Andrew Quinn: Some Wheels Need Reinventing

Andrew Quinn says that when programming, it's easy to fall into guilt over the fact that the tool you're building right now might have already existed as a better implementation 30 to 40 years ago. For example, rather than building your own TSV-aware search-and-replace tool, learning awk would solve the same problem in one shot.

But he sees that kind of thinking as actually a trap. He says you need to reinvent a few wheels yourself to reach the edge of knowledge—not too many, and not too few.

In most fields, 4 to 5 is enough, but in rigorous domains like mathematics or computer science, he thinks you might need as many as 20 to 30. His claim is that the questions you ask while building things yourself get you to the real frontier faster than passive learning does.