sarupbanskota

Why do you think LLMs don't do a good job at generating good selectors by default?

Thank you! I _think_ we learnt that from CodeCademy who was among the first to introduce the pattern.

Eventually, we would love to. In the near/mid-term, it's unlikely.

The reason we specifically chose Rust is because:

* It's rarely someone's first language, so we'll get targeted feedback from users with the right experience level

* It's a somewhat non-intuitive language and has "new/unique" ideas (e.g borrows), so the pain point is particularly strong

* It lacks learning resources (my guess is because the number of Rust developers is relatively small and there's higher barrier to producing content)

It's a change we'd love to see taking place.

We are aware of tech teams that use our challenges instead of LeetCode style questions for interviewing. We run our interviews using this system. Candidates seem to enjoy the experience, and it's much harder to "fake".

There are two main challenges today:

* Interviewers themselves are used to the standard format — they have answers for, a way to benchmark, tools they're familiar with over years, etc.

* While much harder to "cheat" on these challenges (since there are so many touch-points), offering these as take-homes might be lengthy for some engineers.

If you were in our place, what would you try to do for increasing adoption (whether for hiring or learning)?