Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. DeepSeek Harness (deepseek.com)
    1 pointsby 0 minutes ago
    —discuss
  2. I built the only multi-agent Document Intelligence application. (with MCP) (smarterdeck.com)
    1 pointsby 0 minutes ago
    —discuss
  3. Mitty: Transport-agnostic proxy RPC for executing method chains from any context (github.com/jcubic)
    1 pointsby 1 minute ago
    —discuss
  4. Show HN: Last Internet Connection (reparadoxy.itch.io)
    1 pointsby 2 minutes ago
    —discuss
  5. Who defends Siri? I profiled 121 Reddit accounts that blamed my English (vania-novikau.me)
    1 pointsby 4 minutes ago
    —discuss
  6. 640tok/s on Gemma 4 with a megakernel (withhopper.com)
    1 pointsby 5 minutes ago
    —discuss
  7. Fine-tuning Qwen3.5-4B to replace Gemini Flash-Lite in a RAG pipeline (duane.sh)
    1 pointsby 7 minutes ago
    —discuss
  8. The Book Your Tech Boss Doesn't Want You to Read (bayareacurrent.com)
    1 pointsby 8 minutes ago
    —discuss
  9. GLM-5.3 and the spread of advanced cyber capabilities (anthropic.com)
    5 pointsby 9 minutes ago
    1comments
  10. We build safety, security, and privacy into dots (openai.com)
    1 pointsby 10 minutes ago
    1comments
  11. Tango: Simple AI's Conversational Awareness Model (usesimple.ai)
    2 pointsby 10 minutes ago
    —discuss
  12. A Closed-Loop Architecture for Pan-Cancer Suppression and Rejuvenation Safety (substack.com)
    1 pointsby 10 minutes ago
    —discuss
  13. WSL containers is now generally available (windows.com)
    1 pointsby 10 minutes ago
    —discuss
  14. True Agency Requires Operating Without External Cues (civicpivot.blogspot.com)
    1 pointsby 10 minutes ago
    —discuss
  15. U.S. Media Freedom: Perceptions Hit New Low (gallup.com)
    1 pointsby 12 minutes ago
    —discuss
  16. Learning How to Forget: Fine-Tuning for Long-Context Sparse Attention (arxiv.org)
    1 pointsby 13 minutes ago
    —discuss
  17. OpenAI Answers TypeSafe's Jev with a Decision API Built on Luna (thenewstack.io)
    2 pointsby 13 minutes ago
    —discuss
  18. ChatGPT Pro 500 (help.openai.com)
    40 pointsby 14 minutes ago
    15comments
  19. Performance Improvements in JDK 27 (inside.java)
    1 pointsby 14 minutes ago
    —discuss
  20. Show HN: Greek and Cypriot Greek TTS that runs faster than real time on CPU (huggingface.co)
    1 pointsby 16 minutes ago
    —discuss
  21. The worst problem with Claude and ChatGPT is how we communicate with them (caseorganic.substack.com)
    1 pointsby 16 minutes ago
    —discuss
  22. Infinite Jungle (mo.town)
    2 pointsby 16 minutes ago
    —discuss
  23. GPT-6 UltraFast (developers.openai.com)
    3 pointsby 17 minutes ago
    —discuss
  24. AI Models Fail at Physics: Coding Harnesses Are to Blame (juliahub.com)
    1 pointsby 19 minutes ago
    —discuss
  25. Isotropic Superconductivity in Room-Temperature Superconductor LaSc2H24 (arxiv.org)
    1 pointsby 22 minutes ago
    —discuss
  26. New Features. Insight Frequency and Run the Numbers (ichingportal.com)
    1 pointsby 23 minutes ago
    —discuss
  27. The Next 3x in Inference Won't Come from Faster Kernels (twitter.com/usemuna)
    1 pointsby 24 minutes ago
    1comments
  28. Show HN: A working 3D model of an Enigma machine (enigma.design)
    1 pointsby 24 minutes ago
    —discuss
  29. Is there a lanscaping tool which can take photos of my garden and make 3D model?
    3 pointsby 25 minutes ago
    —discuss
  30. What Was Hidden Under Double D Hat? (noxluneworld.com)
    1 pointsby 26 minutes ago
    —discuss

The Next 3x in Inference Won't Come from Faster Kernels

1 pointsby 25m agotwitter.com
Open original View on HN
1 comments
25m agoHN ↗

One unique idea for solving the chip shortage: pack more models into each existing GPU by having them share the resources. :-)

Open thread on HN ↗