Anthropic: "Applicants should not use AI assistants"

Original Article ↗ Hacker News Discussion ↗

Reasonableness of the “no AI in applications” rule

  • Some see it as analogous to “no Google during a coding test” or “no calculator for basic arithmetic”: they want to see your motivation and communication, not a tool’s.
  • Others call the request unrealistic and unenforceable: careful AI use is undetectable, so the rule mostly penalizes honest applicants.
  • Several argue the underlying question (“Why do you want to work here?”) is itself bad and mainly tests one’s ability to produce socially acceptable bullshit, not genuine motivation.

Dogfooding, irony, and optics

  • Many highlight the irony: an AI company discouraging AI use exactly in the domain (writing/communication) where it markets AI as helpful.
  • Some interpret it as tacit admission that AI-generated writing is generic “slop” and not what they actually want to read.
  • Others argue the analogy is closer to “brewery asks you not to drink before your interview”—they sell a tool, but still need to see the human.

Assessment quality and interview design

  • One camp: if AI can ace your screen, your test is bad or not representative of the real job; fix the interview rather than banning tools.
  • Opposite view: interviews are about evaluating reasoning, tradeoffs, and communication; AI obscures the signal and wastes interviewers’ time.
  • Some suggest explicit dual-mode processes: allow AI for take‑home tasks, then follow up in-person/AI‑free to verify understanding.

AI as tool vs crutch and impact on learning

  • Long subthread debates whether frequent AI use erodes fundamental skills (like map-reading vs GPS, or coding vs Copilot):
    • Critics say over-reliance produces shallow understanding, poor mental models, and stagnation once AI output goes off the rails.
    • Supporters counter that effective AI prompting and verification is itself a real skill, comparable to using Google, libraries, or higher‑level languages.
    • Some distinguish between seniors (who can safely offload routine work) and learners, for whom AI shortcuts may severely harm skill development.

Fairness, honesty, and power dynamics

  • Several argue ignoring a polite “no AI” request is simply dishonest, akin to lying on a first date.
  • Others respond that hiring is already adversarial—ATS filters, form rejections, arbitrary hoops—so candidates feel justified optimizing however they can.
  • A few note the asymmetry: applicants are told not to use AI while suspecting the company uses AI/ATS to screen them.

Accessibility and neurodivergent concerns

  • A key comment from an autistic/dyslexic perspective: LLMs function as assistive tech to convert visual thinking into acceptable written English.
  • For such candidates, “non‑AI‑assisted communication” means “without the tools that make my real thoughts legible,” which can be exclusionary.
  • This persuades some initially pro‑ban commenters that a blanket prohibition is too broad; they advocate transparency and evaluating how someone uses AI rather than banning it.

AI in hiring pipelines and broader frustration

  • Multiple people speculate Anthropic wants “non‑AI” text partly as cleaner training data.
  • Others suspect they still use AI (or scoring platforms like CodeSignal) to triage candidates, making the rule feel one‑sided.
  • There’s broader resentment at modern hiring: automated rejections, heavy reliance on tests that don’t match real work, and now a meta-layer of “AI vs anti‑AI” gamesmanship.