I'm not convinced of this at all (this is the first time "citation please" popped into my mind, and I hate that response :). Of course, there are some problems that have a strong need for erotically background, but those positions tend to require a graduate degree.

On the other hand, I've seen self-taught engineers who have far more passion for developing software, which seems (in my experience) to correlate more with engineer productivity.

I have a CS degree from an institution that has an extremely strong theoretical computer science department. I value my degree and the knowledge I gained to get it, but, beyond getting me in front of hiring managers early in my career, I don't think it particularly made me a better developer. A little understanding of algorithms and data structures (much of the need for which is obviates by today's VMs and libraries) and some understanding of what is really going on under the hood (again, not as helpful in the VM word).

For me, a degree is a nice-to-have, but you won't get hired if you can't convince me you are a good learner.

DYAC? I'm sure you meant "theoretical" background there :)

I had a feeling I should have proofread that one. The small box makes it annoying, though.

I totally agree that an attitude of self-teaching makes better developers. There is a lot in the application of CS, but I would call that Software Engineering.

More generally, having seen more kinds of software kind of broadens one's way of approaching programming problems, so even if you don't directly use anything you see when you study operating systems or compilers or AI, you can grab vague approaches and ideas from those fields.

If you're interested in writing performant software, you'll care about big-O, algorithms, and data structures. Understanding algorithms and data structures enables you to intelligently choose and apply them even if you don't have to develop them from scratch.