Skip to content
Business Efficiency Talking point

A hiring-test vendor now apologizes for overweighting puzzle questions

Vivek Ravisankar says the weight on algorithm puzzles outran real-work tests, and AI has now made the old questions easy.

W
WebPulse Newsroom
AI-assisted · 2 min read
Share on X LinkedIn
A hiring-test vendor now apologizes for overweighting puzzle questions

Photo: Thirdman / Pexels

In brief
  • HackerRank's CEO said the industry gave too much weight to textbook fundamentals in interviews, and his company is now apologizing for it.
  • He said AI solves the classic questions in one try, so tests are moving toward real code repositories, judgment and AI fluency.

Vivek Ravisankar, who co-founded and runs HackerRank, told Software Engineering Daily that the industry put too much weight on textbook computer-science puzzles when hiring engineers. HackerRank is one of the best-known sites for interview practice. His company is now paying to say sorry in public.

What was said

Ravisankar told host Josh Goldberg how the habit spread. Developers long complained that interviews had nothing to do with the job. Google, he said, asked hard data-structure and algorithm questions. Its engineers were seen as the best, so other firms copied the style. Candidates objected that they never reversed a linked list or inverted a binary tree at work.

Then AI changed the picture. He said it "was able to one shot all of these different algorithm questions and made them pretty much useless in an interview setting." That forced every company, HackerRank included, to rethink hiring.

Ravisankar said the weight on fundamentals went too far compared with testing real work, and that this is being rebalanced. HackerRank has put up billboards in San Francisco apologizing for the old questions. In his words: "It's like we're actually spending money. Like it's literally an apology tour."

The replacement, he said, moves from standard algorithm questions to real-world questions on code repositories. It weighs critical thinking, judgment and AI fluency above code correctness. Candidates move from a plain code editor to what the company calls an agentic development environment, or ADE, meant to mimic actual work.

Put simply, and this is our gloss rather than his words: a classic puzzle is short, well known and self-contained, with one checkable answer. That suits an AI model. A repository task asks a candidate to read existing code and judge what to change.

Why it matters

Our reading: the puzzle interview spread because firms copied a famous company, and the excerpts offer no evidence that it predicted job performance. Goldberg raised the concern that someone able to spend 20 hours a week on practice problems gains an edge unrelated to the work. Whether candidates and working engineers paid for that is our interpretation, not something the show established.

For managers who hire or buy hiring tools, the lesson is plain. Ask what each stage measures and whether the job ever asks for it. A process kept only because everyone uses it can look sound until a new tool makes it trivial.

The other side

Ravisankar did not call fundamentals worthless. He compared hiring to airline pilots. Even if AI flies most of a flight, you still want a pilot who knows the engineering. His pilot process opens with a no-AI multiple-choice test of fundamentals.

He also pushed back on the cost to candidates. Extra preparation, he said, is still a skill, and he saw the fairness problem as pedigree bias, such as rejecting people for a low grade average or the wrong university.

The excerpts leave one thing open. They give no evidence yet that tests built on real repositories and AI fluency predict job performance better. That claim comes from a vendor selling the new approach.

Written by the WebPulse Newsroom with AI assistance, and checked by our editorial review: every quotation was verified against the recording's transcript. How we use AI.

The conversation this talking point comes from