Earlier quoted context omitted.
I don't see how expecting candidates to know algorithms by heart or solving complex algorithms is useful. Whenever I encounter that at work I research the topic extensively and learn or re-learn whatever it is I need. I completely agree on testing on real tasks, given some candidates who perform terribly on those manage to squeeze by (even thanks to them passing leetcode bs interview without having real world experie…
I think I broadly agree for most programming jobs. Algorithm problems are based on the philosophy that if you can implement a binary tree in 20 minutes, you’re smart enough to figure out just about anything else that comes up. They were popularised by Google, who hire with the goal of never hiring incompetent people even if it means missing out on some good people. There’s two problems with algorithm problem intervie…
You can probably learn more by wrapping it as a tiny 'project' note: not a take home -- just a DSA/LC style 'problem' diguised inside a git repo. I'd rather see 2Sum with a Cargo.toml in a repo, seeing them scaffold this in the lang of their choice, or within an existing codebase.
Justifying the full end-to-end "how is this going to be deployed and maintained", "how would you test X" is a bit more informative than "please repeat Djikstra's algorithm into the void".
Of course, being good @ LC in isolation is still a proxy for (math) intelligence perhaps unpopular - but in SWE being smart isn't enough, you gotta grift, ship or hustle your way through; thinking is mostly optional.