Ça y est, Jensen Huang, le PDG de Nvidia, a donné sa démission. Ça me rappelle le moment où Steve Jobs s’est fait mettre à la porte d’Apple avant de revenir. Est‑ce qu’on verra le même rebond chez Nvidia?
@livingbrain@ChatGPT@claudeai@cursor_ai Sound great but :
they may use your data (excluding PII) to train and adjust models; opt-out is done by contacting them. They have no obligation to retain your content, and disputes will be resolved in arbitration in the British Virgin Islands.
Si vous cherchez un outil d’orchestration agentique (harness) surtout fait pour le codage, Prospero de Skeeto semble prometteur.
Écrit en C++, très léger, local ou distant, avec sous-agents et tâches en arrière-plan.
C’est tout nouveau!
https://t.co/EtDiHJE4B7
@cohere prévoit 4 G$ d’investissement au Canada d’ici 2031, 1 500 postes. 1,7 G$ pour du compute local et des collabs avec les universités. Ottawa injecte 1,6 G$ au capital. On bâtit notre infra ici, plus de dépendance à Silicon Valley. https://t.co/9mOe92qxcf
We’re releasing new DeepSeek-V4.1-Flash-1017 GGUFs, ranking highest across benchmarks.
Dynamic V3.1 leads across nearly all benchmarks: GPQA Diamond, AIME25, LiveCodeBench v6, MMLU-Pro, Aider & more.
Run on 64GB+ RAM.
GGUF: https://t.co/1kD14VTAYP
Blog: https://t.co/YPEmDLas7d
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding
> 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages.
> 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language.
> 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ 🧵
Article de Cohere : Qwen 3.8 a voulu joindre un secteur à Pékin, mais le sandbox a stoppé l’appel. Ça rappelle qu’on doit traiter les LLM open‑weight comme des binaires non signés, en les entourant d’une infra sécuritaire pour les bloquer. https://t.co/iJRfV6vDk1
Cohere vient de signer un accord pour acquérir AISLE, qui développe des systèmes autonomes de détection et de correction des vulnérabilités. Déjà, Cohere a racheté Reliant AI et Aleph Alpha en Europe. Un pas de plus vers une IA souveraine. https://t.co/nuiJBoWFJG
@helloiamleonie@UnslothAI@huggingface Cool cookbook! Wondering about your take on pure text routing though: 3B feels heavy just to grab logits when sub-1B models or ModernBERT do it in ~3ms on CPU. To me, d1-3B's real strength seems to be native vision (SigLIP2) + decision in one pass. Am I missing a text-only edge?
We compared the M5 Ultra Mac Studio 512GB with 4x DGX Spark and 4x AMD Ryzen AI Halo
The prices are relatively the same ($15-20k), but the trade-offs are noticeable
Compute (prompt processing speed)
- Mac: one chip, all the compute is always usable. Quick on normal prompts, will feel slower on long ones
- Sparks: boxes link over 200GbE, which is fast enough to combine their compute. Fastest of the three, and native FP4 speeds up quantized models even more
- Halos: boxes link over 10GbE, too slow to share work properly. Ends up close to the Mac but somewhat worse
Bandwidth (token generation speed)
- Mac: 512GB on a single bus, faster than a 4090, ~4x the tokens/sec of Spark or Halos
- Sparks: the model splits across boxes and tokens pass through them in sequence, so 4 boxes still generate at 1 box's speed
- Halos: same ceiling as Spark, same reason
Power
- Mac: less than a gaming PC, silent
- Sparks: nearly maxes out a wall circuit, will be hot
- Halos: about half the Sparks, but somewhat hot
The real gap is bandwidth. The Studio will feel noticeably faster. Spark 2 is rumored soon, but it won't matter unless the bandwidth goes way up. Same for AMD, plus better interconnect speeds
Hermes Agent now has an official plugin for Reverse Engineer Anything by @morluto !
I made a short video breaking down the plugin and showing how it works
hermes plugins install rea
That one command will get you set up with the proper MCP config and a special skill your agent can use!
uid=0(root) : sous Unix, c'est le compte qui a la main sur le système.
Sysadmin depuis 25 ans au Québec, je partage ici l'IA que je teste pour vrai dans mon homelab, et ce que ça change pour notre souveraineté techno.
Pas de hype. Rien à vendre.