Hiring good people used to mean finding the ones who actually knew their stuff.
Now it mostly means finding someone who can prompt GPT better than you.
Let’s be honest, technical interviews were already broken long before AI. But at least you could still separate the “can build” from the “can talk”. Now that line’s completely gone. With AI tools, anyone can look like a 10x developer for 30 minutes straight.
So how do you even evaluate skill in 2025?
The Collapse of the Skill Signal
If a candidate uses Cursor to write a Kubernetes manifest or a Python API during an interview (and it works) did they cheat, or just work smarter?
Because if the same person would use AI on the job (and everyone does), then maybe banning it in interviews makes as much sense as banning Stack Overflow in 2012.
The problem is, AI has flattened the visible skill curve.
Everyone can now produce something that looks senior-level until the moment something breaks and the AI stops being helpful. Then, suddenly, it’s debugging time, and you find out who actually understands what’s happening under the hood.
Do We Even Need Skilled People?
Here’s the uncomfortable question: if AI makes anyone look competent, do we still need competence?
If Copilot and Gemini can build features, explain architecture, and even generate test plans, maybe “skill” isn’t what we should be hiring for anymore.
Maybe we should hire for judgment.
Because anyone can generate code.
But not everyone knows when the AI is confidently wrong.
The best engineers I know don’t resist AI, they wrestle with it. They use it to move faster, but they double-check every step. They know where to trust automation and where to think like a human.
Interviews Are Now a Mirror of Real Work
Let’s stop pretending interviews happen in a vacuum.
Nobody codes without tools anymore. The real question isn’t “can you solve this from memory?” — it’s “can you solve this with all the modern tools at your disposal and still know what’s going on?”
So maybe the right interview now looks like this:
“Here’s a real-world task. Use any AI you want. Just narrate your thinking as you go.”
That’s how you find the ones who don’t just use AI, they command it.
The New Definition of Skill
In this era, skill isn’t about syntax or frameworks. It’s about adaptability, verification, and ownership.
The unskilled engineer trusts the AI blindly.
The skilled one assumes it’s lying, and still manages to deliver.
