Do developers genuinely believe they produce better code than they do? Would their claims stand up under scrutiny, or is it a simple manifestation of the Dunning-Kruger effect? Given the perceived intellectual demands of coding - particularly green field coding, it fosters a strong sense of self-confidence and personal affinity with the output. I guess anyone that's labored over a body of code and got it to work has felt that.
But if we take a moment to reflect on the vast majority of the lines of code produced by developers today, would we not find to be a rather mundane mix of RESTful APIs, CRUD functions, form validations, or CSS wranglings (repeated ad nauseum) rather than anything new, novel or groundbreaking in the way that say Hoare's QuickSort or various Traveling Salesman solutions are?
Code created by generative AI seems to elicit similar reactions to developers asked to take ownership of another developer’s work. They invariably describe it as poor, usually with comments like "I wouldn't have written in like that!", followed by moves to push for refactoring, likely to realign it with their own world view (until the next handoff, in a series of "rinse and repeat" activities). But clearly, the code was good enough and fit for purpose at some point in it's lifetime?
So it begs the question, is generative code quantifiably and qualitativly as good as human code? It certainly seems to be, and it as verified in numerous benchmarks, even if you ignore the truest-me-techbro benchmarks) Perhaps our new problems are not in code generation itself, but in the intellectual atrophy it could create over time as talent developers try to deal with the sheer volume of generative code output. This splurge amplifies the demand on time, leading to nothing new or novel being created until such a point where true innovation grinds to a halt and everything new simply becomes a new version of something that already exists. Grim!