IdeaWell

Project ideas from Hacker News discussions.

Why is Anthropic's public writing style so unlike Claude's?

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Theme 1 – Distrust of AI‑generated “slop”
Many commenters say they can instantly spot Claude‑written text and dismiss it if the author didn’t write it themselves.

“On the bright side, I’ll know immediately a post is Claude‑generated. If the author didn’t bother writing it, I don’t bother reading it.” – wiradikusuma

Theme 2 – Language of “collaboration” vs. tool use
There’s tension over whether describing interaction with LLMs as “collaborating” unfairly shifts responsibility or builds trust.

“I find the language of ‘collaborating with Claude’ off‑putting. I don’t collaborate with Claude, I use Claude. It’s a tool in my hands, not a colleague or a friend.” – names_are_hard
“I think the opposite. AI Companies use the term collaborate to increase the trust in the LLM output.” – einszwei

Theme 3 – User accountability and corporate caution
Commenters stress that users bear responsibility for LLM output, and note that companies like Anthropic avoid using their own models for public writing.

“Anthropic certainly isn't planning to take responsibility for Claude's mistakes - that's the user's responsibility.” – michaelt
“Anthropic's public writing style is very different from Claude's because it's written by humans. Even in their job ads they ask candidates to not use AI for any writing…” – salvozappa

🚀 Project Ideas

ClaudeSlop Detector

Summary

  • A browser extension that analyzes text in real‑time to estimate the probability it was generated by Claude or similar LLMs, flagging low‑effort “AI slop” posts.
  • Core value: lets readers instantly decide whether to invest time in a comment or article, reducing noise from AI‑generated content.

Details

Key Value
Target Audience Hacker News readers, moderators, and any community forum users
Core Feature Inline AI‑generated likelihood score (0‑100%) with optional auto‑hide for low‑scoring content
Tech Stack JavaScript/TypeScript, WebAssembly for fast inference, optional Rust backend for model serving; uses open‑source detectors like DetectGPT or a fine‑tuned RoBERTa classifier
Difficulty Medium
Monetization Revenue-ready: SaaS API subscription (tiered by request volume) + free basic extension

Notes

  • HN users expressed frustration: “I'll know immediately a post is Claude-generated. If the author didn't bother writing it, I don't bother reading it.” – wiradikusuma
  • Provides a concrete tool to act on that sentiment, sparking discussion about AI transparency and content quality.

HumanWrite Assist

Summary

  • A writing assistant that helps non‑native English speakers improve their prose while deliberately avoiding typical LLM phrasing, preserving a human voice.
  • Core value: produces clear, natural‑sounding text without the “AI slop” that triggers skepticism.

Details

Key Value
Target Audience Non‑native English speakers, bloggers, job applicants, anyone wary of sounding AI‑generated
Core Feature Real‑time suggestions that highlight AI‑like patterns and offer human‑style rewrites (style‑guided, not autocomplete)
Tech Stack Frontend: React/Vue; Backend: Python (FastAPI) with a small open‑source LLM (e.g., LLaMA‑2‑7B) + rule‑based style checker; optional WASM for client‑side checks
Difficulty Medium
Monetization Hobby (open‑source) or Revenue-ready: freemium model – free basic suggestions, premium for advanced style guides and team sharing

Notes

  • Commenters noted the tension: “non native English speakers using AI to blogslop… is actually a good tool!” – reddozen; this tool gives them a way to benefit from assistance without sounding AI‑generated.
  • Encourages discussion about ethical AI use and authentic voice.

AI Content Badge Service

Summary

  • A verification platform where publishers can obtain a “Human‑Written” badge after passing AI‑detection checks or human review, displayable on articles, newsletters, or forum posts.
  • Core value: builds trust with audiences tired of hidden AI authorship, letting creators prove their content’s authenticity.

Details

Key Value
Target Audience Bloggers, news outlets, newsletters, forums like HN, job application portals
Core Feature Submission flow: text scanned by AI detector + optional human moderator; successful submissions receive embeddable badge SVG/HTML
Tech Stack Node.js/Express, PostgreSQL for records, AWS Lambda for detector calls, React dashboard for badge management; badge embed via simple script tag
Difficulty Low-Medium
Monetization Revenue-ready: pay‑per‑verification (e.g., $0.05 per check) or subscription for unlimited badges

Notes

  • HN discussion highlighted distrust: “Anthropic's public writing style is very different from Claude's because it's written by humans.” – salvozappa; a badge would let creators prove the human origin.
  • Provides a tangible way for communities to reward authentic content and debate AI disclosure norms.

Read Later