The coding interview is being rewritten.

Meta, Google, Canva, and thousands of companies now let candidates use AI during technical interviews. The data is hard to ignore: 91% of engineers already use AI tools at work. 75% shipped AI-generated production code in the last six months. Over 35,000 AI-assisted interviews have already been run on CoderPad alone.

The interview is finally catching up to the job.

But here's what most people miss:

When everyone has the same AI, code generation becomes a commodity. So the interview shifts to what AI can't do — judgment, verification, tradeoff reasoning, and communication.

I've been on both sides of this. The pattern is clear:

• Candidates who think first and prompt second outperform everyone
• The strongest signal is verification — catching what the AI gets wrong
• Going silent while pasting AI code is the fastest way to fail
• Selective AI usage shows more confidence than using AI for everything
• 71% of engineering leaders say AI is making technical skills harder to assess, not easier

The engineer's role is shifting from writer to editor-in-chief. And editing — knowing what's good, what's wrong, and what's missing — is harder than writing ever was.

I wrote a deep-dive on what's actually changing, the mistakes I've seen, and how to prepare.

The interview isn't testing whether you can code. It's testing whether you can think.