@matejcerny@witter.cz
I used to think LLMs would make #Python and other mainstream languages dominate forever. But if AI is writing the code, you don't need a language just because it's easy for beginners or hiring. You don't even need to fear refactoring anymore.
What you really need is a language that catches the AI's mistakes. You need strong compilers and strict types (like #Rust or #Scala) to act as safety nets when the AI messes up.
This article changed my perspective 
Alexandru Nedelcu
Boring Tech May Become Irrelevant TechOn programming and personal projects
First, it is based on false premises like "the cost of refactoring goes to zero", etc. Second, it is just wrong.
Assuming that AI would be used more for writing code at larger scale (this assumption would not necessary hold, we'll see) you need a readable language because you need to review the AI's PRs. If AI would be writing "10x" more code, you have 10x more code to review.
Type checking and good linting is always helpful, but for a different reason. C'mon, if you don't believe AI can write a syntactically correct code, or get the types right, how could you trust it to get the logic correctly..? That's the harder part that you need to verify, so you need a easily readable language (especially for juniors) for that.
If you write a code in a language that has good static analysis tools and type checker, like Rust, it is often the case that "if it compiles, it works" because you start with good understanding of logic, so when you translate it to a coherent code, it auto-validates itself to some degree. AI only mimics reasoning, so you can never trust it in getting the logic right. It can easily write a syntactically valid code that does not have any sense, from my experience, it very often does.
