Project ideas from Hacker News discussions.

UK accounting body to halt remote exams amid AI cheating

πŸ“ Discussion Summary (Click to expand)

1. AI Enables Easy Cheating on Remote Exams, Necessitating In-Person Testing

AI makes remote proctoring ineffective, prompting a return to supervised, paper-based exams.
"AI has taken it to the next level. Now you don’t need to know anything at all." (Aurornis:)
"Remote proctoring has always been security theater... The only reliable 'air gap' from AI today is a physical room and paper." (KurSix:)
Many note pre-AI cheating (e.g., VMs, frat help) but AI scales it massively.

2. AI Reliance Undermines Learning, Producing Unqualified Graduates

Students using AI for exams skip deep understanding, risking job market failure.
"The damage has already been done. Much like how if you stop going gym you lose muscle mass, the same happens with knowledge and understanding with the brain." (design2203:)
"People who have learned how to learn can learn more. People who only used AI never learned how to learn." (nkrisc:)
Concerns include "slop-posting" in forums and real-world incompetence.

3. Certifications' Relevance Shifts in AI Era: From Irrelevant to More Gatekept

Certifications seen as outdated gatekeeping or poised for stricter validation.
"Certifications have always been irrelevant for me... With the advent of modern LLM tech... certifications... will be significantly more relevant... a much more in depth, in person, and gatekeepery certification." (Verdex:)
"We are doing what we can to hang on to relevancy as gatekeepers who already held way too much authority over a field." (random9749832:)
Job market will favor genuine skills over AI-aided certs.


πŸš€ Project Ideas

AI-Proof Proctoring Kit

Summary

  • A hardware-software bundle for secure remote exams using locked-down VMs, VM detection, and real-time AI hallucination checks via multiple LLMs cross-verifying answers.
  • Solves rampant AI cheating on remote professional certifications like ACCA, ensuring genuine competency without full in-person requirements.

Details

Key Value
Target Audience Exam providers (ACCA, universities), certification bodies, employers
Core Feature Browser extension + USB dongle enforces VM isolation, blocks screenshots/copying, runs parallel LLM checks on answers for consistency
Tech Stack Electron.js, WebAssembly for client, Node.js backend, OpenAI/Claude APIs for verification
Difficulty Medium
Monetization Revenue-ready: Subscription ($10/exam seat/month)

Notes

  • HN users lament "remote proctoring has always been security theater" (KurSix) and VM cheating ease (drnick1); this arms race winner with hardware would spark "do things that don't scale" debates (recursivedoubts).
  • High utility for scaling in-person rigor remotely; viral potential via open-source core.

CompetencyForge: Project-Based Cert Verifier

Summary

  • Platform for AI-resistant certifications via timed, real-world projects (e.g., balance a mock ledger with edge cases) submitted as annotated code/videos, auto-graded for principles over rote.
  • Addresses irrelevance of AI-passable exams, proving "real people who actually know the subject" (nkrisc) for fields like accounting.

Details

Key Value
Target Audience Job seekers in accounting/CS, hiring managers skeptical of certs
Core Feature Generates unique scenarios from standards (GAAP/IFRS), rubric-based grading mixing AI pre-scan + human review for explanations
Tech Stack React/Next.js frontend, Python/Django backend, LangChain for scenario gen, Vercel hosting
Difficulty High
Monetization Revenue-ready: Freemium ($49/cert attempt, $99 verified badge)

Notes

  • Echoes "certifications... more relevant... in-depth, in-person, gatekeepery" (Verdex) but digitized; users like general1465 want "real world examples... explain why correct."
  • Fuels HN threads on hiring signals; practical for portfolios beating "bleak" job markets (random9749832).

FlashVerify Study Aid

Summary

  • LLM-powered flashcard generator with built-in accuracy verifier: cross-checks against official sources (e.g., FASB standards) and flags hallucinations, plus spaced repetition with explanation prompts.
  • Fixes "LLM summarisation is broken" (fao_) for exam prep, enabling effective Anki-style learning without "shallow knowledge" pitfalls.

Details

Key Value
Target Audience Students prepping for certs/exams (accounting, CS), self-learners
Core Feature Upload syllabus/test material β†’ generates/validates flashcards β†’ Anki export + quiz mode requiring "why?" explanations
Tech Stack Streamlit/Tauri for app, Hugging Face models + custom scrapers for standards, Supabase DB
Difficulty Low
Monetization Revenue-ready: $4.99/month pro tier

Notes

  • Directly tackles amitav1/warmedcookie's AI flashcard workflow but with "if inaccurate, somebody should let the test know"; baq/samrus praise manual construction benefits, this hybrid scales it.
  • HN loves efficient tools like Anki; great for "using AI to boost programming" courses (seanmcdirmid).

Read Later