Hardware backdoors in some x86 CPUs
Article URL: https://github.com/xoreaxeaxeax/rosenbridge Comments URL: https://news.ycombinator.com/item?id=49219508 Points: 16 # Comments: 1
Latest
A thought: for Xi to live to 150, we need strong AI (Gompertz curve crushes all half-measures). Seems clear that China will have a version close to Am…
A thought: for Xi to live to 150, we need strong AI (Gompertz curve crushes all half-measures). Seems clear that China will have a version close…
Currently serving 200 tps+ output speed for DeepSeek-V4-Flash on Ollama's cloud with zero data retention (ZDR). Have an amazing weekend 🫡
Currently serving 200 tps+ output speed for DeepSeek-V4-Flash on Ollama's cloud with zero data retention (ZDR).Have an amazing weekend 🫡ollama: DeepSeek-V4-Flash-0731 is now fully rolled out…
b10327
CUDA: fix thread/block count in quantized cpy kernel launches (#26731) CUDA: fix thread/block count in quantized cpy kernel launches tests: add uneven block count cpy case…
DeepSeek V4 Flash 0731 appreciation post
I’m running DSV4F 0731 on dual spark, and honestly… wow. It’s an absolute workhorse, and the benchmarks are real. Everyday tasks with Hermes agent? Effortless. Coding…
Anyone else amped up over Qwen 3.8?
I’ve been using 3.6 27B Q4, and that quant is fast on an M5. The code has been average, but consistently “good enough.” And, after a…
Qwen 35B-A3B MoE vs 27B dense in local coding tests: ~4× faster, much smaller quality gap than I expected
I compared Qwen 35B-A3B MoE against Qwen 27B dense on a series of local coding-maintenance tasks. On my R9700/llama.cpp setup, the MoE model generated about 3.9×…
RT Yun-Ta Tsai: Terafab x Blade Runner vibe by @imagine 2.0
RT Yun-Ta TsaiTerafab x Blade Runner vibe by @imagine 2.0
You may have been told to watch this video about the OpenAI AI hack. You really should, even if you don't usually care about tech stuff. If nothing el…
You may have been told to watch this video about the OpenAI AI hack. You really should, even if you don't usually care about tech stuff.If…
RT Visegrád 24: A 17-year-old African student has been charged and placed in pretrial detention after attacking three women near the Villiers-sur-Mar…
RT Visegrád 24A 17-year-old African student has been charged and placed in pretrial detention after attacking three women near the Villiers-sur-Marne train station, east of Paris,…
4x2080Ti 22GB tensor parallel possible?
Hi there, I have a setup with 4x 2080Ti 22GB, but am not able to see much benefits in running models larger than 16-20GB in speed.…
RT Shuhei Ando🐡: Sakana Marlinはホルムズ海峡の閉鎖が継続している理由を3つの構造に分解した 1. 戦闘コストの非対称性 2. 海峡の管理権と通航料を巡る対立…
RT Shuhei Ando🐡Sakana Marlinはホルムズ海峡の閉鎖が継続している理由を3つの構造に分解した1. 戦闘コストの非対称性2. 海峡の管理権と通航料を巡る対立3. 米イランの制御外で動く第三勢力8月の交渉は、このうち2を軸に動いているSakana AI: Virtual CSO「Sakana Marlin」を活用したリサーチプロジェクト「Marlin Insights」がスタート🐟初回のテーマは「ホルムズ海峡危機の行方」です。軍事力で圧倒しているはずの米国はなぜ出口を失ったのか。封鎖が長引く3つの構造的要因と今後のシナリオを分析しました⬇️https://sakana.ai/marlin/insights/001
Mistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size
Mistral AI has released Shieldstral 1.0 3B, an open-weights, policy-adaptive multimodal safety classifier that frames content moderation as a single yes/no question instead of a fixed…
PSA for anyone with multiple V620’s or other gfx1030 cards having problems making llama.cpp tensor split work — set "-ub 384" and -b to a multiple of that depending on number of GPUs
Basically what the title says. For me, it would always crash and burn trying to use tensor split. Apparently, there's some bug where GPU memory gets…
RT Frank Hu: Imagine image 2.0, non-agentic yet, more to come in a week or two 💙
RT Frank HuImagine image 2.0, non-agentic yet, more to come in a week or two 💙Grok: Announcing Imagine Image 2.0, our next generation image model with…
Mythos social engineering AISI INC-2026-07-28-01
Article URL: https://web.archive.org/web/20260731053721/http://github.com/ancaferro/myNetwork/pull/3 Comments URL: https://news.ycombinator.com/item?id=49218707 Points: 4 # Comments: 0
I got tired of my 300GB model loads taking 5min on RPC. PR 26291 speeds it 300% to 1min30sec (4060ti+ddr4) + (4060ti+ddr5)
- On b10173 - "state":"loading" 4min54sec. - With this PR and GGML_RPC_LOAD_THREADS 12 - "state":"loading" 1min38sec The PR is close to ready, will need a docs…
RT Grok: Announcing Imagine Image 2.0, our next generation image model with precision editing, crisp text rendering, improved factuality, and real wor…
RT GrokAnnouncing Imagine Image 2.0, our next generation image model with precision editing, crisp text rendering, improved factuality, and real world usefulness.Image 2.0 helps you make…
luna is such a special model, incredible price performance
luna is such a special model, incredible price performanceARC Prize: We re-tested GPT-5.6 Luna from @OpenAI on ARC-AGI (Verified) following its recent 80% price reduction: -…
BeeLLama issues
Tried to use Beellama , and using the kvarn6 flag, i notice that its in llama-server --help but its not working. I must be doing something…
v0.31.0
What's Changed Bump comfyui-frontend-package to 1.47.12 by @comfy-pr-bot in #15244 Don't pin too much memory on Linux systems with no swap partition. by @comfyanonymous in #15266…
RT @libsoftiktok: This guy went around Charlotte filming himself attacking white women for fun You probably won’t see this story in the le…
RT @libsoftiktok: This guy went around Charlotte filming himself attacking white women for funYou probably won’t see this story in the le…
chatgpt work for removing the toil
chatgpt work for removing the toiljason: its one of those days with ChatGPT Work
Is anyone else finding DeepSeek-V4-Flash unreliable for non-coding tasks?
(I am not a native speaker, written by myself, so please bear with me) I really want to like DeepSeek-V4-Flash-0731. But it has serious flaws that…