

I think these skills are more effective (towards your end goal) than simply being able to quickly code some brilliant labyrinth of an algorithm in less SLOC than a 1 or 2-xer. I feel like the emphasis is always put on the latter, and these skills are overlooked.

Combine these with effective communication, and then you are talking about someone who can get things done.

You're missing the point that the term rock-star programmer doesn't mean they will produce world changing programs. It means they'll do an order of magnitude better/faster than their non-rockstar peers.

That might be a bit contradicting to what Joel is explaining. From what I've experienced first hand, there's programmers that can produce similar code/software in a week that takes 5 mediocre programmers to produce in a month. Those are extreme cases though.

There's not enough "rock star programmers" to go around, so not everyone can have a piece of that cake. Much of this is also very domain specific, it's not a given that a "rock star" game programmer is going to be a orders-of-magnitude better at developing business support systems. Finally, this is not that specific to programming. Not everyone's Einstein, Feynman or Michelangelo.

Well, in the linked article he used a Mozart/Salieri analogy, implying that rock star programmers write masterpieces, not simple bug trackers. If all Fog Creek programmers were truly rock stars, wouldn't they be beating Google instead of writing menial money-farm stuff? In my experience, those of us who move at a faster pace than our peers tend to get the most interesting projects. There's a reason that ERP software is so bad, after all.

In the rock star analogy, it'd be like getting the Beatles to write music for Justin Bieber fans, or getting Metallica to switch to the "easy listening" genre. That's not what you do with rock stars.

It’s interesting that in the research from Peopleware mentioned in the article, the factor of 10 difference was mostly between different organisations, while individuals within an organisation were typically quite close in performance.

This suggests to me that the biggest influences on productivity for the organisation as a whole are probably to do with collaboration, culture and environment rather than extraordinary individual skill.

Maybe the very best programmers aren’t extraordinary just because they can solve the most difficult problems, but also because they are better at presenting their solutions in ways that are more accessible to other developers. That would have the twin benefits of improving productivity in its own right and of implicitly teaching by example so less experienced developers in the organisation would tend to develop the same skills themselves over time.

If you had a seed of programmers on that level and a culture that promoted mentoring, pair programming, code reviews or similar collaborative exercises, it’s not hard to imagine that such an organisation would develop significantly higher collective performance over time than another organisation without those benefits, even if most people joining each organisation were of a similar level of skill and experience to start with.

I think it is always important to consider time-lines on stuff like this. In this field (many fields actually, but it seems especially this one), the first version of something may be revolutionary or particularly good, and then everyone copies that as obvious, making the first one seem only OK or even bad. TV-Tropes covers this effect nicely in the world of fictional television at: http://tvtropes.org/pmwiki/pmwiki.php/Main/SeinfeldIsUnfunny I'm not sure the timeline on FogBugz, but it wouldn't surprise me if this was a case of it.