Project ideas from Hacker News discussions.

College instructor turns to typewriters to curb AI-written work

📝 Discussion Summary (Click to expand)

5Prevalent Themes in the Discussion

# Theme Supporting Quote
1 Traditional in‑person exams are still valued, and AI is seen as a threat to that model zamadatix: “The point is more about whether the graded work is actively reviewed than which individual choice is ideal or not though. Whether it’s electronic or written, remote or in‑person, weighted towards exams vs continuous are all orthogonal debates to the problem of cheating/falsely claiming work.”
2 Memorization & foundational knowledge remain essential acbart: “How can you debug a complex application if you have to keep looking up every operator and keyword in the language you’re using? It’d be like trying to interpret poetry in a foreign language but you have to look up every single noun… Some amount of memorization is key.”
3 AI makes cheating easier and forces a rethink of assessment integrity al‑Khwarizmi: “If poor students have capable models but rich students have much better models that go the extra mile for a great mark and do everything in a single prompt, it would still be unfair.”
4 Shifting from high‑stakes exams to project‑based or continuous assessment recursivedoubts: “I now do 50% project work, 50% in‑person quizzes, pencil on paper on page of notes… I’m increasingly going to paper‑driven workflows.”
5 Slow, deliberate writing (typewriters, handwritten work) promotes deeper thinking eichin: “One of the reasons I’ve always encouraged software people to learn to touch type has nothing to do with typing speed—it’s about reducing/eliminating the cognitive load of typing, you want to be thinking in expressions (sentences) not letters.”

All quotations are taken verbatim, with double quotes and author attribution, and HTML entities have been corrected.


🚀 Project Ideas

PencilCheck

Summary

  • Capture handwritten exam responses via digital pen, automatically verify they weren’t AI‑generated.
  • Provide instant authenticity scoring for instructors, eliminating the need for manual review.

Details

Key Value
Target Audience University instructors and exam proctors
Core Feature Handwritten exam capture with AI detection of AI-generated content
Tech Stack Electron front‑end, Python backend, OpenCV pen‑stroke analysis, TensorFlow detection model
Difficulty Medium
Monetization Hobby

Notes

  • HN commenters often lament AI cheating; this tool directly solves that pain point – “the only way to prove you actually wrote it.”
  • Could spark discussion about reviving paper‑based verification while leveraging AI for integrity checks.

MasteryPath

Summary- Continuous micro‑assessments unlock subsequent course material only after meeting mastery thresholds.

  • Reduces reliance on high‑stakes final exams and rewards consistent learning.

Details

Key Value
Target Audience Instructors designing competency‑based curricula, students
Core Feature Adaptive learning path that gates content on proven mastery
Tech Stack React front‑end, Node.js API, PostgreSQL DB, ML recommendation engine
Difficulty Medium
Monetization Revenue-ready: Subscription $5 per student per month

Notes

  • Commenters praise any system that moves away from “one‑shot” exams – they’d love a platform that makes learning the metric.
  • Potential to generate debate about replacing final exams with incremental mastery.

TypeTrap

Summary

  • Enforces typewriter‑style constraints (no backspace, limited edits) to force original composition.
  • Logs any AI‑assisted writing attempts for later review by instructors.

Details

Key Value
Target Audience Writing‑intensive courses, liberal arts professors
Core Feature Constrained editor that logs AI tool usage and prohibits backspacing
Tech Stack Web (TypeScript), Monaco editor wrapper, serverless audit log
Difficulty Low
Monetization Hobby

Notes- HN users nostalgic for typewriters would appreciate a modern twist that also combats AI plagiarism.

  • Sparks conversation about the pedagogical value of “slow writing.”

LiveRecall

Summary

  • Recorded oral exam sessions with AI‑transcribed answers and random follow‑up questions.
  • Provides instant competency scoring while preserving the human interaction of oral exams.

Details

Key Value
Target Audience Instructors of courses that rely on oral assessment, language programs
Core Feature Proctored oral exam platform with AI transcription and competency scoring
Tech Stack WebRTC, Node.js, Whisper transcription, GPT‑4 based evaluation
Difficulty High
Monetization Revenue-ready: $20 per examined student

Notes

  • Commenters who miss face‑to‑face interaction would value a tool that keeps the personal touch but adds analytics.
  • Could generate debate about authenticity vs. automation in oral assessment.

ExamCraft

Summary

  • Generates exam questions explicitly designed to be unanswerable by LLMs without foundational knowledge.
  • Auto‑creates answer keys with explicit retrieval cues to test true understanding.

Details

Key Value
Target Audience Course designers, university assessment teams
Core Feature Constraint‑based question generator producing LLM‑resistant exam items
Tech Stack Python, GPT‑4 API, SAT solver, relational DB
Difficulty High
Monetization Revenue-ready: Tiered $50/month for institutional licensing

Notes

  • Directly addresses concerns that “exams are artificial” by producing questions that force recall, not lookup.
  • HN community would likely discuss the merits of LLM‑proof assessments and their impact on credential signaling.

Read Later