Cadre has been recruiting engineers for startups since 2007, and in 2026 I’ve never seen this much disagreement about how to interview.
On our weekly team meetings we compare interview loops across our early-stage clients, and the approaches are all over the place. Some companies are banning AI in interviews entirely. Others are telling candidates: use it as much as you want, or as you see fit. There’s no clear consensus yet.
But there is a clear shift: the way engineers actually work has changed, and interview processes haven’t caught up — especially on teams that don’t come from Silicon Valley DNA.
For years, LeetCode-style interviews were the default across our clients — tier-one, VC-backed startups with world-class engineering teams — designed to test fundamentals, pattern recognition, and problem solving under pressure. But those circuits were built in a different era. Today the strongest engineers use AI to move faster and build real things. Asking them not to use it in an interview creates a scenario that doesn’t reflect the job.
What we’re seeing from forward-thinking teams: give real, open-ended problems. Let candidates use AI. Evaluate how they think, not just what they produce — we want to see some business logic. Do they understand what’s being built? Can they make good decisions in ambiguous situations?
One thing we’re not saying: that vibe-coded projects are the future. It’s the opposite. Anyone can spin something up quickly with AI — but without structure, testing, and thoughtful design, those projects don’t hold up. They break, they don’t scale, and they’re hard to maintain.
At Cadre we’re increasingly aligned: LeetCode-style interviews are probably no longer the best proxy. Not because fundamentals don’t matter — but because they’re no longer what separates great candidates. The gap now is in how someone thinks, adapts, and builds with modern tools.
If your interview process looks like it did three years ago, you’re not just behind — you’re screening out the exact people you want.
