Ask any seasoned engineer what they hate most about job hunting. The answer is almost always the same — 'Google-style' rounds: whiteboard algorithms, LeetCode puzzles, dynamic programming on arrays they will never touch on the job.

The candidate burns a vacation day, grinds tree-balancing on LeetCode for weeks, sweats in front of a bored interviewer trying to write code in a Google Doc with no syntax highlighting. The next morning they start the actual job and spend it moving JSON between microservices, tweaking SQL queries, and patching legacy.

Where did this gap between real work and how we evaluate it come from?

The cargo cult

Hard-algorithm interviews became fashionable because the biggest tech companies use them. When fifty thousand people apply to your role with a fat compensation package, you need a harsh, mathematically-clean filter. That kind of interview does not test genius. It tests grind, willingness to memorise, and tolerance for arbitrary corporate ritual.

But when a 40-person startup, a regional product company, or a systems integrator blindly copies that format — it is working against itself. You are not BigTech. By asking candidates to solve olympiad puzzles you push away exactly the pragmatic engineers you need: people who ship features fast, glue together messy business logic, and keep production up under peak load.

Strong practitioners walk to competitors who respect their practical experience rather than make them sit a computer-science final. Your filter is selecting for patience and memorisation, not the work.

What works instead

Evaluate candidates against work they will actually do.

  • Code review of a real piece — show them a non-critical fragment of your codebase and ask what they would improve.
  • Pair-programming on a real ticket — even a 60-minute session tells you more than three leetcode rounds.
  • Architecture conversation — discuss how they would design the next module given your current constraints, your stack, your scale.
  • Reading their previous work — public GitHub, an open-source contribution, or a documented decision in a past project.

None of that takes longer than a LeetCode round — and it tells you whether you actually want to work with the person.

If you remember nothing else

Algorithm rounds at scale make sense for companies fielding tens of thousands of applications. For everyone else they are an expensive theatre that filters out the engineers most worth hiring.

✉
Get in touch
Want to redesign your tech interview?

Email [email protected] with a few lines about your current process — we will respond with a candid view and concrete next steps.

Email Anastasia →