UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Наталя ХандусенкоAI Eng
29 June 2026, 14:38
2026-06-29
Senior developer found a simple way to "catch" candidates using AI during technical interviews
Senior FullStack Developer Petro Snieda noticed that candidates were increasingly using AI in technical interviews. So to test their real knowledge, he came up with a few tricks.
Senior FullStack Developer Petro Snieda noticed that candidates were increasingly using AI in technical interviews. So to test their real knowledge, he came up with a few tricks.
Petro Snieda is a developer with over five years of experience working as a Frontend Developer at the Ukrainian IT company Yalantis. He is also the founder of the interview preparation platform ITlead , where he conducts technical interviews for subscribers in React, Node.js, application architecture, and System Design.
As Petro explained in his LinkedIn post, there are three main ways to recognize the use of AI:
firstly, this is a contrast in knowledge: the candidate operates with the names of complex tools, but does not understand the basic logic of the banal useEffect;
secondly, “blind” generation: a person produces flawless code, but is unable to explain a single line of his own;
and thirdly, the inability to go beyond the template: the developer confidently quotes terms, but gets lost if asked to rephrase them in simple words.
"Usually it's the other way around: you can forget the name of a hook or an array method, but understand the foundation. But here there are details, but no understanding," the senior developer added.
To test the candidates' real knowledge, Petro is currently testing two techniques.
The first one asks developers to answer some questions with their eyes closed. According to him, this is not done to catch someone. It helps to understand whether the person is remembering the information himself or is peeking at another monitor.
The second life hack is to use traps in the form of false-positive questions.
"It doesn't always work (models get stronger), but sometimes the answer sounds very confident and turns out to be wrong. This is especially noticeable in questions about the event loop, microtasks, or the inner workings of React rendering — I can see for myself what nonsense neural networks talk about reconciliation, for example," Petro noted.
According to him, candidates usually understand this, because he immediately explains the purpose of such an approach.
“Don’t just write code. Write your stories,” — Ukrainian Senior Frontend Engineer on Interviewing at Microsoft and the Ability to Think Quickly Under Pressure
"As the times go, so do the interviews." Team Lead found that three out of seven Software Engineer candidates used AI during the interview. Do IT professionals see this as a problem? LinkedIn discussion