I’m not a fan of algorithmic tests as a first-line screening tool.
As someone who didn’t study computer science and never made a point of memorising algorithms, I’ve always felt these tests were a poor reflection of my abilities as an engineer. I’ve spent my career building software, solving real-world problems, and learning through experience. I had been lucky over the years that I found work where my employers had appreciated my experience and override these algorithmnic test results during my evaluation.
Over the past decade, Big Tech companies have increasingly standardised algorithmic tests through platforms like LeetCode and HackerRank. Smaller companies and startups followed suit, making these tests a common gateway into the industry.
This has made it harder for self-taught engineers, who may lack formal training but have spent years building systems and honing their craft, to even get their foot in the door.
I thought the age of AI might finally change this. If companies are increasingly encouraging, or even mandating, AI use in software development, shouldn’t hiring focus more on how well engineers build with these tools?
But after speaking with people at some companies, I’m seeing the opposite. AI seems to be giving companies another reason to double down on algorithmic tests, to verify that candidates actually possess the skills they claim to have. Some companies that previously didn’t rely on these tests are now placing greater weight on them during screening.
The irony is that we’re using the same old tests to address a new problem. Being able to solve an algorithmic puzzle without AI doesn’t necessarily mean you’re good at building production systems. And being good at building production systems doesn’t necessarily mean you’ll excel at these puzzles.
I’m concerned that experienced engineers may now face an even higher barrier to entry, having to prove their abilities through tests that may not reflect the skills they’ve spent years developing.
If AI is changing how we build software, shouldn’t we also rethink how we evaluate the people who build it?
P/S: I understand we can argue that then they should be able to ace their technical interviews instead, but this assumes they can get past the first screening.