Don't post generated/AI-edited comments. HN is for conversation between humans
TL;DR Highlight
Hacker News officially added a rule banning AI-generated or AI-edited comments — HN discusses what this means and whether it'll work.
Who Should Read
Anyone who participates in online technical communities and cares about the quality of discourse, and developers thinking about AI content moderation.
Core Mechanics
- Hacker News updated its official guidelines to explicitly prohibit comments generated or substantially edited by AI.
- The rule targets both fully AI-generated comments and comments where a human used AI to polish or expand their writing.
- Enforcement is necessarily imperfect — HN can't reliably detect AI-generated text and relies on community flagging and moderator judgment.
- The rationale: AI-generated comments dilute the distinctive HN voice, reduce authentic discourse, and can be produced at scale to manipulate discussion.
- This puts HN in a different posture than most platforms, which have taken a permissive or hands-off approach to AI-assisted content.
Evidence
- The HN guidelines update was linked in the announcement thread, with 'dang' (the main HN moderator) explaining the reasoning.
- Community reaction was mixed: many welcomed it as protecting HN's signal quality, others argued it's unenforceable and draws an arbitrary line.
- Practical debate: is grammar-correcting AI different from spell-check? Where's the line between 'AI assistance' and 'AI generation'?
- Some noted that a skilled human using AI assistance to write a thoughtful comment is probably better for discourse than a careless human writing without it.
How to Apply
- For online community managers: HN's approach of explicit prohibition with community norm enforcement is worth watching — the rule's main value may be establishing a community norm rather than perfect technical enforcement.
- If you use AI to help with writing in online communities, read the specific platform rules — 'AI assistance' policies vary significantly across communities.
- For AI content detection: HN's approach implicitly acknowledges that AI detection tools are unreliable — community judgment and norms may be more effective than technical detection.
Terminology
Related Papers
What happened after 2k people tried to hack my AI assistant
실제로 6,000개 이상의 이메일로 AI 에이전트에 prompt injection 공격을 시도한 공개 실험 결과로, Claude Opus 4.6이 비밀 파일 유출을 한 번도 허용하지 않았지만 실험 설계의 현실성에 대한 논란이 뜨거웠다.
When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models
여러 LLM을 조합해도 '모든 모델이 동시에 틀리는 비율(β)'이 성능 상한선이며, 업계가 쓰는 pairwise 상관계수(ρ)는 이 상한선을 예측하지 못한다.
Beyond Function Calling: Benchmarking Tool-Using Agents under Tool-Environment Unreliability
실제 환경처럼 API가 망가지거나 결과가 이상할 때 LLM 에이전트가 얼마나 잘 버티는지 측정하는 벤치마크 ToolBench-X 공개.
Nearly Half of LG Smart TV Apps Contain Residential Proxy SDKs
6,038개의 LG·Samsung 스마트 TV 앱을 스캔했더니 2,058개에서 사용자의 IP를 몰래 팔아 트래픽을 중계하는 Residential Proxy SDK가 발견됐다. TV는 컴퓨터처럼 감시받지 않아서 프록시 호스트로 거의 이상적인 환경이다.
Prompt Injection as Role Confusion
LLM이 시스템 프롬프트, 사용자 입력, 툴 출력을 구분하지 못하는 구조적 결함이 prompt injection의 근본 원인이라는 ICML 2026 논문으로, 현재 LLM 보안 아키텍처의 한계를 명확히 분석한다.
GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2
모델 크기가 커질수록 성능이 좋아진다는 통념에 반해, 오픈소스 753B 모델 GLM-5.2가 추정 1~2T 규모의 GPT-5.5보다 환각 비율이 3배 낮다는 벤치마크 결과가 나왔다. 단순히 파라미터 수와 벤치마크 점수만으로 모델을 선택하면 실제 업무에서 낭패를 볼 수 있다는 경고다.