Coding interviews are testing the wrong thing.

Traditional coding interviews are becoming irrelevant, and the reason is straightforward: AI now writes a large share of production code. Engineers aren't hand-coding everything from scratch anymore, they're thinking, planning and problem-solving with it.

Yet most interviews still test one thing: “Can you write a perfect binary tree traversal in 45 minutes?” Wrong question.

The questions worth asking instead

• Can you break down a complex, ambiguous problem?
• Can you reason through edge cases with AI assistance?
• Do you know when to trust AI, and when to intervene?
• Can you communicate your thought process clearly?

The engineers who'll dominate the next decade aren't just coders. They're AI collaborators who can think strategically, reason deeply, and build faster than ever.

Which leaves the open question, and I'm genuinely interested in how other people are solving it: how would you evaluate a candidate's ability to work with AI in a technical interview?