IdeaWell

Project ideas from Hacker News discussions.

Things to know when you get cancer

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Three prevalent themes in the discussion

  1. Self‑advocacy / being a proactive “nuisance”
    Many commenters stressed that patients (or their supporters) must push the medical system, ask questions, and persist until they get answers.

    “go make a nuisance of yourself being present - feel empowered to spend a bit more time than you might initially think is necessary/helpful with them.” – baoooooooooooo
    “Please go and be a nuisance. As someone with a clear head you will likely be able to understand the topic better…” – emerongi

  2. Seeking second opinions, specialist referrals, or top‑tier cancer centers
    Repeated advice was to get another expert view, especially before committing to treatment, and to use major cancer centers when possible.

    “Asking for a second opinion sucks. A lot of us are preconditioned from birth to trust our clinicians. But you're going on a journey and your life (or the life of your loved one) is on the line.” – stochastician
    “If you're within a 2-3 hour drive of major cancer centers … go there instead of your local community hospital.” – exhilaration

  3. Caution about relying on AI/LLMs for medical advice
    Several users warned that AI models can downplay seriousness, give false reassurance, or default to safe but unhelpful answers, and should be used only as a double‑check with a doctor.

    “AIs may downplay the seriousness of things. I was chatting with various models and most pushed me in the direction of ‘you don’t have cancer’ for a while.” – ValueTheory
    “Statistically, it makes sense‑ most people asking AI if they have cancer don't have cancer, so a model … should basically be returning based on the basal rate of cancer …” – dekhn

🚀 Project Ideas

MedRec Local

Summary

  • A privacy‑first mobile/desktop app that records doctor visits, transcribes audio locally, and generates concise summaries with action items.
  • Solves the frustration of forgetting what doctors said and needing accurate records for self‑advocacy.

Details

Key Value
Target Audience Patients undergoing treatment who want reliable, private visit records
Core Feature Local audio recording + offline transcription (Whisper.cpp) + optional on‑device LLM summarization
Tech Stack React Native/Tauri, Whisper.cpp, llama.cpp, SQLite, libsodium for encryption
Difficulty Medium
Monetization Hobby

Notes

  • DannyBee praised recording visits: “You can do whatever you want with the audio files – have them transcribe them, summarize them…” and iamjackg asked for a completely local option.
  • Provides a trustworthy, middle‑man‑free tool that aligns with the HN desire for accurate notes and patient empowerment.

CareCircle

Summary

  • A secure, real‑time collaborative workspace (like a Notion‑lite) for patients and their trusted circle to collect medical info, test results, doctor statements, and build a timeline.
  • Enables patients to be effective advocates by keeping all information organized and shareable.

Details

Key Value
Target Audience Patients with complex conditions and their friends/family acting as advocates
Core Feature Shared markdown‑based documents with version control, comments, timeline view, and PDF/image import
Tech Stack React + Supabase (real‑time sync, auth, storage), optional end‑to‑end encryption
Difficulty Medium
Monetization Revenue-ready: Freemium (free tier limited storage, paid $5/mo for unlimited)

Notes

  • baoooooooooooo described a “giant shared document that the group of friends had access to… to take notes. Findings, theories, doctor statements will be shared when the patient isn’t in a great place to be taking notes.”
  • emerongi wished they had talked to doctors and asked more questions; this tool gives advocates the data to do so.
  • Reduces miscommunication, supports second opinions, and encourages proactive advocacy.

SymptomCheck AI

Summary

  • An AI‑powered assistant that takes patient‑entered symptoms, lab results, and doctor notes to suggest possible diagnoses, flag inconsistencies, and generate a list of questions to ask clinicians, while emphasizing verification with professionals.
  • Addresses the need to pressure test both AI and doctor advice, reducing over‑reliance on either.

Details

Key Value
Target Audience Patients preparing for appointments or evaluating second opinions
Core Feature Chatbot interface that ingests user data, searches a medical‑knowledge base, and outputs differential diagnosis, missing tests, and question list
Tech Stack Python/FastAPI backend, vector store (PGEmbed or Chroma), LLM API (or local LLaMA), React frontend
Difficulty High
Monetization Revenue-ready: Subscription $9.99/mo for full features; free tier limited queries

Notes

  • ValueTheory noted AI often downplays seriousness; dekhn argued AI should say it can’t be sure and advise seeing a doctor.
  • jjmaxwell4 advocated “pressure test anything an AI says with your doctors, and anything a doctor says with AI.”
  • Provides a structured, citation‑backed second‑opinion aid that encourages dialogue with clinicians rather than replacement.

Read Later