These artsy fartsy takes on a subject that would be better studied as a science are fun but are they useful? I am not saying there is no such thing as taste and neither will I attempt to define taste except to say "tasteless" is probably the opposite. "Judgement" seems more valuable as some have mentioned, maybe judgement and taste together identify something we want?
Still, I like better when we actually do try to name what we're talking about. We've probably all read the motorcycle guys take on quality and things like that right? I guess the theme some are going with is that quality can't be objectively quantified and to define it is futile but I've never liked that take. If we think LLM generated code is low quality then it certainly makes sense to start our attempts to define or at least describe what makes the quality needle move, because language is how we encode lots of information and information is what feeds LLMs.
People also like to rag on uncle Bobs takes, clean code, etc but I like how at least he tried. I've been a fan of his for awhile but I'm not big on objects so don't care about his pushes on that sort of thing. The best take that offered hints that I would call quality is "A Philosophy of Software Design" by John Ousterhout, there is plenty of stuff in this short little book that was very satisfying to me, like for instance where Bob wants short/small functions, Mr Ousterhout was much more flexible on this, like for instance maybe some functions can be bigger if the need arises, things like that.
Still, I like better when we actually do try to name what we're talking about. We've probably all read the motorcycle guys take on quality and things like that right? I guess the theme some are going with is that quality can't be objectively quantified and to define it is futile but I've never liked that take. If we think LLM generated code is low quality then it certainly makes sense to start our attempts to define or at least describe what makes the quality needle move, because language is how we encode lots of information and information is what feeds LLMs.
People also like to rag on uncle Bobs takes, clean code, etc but I like how at least he tried. I've been a fan of his for awhile but I'm not big on objects so don't care about his pushes on that sort of thing. The best take that offered hints that I would call quality is "A Philosophy of Software Design" by John Ousterhout, there is plenty of stuff in this short little book that was very satisfying to me, like for instance where Bob wants short/small functions, Mr Ousterhout was much more flexible on this, like for instance maybe some functions can be bigger if the need arises, things like that.