IdeaWell

Project ideas from Hacker News discussions.

The Slow Formation of Durable Software

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Theme 1: Speed vs. Quality/Durability
Many commenters argue that rapid development encouraged by AI often sacrifices long‑term reliability and maintainability.

“speed kills quality. it’s literally impossible to make anything good fast.” – ORDINAND_PIZZA
“The trade‑off between quality and speed of development has always been a tenet of software engineering.” – danmaz74
“People say that agentic development is great because you can churn out so much so fast. But that doesn't mean that any of it will be truly good and reliable.” – adamddev1

Theme 2: LLMs Excel at Adding/Prototyping but Struggle with Removal/Refactoring
AI agents are praised for quickly generating code and prototypes, yet they are seen as poor at simplifying or deleting existing code, leading to ever‑growing codebases.

“I have personally detailed entire projects because adding a feature seemed easy with AI. It is very difficult to vibe code and not add a bunch of useless crap features.” – ghoshbishakh
“Really? I find LLMs quite bad at deleting code. … the codebase still grows. Every time I've tried it, llms have failed to simplify code via refactoring.” – josephg
“To delete text using a text generator, you have to emit the original thing taking care to omit the deleted stuff during emission. It's more work.” – lelanthran

Theme 3: Clear Vision, User Research, and Iterative Feedback Remain Essential
Even with AI, successful software requires understanding the problem, gathering user needs, and iterating—LLMs cannot replace this discovery process.

“we could not have accelerated Zotero’s conception, because we did not know exactly what we wanted, and so could not have written coherent prompts for an LLM.” – kstenerud
“The hard part of evolving Scribe and Web Scrapbook was discovering that a browser extension manipulating a local SQLite database was the only architecture that could reconcile local offline persistence with live DOM scraping …” – scruple
“It is MUCH easier to make something people want, than to make them want something you made.” – fw

🚀 Project Ideas

AI-Powered Code Deletion Assistant (ChatDPT)

Summary

  • An interactive AI pair‑programmer that helps developers safely delete unused code, refactor bloated features, and keep repositories minimal while preserving functionality.
  • Core value proposition: reduces code‑base growth caused by LLMs that excel at adding but struggle at deleting, turning “vibe coding” into disciplined, lean development.

Details

Key Value
Target Audience Software engineers and teams using LLMs for code generation who notice uncontrolled code bloat
Core Feature Chat‑driven proposal of safe deletions, impact analysis via static analysis and test runs, diff generation with one‑click apply
Tech Stack LLM backend (CodeLlama / GPT‑4 via API), tree‑sitter AST parser, pytest/jest test harness, VS Code extension (TypeScript)
Difficulty Medium
Monetization Revenue-ready: subscription per developer ($10/mo) or usage‑based API credits

Notes

  • HN commenters explicitly asked for “ChatDPT” to counter LLMs’ inability to delete code (iamnothere, lelanthran). This tool directly satisfies that demand.
  • Provides a concrete way to combat feature creep highlighted by ghoshbishakh and josephg, encouraging discussion on sustainable coding practices.

Self‑Hostable, Open‑Source Zotero Alternative with One‑Click Sync

Summary

  • A desktop‑first reference manager that stores data in open formats (BibTeX, JSON, Markdown, CSV) and syncs effortlessly via WebDAV or IPFS, eliminating vendor lock‑in.
  • Core value proposition: gives academics and power users full control over their research library while retaining Zotero‑level usability and plugin extensibility.

Details

Key Value
Target Audience Researchers, academics, and knowledge workers frustrated with Zotero’s self‑hosting complexity and desire data portability
Core Feature One‑click local server setup with automatic export to multiple open formats and seamless WebDAV/IPFS sync across devices
Tech Stack Tauri (Rust backend) + React frontend, SQLite storage, WebDAV server integration, IPFS optional layer
Difficulty Medium-High
Monetization Hobby (open source) – optional hosted premium tier for $5/mo for managed backups

Notes

  • Commenters praised Zotero’s “just works” nature but lamented self‑hosting friction (Almondsetat, pajamasam). This project removes that pain point while preserving the open‑data ethos.
  • Enables discussion on durable software (shieldagent, antonyragleap) by emphasizing open formats and vendor‑independent sync, a frequent HN theme.

Feature‑Impact & Bloat Analyzer for Agentic Coding

Summary

  • A GitHub‑action‑integrated tool that evaluates proposed feature additions for likely code‑base growth, maintenance cost, and suggests simpler alternatives before merging.
  • Core value proposition: helps teams avoid the “quick‑add, later‑regret” trap of LLMs by quantifying feature impact and encouraging lean decision‑making.

Details

Key Value
Target Audience Product managers, tech leads, and indie hackers using agentic coding pipelines who want to control scope creep
Core Feature Scans PR description & code, runs complexity metrics, asks LLM for impact summary, returns a score and refactor suggestions
Tech Stack LLM (GPT‑4o or Mistral) for semantic analysis, static analysis tools (SonarQube, CodeQL), GitHub Action (Node.js)
Difficulty Medium
Monetization Revenue-ready: SaaS tiered pricing – free for public repos, $20/mo per private repo

Notes

  • Directly addresses concerns from ghoshbishakh and josephg about LLMs making it easy to add useless features, and from scruple about missing architectural insight.
  • Sparks practical utility debates on balancing speed vs. quality in AI‑assisted development, a recurring HN discussion point.

Read Later