Interviewing a software engineer who prepared with AI

Original Article ↗ Hacker News Discussion ↗

AI vs. Old‑Fashioned Lying

  • Many argue the core problem isn’t “preparing with AI” but fabricating experience, which predates LLMs by decades.
  • AI changes the scale and polish: it can invent plausible projects, resumes, and prep scripts for people who previously lacked the knowledge to even embellish convincingly.
  • Some candidates reportedly pause, “think,” then read out obviously AI‑style paragraphs or contradictory technical claims.

Take‑Home Projects, Coding Tests, and AI

  • Some say simple take‑homes are now trivial for AI, making them poor filters; realistic ones become too heavy for honest candidates.
  • Others still like take‑homes, but only when followed by a deep, interactive walkthrough focusing on trade‑offs, design, extensions, and code quality.
  • Incoherent styles, messy structure, or inability to explain basic decisions are used as signals of AI or proxy work.
  • There’s strong dislike for timed Hackerrank/LeetCode‑style tests (accessibility, speed over quality), but others say they aren’t always strict pass/fail and can be informative.

Interview Format: Remote, In‑Person, and Fairness

  • Several interviewers now insist on camera‑on video to catch obvious cheating (eye tracking, whispers, disappearing and returning with finished code).
  • Autistic and disabled candidates express anxiety about being misread as cheating; some prefer remote for accessibility.
  • A faction predicts a swing back to in‑person days with whiteboards and even suits; many others call this exclusionary, outdated, or “boomer nonsense” and prefer business‑casual and conversation‑based interviews.

Credentials, Baselines, and Screening at Scale

  • Some propose HVAC‑style certifications or bar‑exam‑like fundamentals exams to establish a baseline and reduce arbitrary interviews.
  • Others doubt the industry can agree on what “competence” is, pointing to existing vendor certs that mostly test trivia.
  • With keyword‑matched and AI‑generated resumes flooding pipelines, people expect heavier reliance on referrals and networks.

Detecting AI or Fabrication

  • Effective techniques mentioned:
    • Drill into specific resume bullets (e.g., pagination, rate limiting); ask for concrete data, constraints, and rationale.
    • Require code samples or take‑homes, then spend most of the interview having the candidate explain, critique, and extend them.
    • Favor questions about past work and “how you thought through it” over generic trivia.

Ethics and Tone

  • Some see cheating as a predictable response to opaque, adversarial hiring and AI‑screened resumes; others insist it’s still fraud that hurts honest candidates and teams.
  • The article’s moralizing (“integrity and reputation”) is seen by some as justified; others think lecturing a desperate candidate and blogging their (partially redacted) resume is unprofessional and self‑promotional.