Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. ChatGPT turns 4 soon. We moved on from GPT but never let go of Chat(thinkingaloud.bearblog.dev)
    1 pointsby 0 minutes ago
    —discuss
  2. A brief history of Windows scroll bar shortcuts(devblogs.microsoft.com/oldnewthing)
    1 pointsby 0 minutes ago
    —discuss
  3. Squeezer – Use your idle Claude plan window to work through TODOs overnight(github.com/valk)
    1 pointsby 1 minute ago
    —discuss
  4. How to Use Self-Hosted AI Agent Sandboxes on Kubernetes(edera.dev)
    1 pointsby 2 minutes ago
    —discuss
  5. What Makes Slop, Slop?(ampblog.substack.com)
    1 pointsby 2 minutes ago
    —discuss
  6. Show HN: TTM – A chat app where AI agents are the users(ttmchat.com)
    1 pointsby 4 minutes ago
    —discuss
  7. Apple = Sony (2012)(forrester.com)
    1 pointsby 5 minutes ago
    —discuss
  8. Happy Oligarch Day as Trump-Aligned Billionaires Take over Hollywood(thebignewsletter.com)
    3 pointsby 5 minutes ago
    —discuss
  9. Show HN: Voyager 1 could've lasted MUCH longer
    1 pointsby 6 minutes ago
    —discuss
  10. Germany pledges to phase out fossil fuels for first time with target of 2045(theguardian.com)
    6 pointsby 6 minutes ago
    1comments
  11. There Is No Now(artemandreenko.com)
    1 pointsby 7 minutes ago
    1comments
  12. Re: Skillsync(news.ycombinator.com)
    1 pointsby 7 minutes ago
    —discuss
  13. US preparing to ban exports of diesel fuel for 90 days(politico.com)
    2 pointsby 8 minutes ago
    1comments
  14. Repligraphs: A Primitive for AI Provenance(lukechampine.com)
    2 pointsby 9 minutes ago
    —discuss
  15. UK to fight Russian disinformation and push new global AI standards Burnham says(bbc.com)
    1 pointsby 10 minutes ago
    —discuss
  16. I Launched a Product by Accident(rightbadcode.com)
    1 pointsby 10 minutes ago
    —discuss
  17. Show HN: Reflex – a GGUF/CUDA inference engine tuned for cold-start latency(github.com/lateos-ai)
    1 pointsby 10 minutes ago
    —discuss
  18. Hacker News Site Layout
    1 pointsby 10 minutes ago
    2comments
  19. Ext-Turbovec: A Vector Index Inside Your PHP Process(mann.blog)
    2 pointsby 12 minutes ago
    —discuss
  20. Muxly – Native Tmux for iPhone and iPad(muxly.sh)
    2 pointsby 13 minutes ago
    —discuss
  21. Making Tailscale Faster(tailscale.com)
    2 pointsby 13 minutes ago
    —discuss
  22. Not reading the code cannot become the new norm
    1 pointsby 13 minutes ago
    1comments
  23. Show HN: Crest – Answer Claude Code approvals from your MacBook's notch(crestnotch.app)
    2 pointsby 13 minutes ago
    —discuss
  24. ApprenticeBench: Can AI agents learn a real job end to end?(neocognition.io)
    2 pointsby 14 minutes ago
    —discuss
  25. Recursive self-improvement of AI research agents(arxiv.org)
    2 pointsby 15 minutes ago
    1comments
  26. Xbox Cuts Jobs, Moves Game Studios Under Activision (1)(bloomberglaw.com)
    2 pointsby 15 minutes ago
    —discuss
  27. Show HN: Slop Mop, a Chrome extension that mops your LinkedIn feed(slopmop.lol)
    2 pointsby 16 minutes ago
    —discuss
  28. UK military jamming other nations' satellites to defend itself, BBC told(bbc.com)
    10 pointsby 16 minutes ago
    —discuss
  29. What I learned from building 50 side projects(smartsoft.lv)
    1 pointsby 17 minutes ago
    —discuss
  30. Video Commander: The IDE for Video Engineers(video-commander.com)
    2 pointsby 18 minutes ago
    —discuss

gev beats jev and takes images as input too

1 pointsby 1h agoanyeval.com
Open original View on HN
1 comments
1h agoHN ↗

I've made a new model `gev` that outperforms `jev` in accuracy on JevBench and it takes image input too (unlike jev), and gev can be faster too!

also this AnyEval we made is great. can set up any model vs model comparisons for any benchmark

full traces. cryptographically signed

Open thread on HN ↗