IdeaWell

Project ideas from Hacker News discussions.

A recent experience with ChatGPT 5.5 Pro

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

1. Redefining GraduateMathematics Assessment

"90% of the final grade are in room examinations ... I don’t see why we shouldn’t allow the wholesale use of LLMs." – dyauspitr

2. Credit, Attribution, and the Question of Originality

"If a mathematician solved a major problem by having a long exchange with an LLM ... Would we regard that as a major achievement of the mathematician? I don’t think we would." – pmontra

3. Access Inequality & Budget Constraints

"The $100/month subscription costs less than your office space." – krab

4. Verification and the Limits of LLMs

"I only work with LLMs in domains I’m expert in." – smartmic (emphasizing that expert oversight is still required)

🚀 Project Ideas

VeriMath: Collaborative Proof Verification & Citation Tracker

Summary

  • A web platform where authors upload AI‑generated proofs, automatically tag statements, and invite human experts to verify steps while the system records citation links and verification status.
  • Provides the “repository for AI‑produced results” that the discussion highlighted as missing.

Details

Key Value
Target Audience Academic researchers, PhD candidates, and peer‑review editors who need transparent validation of AI‑assisted mathematics.
Core Feature Step‑by‑step verification UI + automated citation graph that tags each lemma with source references and verification flags.
Tech Stack Backend: Node.js + GraphQL; DB: PostgreSQL + Neo4j (citation graph); Frontend: Vue.js; Verification engine: custom LLM‑checker; Auth: institutional SSO.
Difficulty High
Monetization Hobby

Notes

  • Directly echoes Gowers’ experience where a pre‑print was vetted by a colleague; HN users emphasized the need for rigorous peer‑review of AI‑generated mathematics.
  • Generates discussion on building a trusted audit trail for AI‑originating research, a hot topic among mathematically‑inclined commenters.

EduAI Grader: Adaptive Grading with AI‑Generated Code Detection#Summary

  • A grading assistant that evaluates take‑home assignments, flags excessive AI‑generated code patterns, and requires students to submit “atomic change” logs to prove manual effort.
  • Tackles the grading nightmare described by crocdundae and others who face inflated grades from ClaudeCode.

Details

Key Value
Target Audience University instructors and teaching assistants in programming‑heavy courses (CS, data science, engineering).
Core Feature Automatic similarity detection + “AI‑usage score” that prompts students to annotate modifications and provide version diffs.
Tech Stack Backend: Django + Python; ML component: CodeBERT fine‑tuned for AI‑text detection; Frontend: Angular; DB: MongoDB; CI/CD: GitHub Actions.
Difficulty Medium
Monetization Revenue-ready: per‑course license $250/yr (scalable to department).

Notes

  • Directly addresses crocdundae’s frustration about students bypassing learning via paid AI tools, and the need for a fair grading scheme that still permits AI assistance.
  • Will likely generate lively discussion on balancing AI use with authentic learning outcomes, a recurring theme in the thread.

AI‑Math Repo: Open Platform for Publishing & Auditing AI‑Generated Theorems

Summary

  • A public archive where researchers can publish AI‑produced mathematical results, automatically accompanied by a verification pipeline (prompt logs, model version, validation scripts) and community rating.
  • Provides the “different repository” idea mentioned by rito and the need for a dedicated space for AI‑generated math.

Details

Key Value
Target Audience Pure mathematicians, theoretical computer scientists, and research institutions interested in AI‑assisted discovery.
Core Feature Submission form that captures full prompt chain, modelUsed, and generates a reproducible validation notebook; community can up‑vote and request revisions.
Tech Stack Backend: Flask + Docker; Frontend: Svelte; Validation: Jupyter notebook execution; Storage: S3; Rating: GitHub Discussions‑style.
Difficulty High
Monetization Hobby

Notes

  • Mirrors the sentiment of tag2k and other commenters who want a trusted place for AI‑generated proofs, and would foster the kind of discussion highlighted by multiple participants about transparency and credibility.

Read Later