I think the author is confused.<p>> Writing cool code in obscure languages or rewriting industry-accepted libraries is great as a hackathon project or when you’re at Facebook-scale, but it’s dumb and dangerous otherwise.<p>On the face value, this is quite agreeable. But what is "cool" code? What if someone just writes "a code" in an obscure language - because that language happens to be the best for the code? As for "industry-accepted libraries", those libraries can't be considered the word of almighty while everybody else agrees we have a crisis in the area of software engineering.<p>On the other hand,<p>> There’s a time and a place for innovative custom-written code that advances the edges of engineering and computer science.<p>Again, well said - but what is "innovative"? Solving the same problem for the hundredth time - but doing it well while explaining why - can or cannot be innovative, depending on the definition. Yes, there is time and place for that - so should it be thrown away in a young startup just because it should concentrate its efforts elsewhere?<p>The whole idea of the article is great - the code is written over and over, with small variations because of platforms, infinitisimal changes in requirements etc. Indeed, engineers first learn to produce linked lists, then taylor them for their particular problems - it's like an art of pottery, when the next customer has slightly different requirement because, say, his shelves have different width. But the blame seems to be put to the wrong place.
Industry accepted libraries tend to get bigger and bigger and slower and buggier with lots of dependencies on other libraries with their own set of dependencies.