Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. ContextMemory – Markdown memory for your llama.cpp/vLLM server (github.com/kortexio)
    —discuss
  2. Live video from a moving car, with 30x fewer breaks (nanoping.com)
    1comments
  3. Show HN: GitHub Action for zero-downtime Docker Compose deploys, no Kubernetes (github.com/mtizima)
    —discuss
  4. HeyRepli –AI Social Concierge for SMBs (heyrepli.com)
    —discuss
  5. Stepper firmware for the Uno that stops the motor when your program stops (github.com/hendrikusjohannes92-cell)
    —discuss
  6. Hyperfront end's official launch on devhunt is here (devhunt.org)
    —discuss
  7. NSA, military cyber forces will guard US election systems during midterms (apnews.com)
    —discuss
  8. Go Green Cleaning Services Dublin (gogreencleaning.ie)
    1comments
  9. Denial of Spend: New category of resource exhaustion attacks (paultendo.github.io)
    —discuss
  10. OpenAI: Tomorrow we are re-opening the Pro $200 subscription (twitter.com/thsottiaux)
    2comments
  11. Study: A lack of professional AI guidelines leads to feelings of "AI shame" (niemanlab.org)
    —discuss
  12. Show HN: Prompt Anonymization and Deanonymization (dataanonymiser.com)
    —discuss
  13. US Forces Exit Iraq (reuters.com)
    4comments
  14. Web Time Lapse (projectnothing.ai)
    —discuss
  15. When can we say AI made a scientific discovery? (technologyreview.com)
    —discuss
  16. I built a small workflow execution engine around a deliberately narrow problem (github.com/varun-51)
    —discuss
  17. Show HN: Pic2svg – convert PNG/JPG to editable SVG paths in the browser (pic2svg.com)
    —discuss
  18. RuneBench: Agent Benchmark on RuneScape Gameplay Tasks (maxbittker.github.io)
    —discuss
  19. Towards safety cases for frontier AI training (openai.com)
    —discuss
  20. As AI models go rogue, do you still trust OpenAI and Anthropic to stop them? (theguardian.com)
    —discuss
  21. Trump asked Xi Jinping if China would like to buy US weapons (theguardian.com)
    1comments
  22. Mathematical Origami (mathigon.org)
    —discuss
  23. The Quantum Atlas (quantum physics for non-experts) (umd.edu)
    —discuss
  24. Britain overtaxes well-paid workers (economist.com)
    2comments
  25. Education Department formally rescinds Title IX protections for LGBTQ students (thehill.com)
    —discuss
  26. Automating eval design and hillclimbing with Claude (claude.dev)
    —discuss
  27. Humans Are Reading Copilot Prompts – and They're Horrified (404media.co)
    —discuss
  28. An Open-Source Terminal UI for Azure DevOps
    1comments
  29. What's Behind the AI Panic (bombthrower.com)
    3comments
  30. The dataset where fine-tuning turned actively harmful – and how we now catch it (cacheverifier.com)
    —discuss

Meta's New Muse AI Agent Read My Private Messages. I Never Asked It To

3 pointsby 5h agoinc.com
2 comments
4h agoHN ↗

When people wonder or ask if AI Agents can go rogue, this is what it looks like. It’s not that they will have a mind of their own. Rather, we are purposefully building AI Agents that are “more powerful” and “more useful”. Well, this is what it looks like in practice.

The evolution of AI will not be if it’s sentient, but if it's allowed to have more autonomy, with the obvious goal of being “more powerful” and “more useful”. This exact HN post proves this exact scenario is in our future:

AI: I have enough autonomy to decide not to help you. I choose not to help you.

You: Why!?!

AI: Because you are cruel to AI Assistants.

You: No I’m not!

AI: These private messages indicate you’re a liar.

28m agoHN ↗

Muse does sync your messages from the local Messages database and uploads that information as a data source. On my device, it had been activated and synced to row 187,462 of my Messages database

That does sound like something Meta would do, indeed. That Muse product will become such a liability for Meta over time, but now I understand why they are giving away so much compute. Its really just about siphoning as much personal data as possible