Most software interview methods were shaky even before AI showed up. Now the weak ones are just easier to game.
If somebody can paste your take-home into a model and send back a polished answer, that tells you almost nothing. If they can grind through a whiteboard problem they memorized from LeetCode, that also tells you almost nothing.
So my process is simple: get people talking about real work, then pay attention.
What I Actually Do
If someone has a decent work history, I ask about their resume and then I ask what they’re working on right now.
That second question throws people off a little because most interviewers do not ask it. Good. I do not want the rehearsed answer they have given twelve times already. I want the version where they start talking about something real.
Once they do, I ask the questions I actually have. What is hard about it? What is annoying? What broke? What would they change if they had more time? What are they still unsure about?
That is where people stop performing and start showing you who they are.
If they are earlier in their career, I use the same approach, but I also ask them to show me code they wrote and walk me through it. Then I poke at it a little. Edge cases. Tradeoffs. What they would do differently now. Not to trap them. Just to see how they think once the script runs out.
What I Am Looking For
I care way less about whether somebody looks brilliant in a one hour interview than whether they are honest when things get messy.
Can they admit they do not know something?
Can they listen?
Are they curious?
Are they willing to learn?
Are they going to be a pain in the ass to work with?
That last one matters more than people want to admit. I would rather hire someone solid, humble, and engaged than someone technically sharper who makes every room worse.
When I think someone is bullshitting me, I ask more specific questions until they either settle down and get honest or keep digging. That moment matters. Once people stop pretending, you can usually tell what is actually there.
For junior roles, I want eagerness and honesty. For senior roles, I want humility and judgment. If a senior engineer cannot say “I was wrong” or “I do not know,” that is a much bigger red flag than missing some trivia question.
What AI Changed
AI did not ruin interviewing. It just exposed how many interview loops were built around fake certainty.
If your process depends on candidates being unable to look things up, your process was already fragile. Real work has always involved searching, checking docs, asking questions, and feeling around in the dark for a while before the answer shows up.
The useful signal is not whether someone can produce an answer in isolation. It is whether they can think clearly, communicate honestly, and work through uncertainty without becoming arrogant or useless.
That is harder to measure than a quiz. It is also closer to the job.
The Whole Thing
I do not have research backing this up. My source is basically years of talking to people and seeing who turns out to be great to work with.
That is good enough for me.