Project ideas from Hacker News discussions.

AI Can Make You Suck Faster Too

📝 Discussion Summary (Click to expand)

Four prevalent themes in the discussion

  1. AI speeds up the creation of low‑quality code
  2. “Pre‑AI, the limiter on my bad code was how much code I could type in a day..... Now with AI I can 10x my output and 10x my bad code!” – Lapalux
  3. “As a user they'll probably see something relatively nice, with some maybe strange behavior and non ideal, but what software sin't non ideal? It does most of what we wanted, and you finished it how fast!?” – nomel

  4. Judgments about AI’s usefulness hinge on the model used; citing weak models (e.g., DeepSeek) undermines broad claims

  5. “Why not use actual frontier models, and you know do some real research, before writing a blog post?” – Zakis1
  6. “Seems like they've already done more research than you. Their results not matching your expectations does not indicate a lack of research.” – solid_fuel
  7. “Using DeepSeek to dismiss better models is the bad argument here.” – smokel

  8. AI empowers non‑technical staff and “vibe coding,” leading to maintenance and reliability problems

  9. “At least once a week I have to drop what I'm doing and race in to stop the non dev manager from merging a fully vibe coded patch which will blow up production.” – Gigachad
  10. “CEO hired an AI forward CTO … thousands of lines of misguided, verbose, brittle code … go essentially straight into production.” – NichoPaolucci

  11. When paired with proper harnesses, prompting, and oversight, AI can augment productivity rather than replace skill

  12. “Use a very good model. Set up a good harness. Spend some time on your system prompts and skills…. Don't write it off until you've genuinely seen the upsides.” – mikenew
  13. “You can get around these not by just 'prompting better', but by establishing rules and pipelines that are enforced with standard software engineering tools … AI orchestrating all of the tools that we know work so well.” – ModernMech

🚀 Project Ideas

CodeGuard AI Reviewer

Summary

  • Automated PR review bot that detects AI‑generated code patterns and flags potential hallucinations, security issues, and style violations.
  • Integrates with CI/CD to enforce quality gates before merging, reducing the risk of “vibe‑coded” production bugs.

Details

Key Value
Target Audience Engineering teams that use LLMs for code generation (frontend, backend, infra)
Core Feature LLM‑powered static analysis that scores each diff for confidence, highlights suspicious snippets, and suggests fixes
Tech Stack Python, FastAPI, GitHub Actions, OpenAI/Anthropic APIs, Ruff/Bandit for linting
Difficulty Medium
Monetization Revenue-ready: $10 per active developer per month

Notes

  • HN commenters complained about managers merging AI‑slop without review; this gives them an objective gate. (e.g., “Gigachad: …drop what I’m doing and race in to stop the non dev manager from merging a fully vibe coded patch”)
  • Provides discussion material: teams can tune confidence thresholds and debate trade‑offs between speed and safety.

Read Later