HackerRank open sourced its ATS. My resume scored 90/100. Oh wait 74. No – 88(danunparsed.com)
558 points by sambellll 10 hours ago | 226 comments
tl;dr: HackerRank's open-sourced ATS (hiring-agent) produces wildly inconsistent resume scores—the same resume scored anywhere from 66 to 99 across 100 runs, meaning a candidate could fail an 85-point cutoff 65% of the time purely by chance. The author traces this to flawed design: subjective categories like "projects" rely on LLM vibe-checks with no proper rubric, while "experience" gives everyone 25/25 regardless of seniority due to a two-line prompt. With 65% of the score weighted on open source and projects, the tool filters by noise rather than quality.
HN Discussion:
  • LLM stochasticity is poorly understood and confirms why resume screening is broken
  • Inconsistent randomness in hiring is absurd and unfair to candidates
  • Weighting open source/projects heavily disadvantages candidates with other life commitments
  • From a recruiter's perspective, the randomness is actually an acceptable filtering rate given application volume
  • The tool is poorly built (tiny model, vibe-coded) which explains its flaws