> At some point, execution speed starts to matter more than the elegance of the code.

Having seen bad (human) code behind award winning products, and good (human) code take so long the investors get cold feet, yes.

Some people claim (or seem) to know which corners can be safely cut. But what I've seen suggests those who cut corners got lucky, rather than using skill to know what could be safely ignored.

The other side of this is that I have come to view things like "Clean Code", "SOLID", "VIPER", and use of mocking in unit tests, the same way as self-help books: survivorship bias.

GenAI in code is likely to give us enough feedback fast enough that we can turn the survivorship bias of SWEng self-help guides into actual science; but unless progress stops suddenly (could happen any time if investments stop), humans coders aren't likely to be the beneficiaries of this research.