Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Toxic bosses: why do companies protect managers who make working life hell? (theguardian.com)
    —discuss
  2. Show HN: Multi-agent coding feels like pre-GitHub development (github.com/motif-labs)
    —discuss
  3. Laya "System 1" Code Intelligence for AI Agents (github.com/leonardoventurini)
    —discuss
  4. Full iPhone 4 'Antennagate' Q&A Video Surfaces After 16 Years – MacRumors (macrumors.com)
    —discuss
  5. Jamie Beaton might end up the world's most educated person (nytimes.com)
    1comments
  6. Vitvm: Git for VM State (github.com/numinous-technology)
    —discuss
  7. Docker releases cloud sandboxes, enabling safe agentic workloads in the cloud (docker.com)
    —discuss
  8. Gmux: Tmux for GPUs (github.com/numinous-technology)
    —discuss
  9. Platform-Independent SIMD in Go (go.dev)
    —discuss
  10. Protecting our knowledge of human genetic diversity (projecttimestamper.org)
    —discuss
  11. Show HN: Jev-pilot – Jev picks Claude Code's effort, model and skill per prompt (github.com/akramovic1)
    —discuss
  12. Preventive Software Maintenance (poolp.org)
    —discuss
  13. 4% of Sessions, 65% of the Bill: 30 Days of Claude Code Telemetry (promptarmor.com)
    —discuss
  14. Back and shoulder surgery is often worse than useless (economist.com)
    —discuss
  15. Software-only quantum tool, it breaks RSA-2048 live (dtec9.com)
    —discuss
  16. Danish Health Minister summoned over secret Wegovy court proceedings (dr.dk)
    1comments
  17. Which AI model is best for coding? A hands-on comparison (flaviocopes.com)
    —discuss
  18. C74-6502 Overview (c74project.com)
    1comments
  19. Grieving the Loss of Details (purplesyringa.moe)
    —discuss
  20. Dolphin Emulator Progress Release 2609 (dolphin-emu.org)
    —discuss
  21. Sir Ranulph Fiennes kept in care home illegally (telegraph.co.uk)
    —discuss
  22. Bluh (vivico.space)
    —discuss
  23. Amiga Screens: A Primer (datagubbe.se)
    —discuss
  24. Cheap and Awesome (praf.me)
    —discuss
  25. Adam Conover explains how YouTube ruined everything (theverge.com)
    —discuss
  26. Google's first Suncatcher orbital data center test launches October 1 (arstechnica.com)
    —discuss
  27. Pick OS Is a Living Fossil of Computer History (csixty4.medium.com)
    —discuss
  28. Ask HN: IEEE Expulsion of Fei-Yue Wang?
    —discuss
  29. Redditor Connects an AMD GPU to a MacBook Pro (reddit.com)
    —discuss
  30. More than $170M in crypto moves from Bitget wallets to unidentified address (theblock.co)
    —discuss

Claude can't stop itself from writing code comments

1 pointsby 3h agobayes.net
2 comments
2h agoHN ↗

It's all just training.

LLMs are trained to reproduce what they were trained on. They are trained on tons of source code ... with comments.

It doesn't matter what you put in context, because what you put there will be in direct conflict with what the model "wants" to do (ie. what the patterns in the data it was trained on did).

This is the same reason why it is SO FUCKING HARD to make Claude shut up! Everything in it's training is focused on producing text, so it does it's damnedest to try and produce text.

1h agoHN ↗

That's literally the only language it understands. Text and more text and text again...