Probably, this is not first time we talk about this, but I think it's not necessary to compromise code quality for speed of development.

For instance, although I don't strictly follow TDD, I often find that having tests in place is indispensable - they truly save time rather than waste it.

When a system is built with thoughtfully designed parameters, maintenance becomes much more manageable. It's often simpler to work with a clean, parameterized approach (like using X+Y) rather than dealing with a "throwaway prototype" such as 143+235. Building on a composable platform is far more efficient than navigating through heaps of ad hoc code, even with AI assistance.

While there's no need to paint a Falcon rocket, it's crucial to ensure that differential equations are solved correctly.

Programming is about knowledge. The less (syntactic, solution-domain) clutter your code has, the better, the faster you can go.

Unfortunately, software development is understood wrong, where code (and amount of it) is the Holy Grail. Now with GenAI it's even worse: it can generate huge amounts of mindless code quickly. But at the end of the day it's just knowledge, which matters, and ability to "compile" it to working solution.

As for optimization - why assume that something if born not optimized? If it's a habit.

PS. Thanks for the git-of-theseus hint! I've read for example this https://pragprog.com/titles/atevol/software-design-x-rays/ book, but never really tried any tools.