One thing I look for during an interview is does the candidate use Google as part of their answer workflow. While having a candidate who just "knows" the answer might be great, eventually they wont, and what then? I'd rather see how a candidate goes about finding an answer to a problem, and part of that is looking it up. Why would LLM be any different? The real test is: do they know the difference between a right answer and a wrong one. That's what I'm interviewing for.
Ironically all the top tech companies instead require you to have the answers for ~50 leetcode questions/concepts already memorized. You then pretend this is the first time you ever saw a problem like this and "work your way" to the optimal solution.
