You’re hired for your problem solving ability which includes being able to learn and apply specific algorithms if needed.

Asking questions that chat GPT can easily answer gives 0 insight into problem solving ability.

For 99% of the roles out there, needing to drop into algorithms all the time is an anti pattern that demonstrates lack of knowledge of your tech stack. I’ve had to clean up a lot of code written by people who wanted to show off their algo knowledge to solve a problem that should have been addressed with a few lines of SQL.

Sounds cool in theory, but the question stands

How are you going to improve existing codebase / new code if you aren't aware of the foundations, tools and basic primitives?

How many problems could be solved by just usage of z3 if you were aware of it?

How many problems could be solved by applying most popular combinatorial optimization methods/approaches

instead of writing something times, times slower, but eventually hacked enough to make it work?

You're assuming that that awareness of how to solve a problem and immediate hands-on-keyboard test for materially different success rates while actually employed. My experience is that by the time you've gotten the job, most of the crammed-in Leetcode specifics have leaked back out and you're in the same place as somebody who knows the thing from school or from using it in prior projects, but didn't have it in their fingertips during the interview--you're looking it up again, because you know what you need but don't have the details top-of-mind.

The insistence on being able to spew out code-first answers to contrived problems during an interview situation is silly gatekeeping, but it makes the people who pass feel good, and that's why it persists.