I wonder the same sometimes, but I still do it to brush up my problem-solving skills and fundamental understanding of systems and data structures. AI can make mistakes. I was implementing LRU Cache using a linked list, and AI coding did it well, but it wasn’t fitting for my system. I had a better understanding of the context, so maybe I failed to provide enough problem context to AI.
The meta is definitely shifting. Since AI can now instantly output a flawless solution to almost any medium/hard LeetCode problem, asking someone to write it from memory on a whiteboard feels entirely disconnected from 2026 reality. What I'm seeing more of now is a heavy shift towards System Design, architecture, and "Code Review" interviews where you are given AI-generated code and asked to find the hidden bugs or security flaws.
Mmm, I stopped doing leetcode back in 2024 because I realized that even though I was able to build highly efficient small portions of code, I wasn't actually able to build real software because I wasn't able to put all these small pieces together, so far I think it was a good decision because now I'm able to build real software. As for the current state in interviews I have no idea, but personally I think is not as importart or useful for most situations when developing software.
I wonder the same sometimes, but I still do it to brush up my problem-solving skills and fundamental understanding of systems and data structures. AI can make mistakes. I was implementing LRU Cache using a linked list, and AI coding did it well, but it wasn’t fitting for my system. I had a better understanding of the context, so maybe I failed to provide enough problem context to AI.
The meta is definitely shifting. Since AI can now instantly output a flawless solution to almost any medium/hard LeetCode problem, asking someone to write it from memory on a whiteboard feels entirely disconnected from 2026 reality. What I'm seeing more of now is a heavy shift towards System Design, architecture, and "Code Review" interviews where you are given AI-generated code and asked to find the hidden bugs or security flaws.
Mmm, I stopped doing leetcode back in 2024 because I realized that even though I was able to build highly efficient small portions of code, I wasn't actually able to build real software because I wasn't able to put all these small pieces together, so far I think it was a good decision because now I'm able to build real software. As for the current state in interviews I have no idea, but personally I think is not as importart or useful for most situations when developing software.
What did you do to bridge the gap between small portions of code and real software?
LeetCode is still used, but it’s not the only thing that matters now.
With AI, companies care more about how you think and solve real problems, not just memorizing questions.
I do it - but more like a crossword. Do one a day or so
Same issues though - normally comes down to knowing a trick rather for leetcode than actual coding
It depends on whether nowadays FLAG will still need engineer write code during interview.