IdeaWell

Project ideas from Hacker News discussions.

How to Write with an LLM

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

1. LLMs are helpful as copy‑editors or fact‑checkers, but shouldn’t replace the writer’s voice
- “Don’t take all of the model’s copyediting advice.” – jamiejquinn
- “use them like a copyeditor rather than a ghostwriter” – sixtyj
- “I do something like this with code all the time, ‘is this idiomatic ?’, I find it very helpful.” – benoau

2. Widespread concern that AI‑generated text degrades quality, literacy, and the value of human writing
- “If you can't spend the time to write it, why should anyone read it?” – S‑E‑P
- “Excessive generative AI output devalues all communication media unfortunately.” – jakeinspace
- “I fear in the future people will read even less because of how often people cheap out and use AI to write.” – S‑E‑P

3. Irony/joke about the article’s own use of an LLM‑tell (“load‑bearing”) despite advising against it
- “The placement of the 'Those rethinks are load-bearing parts of your voice' sentence in the article makes me think the author intended the word choice.” – yapfrog
- “I saw 'load-bearing' there and instantly assumed Thomas was making a deliberate point…” – simonw
- “SupremumLimit: …the post includes this chunk of AI slop: 'Those rethinks are load-bearing parts of your voice.'” – SupremumLimit

🚀 Project Ideas

AI Voice Detector & Editor

Summary

  • Highlights LLM‑generated phrasing patterns (e.g., “load‑bearing”, overused AI‑tells) in a writer’s draft.
  • Provides inline suggestions to replace or rework those patterns while preserving the author’s voice.
  • Core value: helps writers keep their text human‑sounding and avoid inadvertent AI‑style artifacts.

Details

Key Value
Target Audience Bloggers, technical writers, students, anyone using LLMs for drafting
Core Feature Real‑time detection of AI‑tells + optional rewrite suggestions
Tech Stack Python (spaCy, transformers), React frontend, WebAssembly for in‑browser processing
Difficulty Medium
Monetization Revenue‑ready: Freemium (free detection, paid advanced style‑bank)

Notes

  • HN commenters lament AI‑tells like “load‑bearing” creeping into their writing; this tool surfaces those exact phrases (see simonw & SupremumLimit comments).
  • Encourages discussion about preserving authorial voice while still benefiting from LLM assistance.

Fact‑Check Companion

Summary

  • Scans a draft for factual claims, retrieves supporting sources, and flags unverified or hallucinated statements.
  • Shows confidence scores and lets the user attach citations without altering the original wording.
  • Core value: gives writers a safety net for accuracy while keeping full control over phrasing.

Details

Key Value
Target Audience Journalists, researchers, content creators, technical bloggers
Core Feature Claim extraction + provenance‑backed verification via search APIs
Tech Stack Node.js/Express backend, Elasticsearch or custom vector DB, Vue.js UI
Difficulty Medium
Monetization Revenue‑ready: Subscription per‑user or per‑document API calls

Notes

  • Many users (e.g., thombles, benoau) value LLMs for fact‑checking but fear hallucinations; this tool offers verifiable citations.
  • Addresses the concern that AI‑generated output devalues trust (S‑E‑P, jakeinspace) by making verification transparent.

Collaborative Style Review Board

Summary

  • A lightweight web app where writers submit drafts and receive style‑focused feedback (passive voice, run‑ons, overused phrases) from configurable rule sets or peer reviewers.
  • Allows toggling between AI‑generated suggestions and human reviewer comments, with a “voice‑preserve” mode that blocks rewriting.
  • Core value: supplies a second‑eye review that targets mechanics without overriding the author’s tone.

Details

Key Value
Target Audience Writers, editors, students, open‑source project maintainers
Core Feature Rule‑based style feedback + optional human reviewer comments
Tech Stack Django REST API, PostgreSQL, React + Redux, optional LLM microservice for suggestions
Difficulty Low
Monetization Hobby (open‑source) or Revenue‑ready: Tiered team plans for private repos

Notes

  • Commenters like ameliaquining and awithrow appreciate LLMs for spotting passive voice and overused tics but worry about losing voice; this tool lets them opt‑in/out.
  • Encourages practical utility: teams can adopt a shared style guide while retaining individual expression, echoing the desire for “second eyes” not “first hands”.

Read Later