I think the implication of this article is that fast = reckless, but I disagree with that. Sometimes you recognize that you need more data points, and writing something that works will move you closer to the correct solution, even if the first version isn't perfect.
But yeah, it depends on the problem domain. If you're asking me how to build a website, I can probably have a pretty solid idea of what that code should look like. It's not a novel problem, usually. If you're asking me to build, I don't know, a voxel renderer and you're trying to figure out if an octree is the right way to go for representing the data, or you need to do something more exotic, I'm gonna have to play around a bit.
But yeah, it depends on the problem domain. If you're asking me how to build a website, I can probably have a pretty solid idea of what that code should look like. It's not a novel problem, usually. If you're asking me to build, I don't know, a voxel renderer and you're trying to figure out if an octree is the right way to go for representing the data, or you need to do something more exotic, I'm gonna have to play around a bit.