Are coding interviews still relevant? AI cheating & the future of tech recruitment

Recently, X has been flooded with discussions about InterviewCoder.co, an AI tool that automatically solves engineer recruitment challenges. People call it an “invisible AI,” a tool that helps candidates pass coding interviews undetected. Imagine showing up to an online coding assessment, turning it on, and letting AI spit out the perfect solution. No stress, no struggle, not even a slight preparation. If this is where we’re heading, then what’s the point of coding interviews anymore?

AI cheating or just adapting to the future?

Back in the day, cheating on coding interviews was such a hassle. The best we could do was sneak in a second monitor, google solutions, or memorize common patterns. Since AI, cheating on coding assessment is on a whole other level. AI can generate full solutions, explain them in detail, and optimize them for better performance. Not only does AI know the answer, but it can also pass interviews better than we can.

The rise of AI in completing coding assessment raises an important question: Is this really cheating or simply the natural evolution of the industry? Netizens on X argue that traditional coding tests, like LeetCode-style challenges, are already a flawed measure of real-world ability. After all, engineers already regularly use AI tools like ChatGPT or Cursor in their day-to-day jobs, be it to speed up generating some boilerplate components or generate some test case possibilities, so why should coding assessment be any different? But on the flip side, if AI can solve all the problems for us, then what are companies even testing for? Is it skill? Or just the ability to game the system?

Recruiters are already catching on, though. If a solution looks too perfect, they’re going to be suspicious. Some companies are scrambling to implement AI-detection measures, while others are starting to rethink the entire interview process. Maybe it’s time for something better.

The real problem: coding interviews were already broken

Let’s be honest: coding interviews were flawed even before AI came into the picture. Many companies still rely on an evaluation system that has little to do with actual engineering work. You get asked to implement quicksort from scratch, but once you land the job, you’re debugging APIs, reviewing pull requests, and dealing with system architecture. The disconnect is real.

Then, there’s the speed coding nonsense. Interviewers expect you to think, code, and explain your solution in minutes. While in reality, real-world engineering involves research, iteration, and collaboration. Now, with AI able to generate solutions instantly, the absurdity of these tests is even more obvious. Are we testing problem-solving skills or just measuring who can type the fastest under pressure?

If not coding assessment, then what?

If coding interviews in their current form are becoming pointless, what’s the alternative? Here’s what might actually work:

• Project-Based Assessments Instead of asking random algorithm problems, have candidates complete a real-world coding challenge. If you’re hiring a frontend engineer, ask them to build an actual UI component. If it’s for a backend role, give them a small API to design. This assesses the candidates’ actual skills instead of memorization.
• Live Coding with Explanation Instead of making candidates rush through code, let them explain their thought process. Give them time to think and discuss trade-offs. This shows how they approach problems rather than just whether they’ve memorized solutions.
• Pair Programming Engineers rarely work in isolation. So, why not assess them by pairing them with an interviewer to solve a problem together? This reveals communication skills, collaboration, and debugging ability, all things that matter far more than whether you can reverse a linked list on demand.
• System Design Interviews AI might be good at spitting out code, but it struggles with big-picture thinking. System design interviews force candidates to consider scalability, trade-offs, and architecture, making it a better test of engineering ability than pure coding challenges.