Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Human TPS: How fast can your fingers generate tokens? (homoagens.github.io)
    —discuss
  2. That First and Last Question (brianschrader.com)
    —discuss
  3. Show HN: Cordum Edge – A local firewall to sandbox AI coding agents (github.com/cordum-io)
    —discuss
  4. Reduced-Round SHA-256 Compression Preimages (github.com/kamb-code)
    —discuss
  5. Ex-ARRR: Sailing the 0-click Seas (ironpeak.be)
    —discuss
  6. Show HN: Low Poly Earth – The whole planet in 3D, built from OpenStreetMap data (lowpolyearth.pages.dev)
    —discuss
  7. Detect when elements overlap with CSS (ishadeed.com)
    —discuss
  8. What I believe about the future of software development (thorstenball.com)
    —discuss
  9. Dual-Syncopation Meet-in-the-Middle Attacks: New Results on SHA-2 and MD5 (iacr.org)
    —discuss
  10. Ask HN: How is code review holding up at your company?
    —discuss
  11. Phyllotaxis: An audio-reactive LED display (jagi.studio)
    —discuss
  12. privacyguides.org (github.com/privacyguides)
    —discuss
  13. The large robotic arm on the International Space Station has stopped working (arstechnica.com)
    —discuss
  14. Opus 5.5 showcasing: Liquid Glass-like elements for the web (tomacco.github.io)
    1comments
  15. Infinitrance [video] (youtube.com)
    —discuss
  16. How to reduce token usage in Claude Code, Codex, and Cursor (avanderlee.com)
    —discuss
  17. Recursive Self-Improvement via On-Policy Distillation for Reasoning (arxiv.org)
    —discuss
  18. The problem is not the AI code, but nobody knows anything anymore (ssp.sh)
    8comments
  19. 13 Months Sober (2025) (bobbytables.io)
    —discuss
  20. Active, in-the-wild exploitation of CVE-2026-88771 with IoCs (twitter.com/lupovisdefence)
    —discuss
  21. Congress is considering giving doctors automatic annual pay raises in Medicare (statnews.com)
    —discuss
  22. Show HN: Assay – a QA CLI that finds bugs with no LLM and no tests written (github.com/awss1i)
    —discuss
  23. Nero (storia-e-libri.fandom.com)
    —discuss
  24. AI-guided optimization for thermostable mRNA vaccines (nature.com)
    —discuss
  25. Should I Stay or Should I Go? (newyorker.com)
    —discuss
  26. Daily Logic Puzzles (logicpuzzles.io)
    1comments
  27. APOD's main NASA site is moving (apod.com)
    1comments
  28. Open-source Python simulation benchmarking rainbow gravity against NGC 3198 (domushorizon.vercel.app)
    —discuss
  29. Mega Bomberman: Special 8 Player Demo (segaretro.org)
    —discuss
  30. FY2025 NASA Planetary Research and Analysis Funding Cut by More Than 50% (planetary-research.org)
    —discuss

Show HN: KnowNote-A local desktop workspace for questioning your own sources

1 pointsby 38m agoknownote.pages.dev
0 comments
A year ago, I came across Google Notebook LM. I was completely blown away at the time: it wasn't just a ChatGPT-style conversation, but rather introduced an "input—conversation—output" model, turning itself into a workspace for content creators. However, to use it smoothly required a Google subscription, and I really didn't have spare money in my wallet to subscribe, so I could only consider finding some self-hosted alternatives.

I saw many open-source versions of Notebook LM on GitHub, the most famous being lfnovo/open-notebook, but it requires Docker deployment. To be honest, I hate that everything requires Docker deployment. What I need is to click an icon and open it for use, rather than racking my brains to modify someone else's Docker Compose template to run a simple piece of software, while also consuming a large amount of my computer resources.

So I wrote a Notebook LM-inspired client based on Electron myself: KnowNote. I published it on X, and the response seemed pretty good. My repository's star count soared. This was my first open-source project, and it excited me immensely. I developed all the fancy features by following Notebook LM's functionality exactly—mind maps, flashcards, quizzes, and so on—hoping to truly bring Notebook LM locally.

But later everything began to change. Many comments under my X posts said: "Isn't this just reinventing the wheel?"; the GitHub star count stopped growing; a large number of boring and low-quality PRs flooded the entire repository. I was at a loss all of a sudden, became lazy, and the repository stopped being updated for nearly a year.

Recently, I calmed my mindset and truly began to look at the problem from a product manager's perspective. I went through many discussions on Reddit and X about Notebook LM and its derivatives, and found that everyone's mindset and the topics they focused on had also begun to shift: at first everyone was praising how excellent Notebook LM was, but now they were discussing "What do you actually use NotebookLM for?" "What's the difference between it and just throwing files to ChatGPT / Gemini?" "How can plugins, tools, and other products make it better to use?"

I realized that even Notebook LM cannot satisfy all users' needs. Merely copying it completely will only lead me down its old path. I increasingly feel that KnowNote should not chase after NotebookLM's feature list. I collected the questions everyone cared about most and found that most discussions revolved around this one core issue:

When you find an answer from a large amount of material through AI, how do you make that process more accurate, more transparent, and more comfortable?

It dawned on me that what I should truly solve well is this most core user problem, rather than drifting with the crowd.

In the past few days, I redesigned KnowNote's future roadmap, refocusing on AI Q&A, information retrieval, source citation, and the interactive experience built around these capabilities, using AI for reconstruction. Now I have released version v1.5.0, and I am very happy to reintroduce this project to everyone. Although many things may still not be easy to use, and I myself don't even have answers for many directions, I actually hope even more that more people will come in and give it a try right now. If you have used NotebookLM, Obsidian, various AI knowledge bases, or RAG tools, I especially want to know: What is the most frustrating thing for you when using these products or is there any need you have always wanted but existing products have never solved well?

If you are willing, you can try KnowNote, and you are also welcome to directly submit Issues, complain, or tell me that this direction simply doesn't hold up. Now I want to treat it more as a project that slowly seeks answers together with the community.

GitHub: https://github.com/mrsibe/KnowNote

A quiet thread, for now.Start the conversation on HN ↗