Hackernews posts about 35B

1. A 35B MoE on a 16 GB GPU, without the offload tax (www.lucebox.com)
   18 points by GreenGames 14 days ago | discuss
2. Running a 35B MoE model on a 2017 AMD RX 580 8GB via Vulkan (no ROCm/CUDA) (github.com)
   4 points by aivisionslab 1 day ago | discuss
3. Qwen and Fable: An open-weights agentic coding model. 35B Mixture-of-Experts (huggingface.co)
   2 points by thunderbong 5 days ago | discuss
4. Story of How Im Running an Unlimited $6/Month AI Provider on 4x RTX 3090s
   7 points by yolo-auto 8 days ago | 3 comments
5. Ask HN: What are some good/fast coding models for Apple Silicon?
   2 points by LoganDark 6 days ago | 3 comments
6. Scammers keep scoring: Brits fleeced for £1.3B as Americans lose $3.5B (www.theregister.com)
   2 points by Bender 6 days ago | discuss
7. A Comprehensive Analysis of the Open-Weight Frontier (May 2026) (deepresearch.ninja)
   2 points by jackalxyz 25 days ago | discuss
8. Health insurance claim denial rates range from 13% to 35% by insurer (www.randalolson.com)
   65 points by brandonb about 14 hours ago | 65 comments
9. Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup (firethering.com)
   38 points by steveharing1 28 days ago | 31 comments
10. Canada considers cancelling part of U.S. F-35 order to buy 60 Swedish Gripen (www.armyrecognition.com)
    25 points by vrganj 20 days ago | 5 comments
11. What five simple habits can do over 35 years (en.wikipedia.org)
    12 points by frans 21 days ago | 4 comments
12. Karmelo Anthony sentenced to 35 years in TX track meet murder of Austin Metcalf (www.cbsnews.com)
    4 points by donsupreme 12 days ago | 1 comments
13. F-35 versus Gripen E: A deep deep dive (hushkit.net)
    4 points by root-parent 26 days ago | discuss
14. A Transoceanic Jet at 35,000 Feet Is In Airspace That Doesn't Legally Exist (simpleflying.com)
    4 points by Stratoscope 29 days ago | discuss
15. VibeThinker 3B – Taking on Giant Models [video] (www.youtube.com)
    3 points by modinfo 1 day ago | 1 comments
16. Unhealthy code makes AI agents consume 35-50% more tokens (codescene.com)
    3 points by ailinter 24 days ago | discuss
17. 35°C Wet-Bulb – Narration of Robinson's 'The Ministry for the Future' Opening (www.youtube.com)
    2 points by SpiralLibrarium about 18 hours ago | discuss
18. Softmax-free ~354M: tile-skip kernels for long-context VRAM savings (sparse) (huggingface.co)
    2 points by Tripstoph about 21 hours ago | discuss
19. Daveigh Chase, star of Lilo and Stitch and The Ring, dies aged 35 (news.sky.com)
    2 points by austinallegro 3 days ago | discuss
20. Maersk Middle East Operational Update 35 (www.maersk.com)
    2 points by mooreds 6 days ago | discuss
21. Only 35 domains link to all 8 of the top project management tools (crawlgraph.com)
    2 points by pucilpet 18 days ago | discuss
22. Show HN: Lowfat – pluggable CLI filter that saved 91.8% of my LLM tokens (github.com)
    156 points by zdkaster 17 days ago | 80 comments
23. Bad MCP design costs your agent 5x more tokens
    17 points by JohnnyZhang483 17 days ago | 1 comments
24. Show HN: Ustps (UDP Speedy Transmission Protocol Secure) and USSH (github.com)
    13 points by x1colegal 13 days ago | 7 comments
25. Show HN: Levi – run AlphaEvolve on your Claude Code/Codex for dirt cheap (ttanv.github.io)
    5 points by ttanv 14 days ago | discuss
26. Show HN: Libfyaml adds a YAML/JSON parse cache; 427 MB reloads in 36 ms (github.com)
    3 points by fypanto 25 days ago | discuss
27. Show HN: Scholar Sidekick – citation verifier for the "real DOI, wrong paper" (scholar-sidekick.com)
    2 points by ProductivePhys 19 days ago | 3 comments
28. Show HN: Mesmereyes – classic HN demos, mashed up and remade for multitouch (kmewhort.github.io)
    2 points by kmewhort 9 days ago | discuss
29. Ask HN: What do you do when you hit Claude subscription limits?
    2 points by sermakarevich 10 days ago | discuss
30. How I stopped wasting Claude tokens (botverse.cloud)
    2 points by MarkTurnerETech 14 days ago | discuss