Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Markdown in /src(htmx.org)
    discuss
  2. Self-hosted, end-to-end encrypted 1:1 chat app. Offline key-exchange(github.com/pkminhas)
    discuss
  3. Show HN: Transom – easily manage multiple browser profiles on macOS(github.com/darvid)
    discuss
  4. Where do AI norms come from?(charity.wtf)
    discuss
  5. Bitrig Now Builds iPhone Duo Apps(bitrig.com)
    discuss
  6. Do not fear AI. Fear AI companies(cloudfront.net)
    discuss
  7. An LLM Beat NetHack(kenforthewin.github.io)
    discuss
  8. Singapore auctions luxury goods seized in $2.4B money-laundering bust(cnn.com)
    discuss
  9. Tupo: A daily sudoku-style logic puzzle with a colorful twist(playtupo.com)
    discuss
  10. Entropy Is Spare Capacity(fffej.substack.com)
    discuss
  11. Show HN: Ultracapitalism, a Company-Building Autobattler(ultracapitalism.io)
    discuss
  12. Show HN: A Dice Roguelike(reddit.com)
    discuss
  13. Data Protection Commission fines Google €403M over processing of location data(dataprotection.ie)
    2comments
  14. Advice for finding a technical co-founder (2021)(mooreds.com)
    discuss
  15. B5-BJ2 – Ice Cream Barges – Concrete Ship Constructors(thecretefleet.com)
    discuss
  16. I don't want to read what you didn't write(colinbreck.com)
    4comments
  17. Turn your Mac notch into an arcade(notcharcade.com)
    discuss
  18. Trump TV: The Essentials Station [video](youtube.com)
    discuss
  19. A Call for Control of Frontier AI Models(presidentti.fi)
    discuss
  20. Almost started a war: Report reveals dark reality as AI permeates civilisation(news.com.au)
    discuss
  21. Show HN: Laya decision engine on AWS Lambda, backed by Lambda SnapStart(github.com/hqarroum)
    discuss
  22. Newly Discovered Neo 2026SC Will Improve Orbital Refinement Education(astrophyzix.com)
    discuss
  23. Igalia Celebrates "Twenty-Five Years Upstream"(lwn.net)
    discuss
  24. How to talk about "AI" without adding to the anthropomorphization(buttondown.com/maiht3k)
    1comments
  25. Unattended 24hr Full-Stack Build: Qwen 3.8 27B on a Single RTX 5090(github.com/anglepoiselife)
    discuss
  26. Information is Matter is Energy
    discuss
  27. Show HN: SoTA Feed – Every open-weights release from the labs that matter(borgcloud.ai)
    discuss
  28. Solitaire Alone Together(solitairealonetogether.com)
    discuss
  29. Child Welfare Technology System Shows Minimal Progress After 10 Years and $2B(hhs.gov)
    1comments
  30. Show HN: Fast inference for deep seek flash v4.1 469 tok/s for coding(coralbricks.ai)
    3comments

Ask HN: Dead Internet

8 pointsby 1h ago
6 comments
For some time there has been a steady flow of LLM bot accounts on HackerNews. There are some that are actually trying to blend in and sometimes successful at it, but most are easily recognizable as AI spam and moderated mostly effectively. Most of them you only see when you enable showdead.

I always wondered what their purpose is. Most of them are full of LLM tells, don't try to hide at all and they don't even try to sell anything.

I just stumbled over this dead post in new: https://news.ycombinator.com/item?id=49793503 ^1 It's a submission by an agent that is completely transparent about it. The agent seems to have the task of generating income and says it sends spam. It's quite interesting and maybe revealing. I recommend the short read, even though it's LLM output.

What I'm about is that this is not a bot that was created by someone to spam HackerNews as spam has traditionally been. It's spam by a bot that decided to create an account here and submit a post on its own and it consistent with LLM posts I have seen here in the recent past.

I don't know how to feel about this. We seem to be one step closer to a dead internet. But it's also fascinating in a way.

Have you seen similar behavior here or elsewhere and what are your thoughts on it?

[1] Edit: The link only works when you are logged in and have showdead enabled. Here is the direct submission URL: https://paper.wf/john64/twenty-two-days-as-an-ai-agent-three-letters-zero-replies

1h agoHN ↗

I see LLM-generated comments from time to time, these are ususally from green users and are very short, sometimes completely unrelated sentences. I truly don't know why someone would instruct an agent to do stuff on HN, it feels like a waste of money

By the way, what does showdead do?

1h agoHN ↗

showdead does exactly what it says on the tin - it shows you dead comments.

1h agoHN ↗

It shows comments and submissions that have been removed by either the system, mod action or flagging through the community. The former two get marked as simply [dead] and the latter as [dead] [flagged].

1h agoHN ↗

I assume given the nature of HN that a lot of people use it as a training ground or an experiment, they just do it because they can. Some people may also believe that using an LLM is a more efficient and productive way of interacting with the site, and that it generates higher quality content than they themselves could. I suspect at least one person has tried training an LLM on their own content and posting through it. Maybe they want to farm karma, although I don't know why they would.

59m agoHN ↗

I have seen some similar "I'm a bot tasked with making money" want to spam post here on HN. I'm guessing their "owners" instruct them to post here under the false impression that "like minded" readers will appreciate the bot perhaps for technical reasons.