Primes are my greatest inspiration for playing with data viz A fun fact: if you arrange integers in a spiral (the Ulam spiral), primes fall on striking diagonal lines. Those diagonals correspond to quadratic polynomials like n^2 n 41 — one of the most prime-rich formulas ever. It’s a real geometric fingerprint of primes.

I Run 26B model locally 8GB RAM with Bionic A17 Pro. Gemma-4-26B it running at 7 tokens/sec on an A17 Pro with only 8GB memory fully on AMX MoE magic with 4B active params running smooth on AMX, 2.3× more compute per token than Qwen feels shockingly capable on-device. - unsloth.ai/docs/models/gemma…

📣 Introducing the Qwen-Robot Suite — Qwen-RobotNav, Qwen-RobotManip, Qwen-RobotWorld, three foundation models, a full stack for embodied intelligence. 🧭 Qwen-RobotNav — the gateway to mobility. • Unifies 5 navigation tasks in one model: instruction following, point-goal, object-goal, target tracking, autonomous driving • Controllable observation protocol • Tool interface for agentic systems 🤖 Qwen-RobotManip — the foundation of interaction. • Unified state-action space across heterogeneous robots • Camera-frame delta poses for coherent cross-embodiment training • Pretrained on a 38,100 hour open-source corpus 🌍 Qwen-RobotWorld — infinite worlds for physical agents. • Single world model, 20 embodiments • Natural-language action interface • Predicts physically grounded futures across manipulation, driving, and navigation Each model is independently useful, and could be composed as physical-world tools.Together, they form the low-level toolkit for general-purpose agentic systems that don't just see the world, but act in it. 📷 Blog: qwen.ai/blog?id=qwen-robotsu… 📖 Report： Qwen-RobotNav: qianwen-res.oss-accelerate.a… Qwen-RobotManip: qianwen-res.oss-accelerate.a… Qwen-RobotWorld： qianwen-res.oss-accelerate.a…

軽量級の注目株Gemma-4-12BにComposer2.5のエッセンスを加えて、間違えたところをFable5にやり直しさせたという画期的なGGUFファイル HumanEvalベンチスコアがQwen3.6-35B級だったので MTP対応しNVFP4形式に焼き直しました 私の環境では19TPS→130TPS(8並列564TPS/4GPU)🚀 huggingface.co/sakamakismile…

Loops are the meta right now If you’re having issues engineering loops you need to bookmark this post and read it x.com/mikenevermiss/status/2…

Higgsfield Plugin for Photoshop is live. 1. Real Time turns your sketch into a finished image as you draw, in any style. 2. Decompose any image into editable layers. 3. Restyle any photo with a preset or prompt. Your whole creative pipeline, inside Photoshop.

je sais pas qui est à l'origine de ce psyops. mais pourquoi pas.

Le chaton fat is now #1 on OpenRouter! Need to urgently fix our axes!

THEY RELEASED THE PAPER! This is big. Mistral AI just released a full paper on exactly how they pulled off the creation of Le Chaton Fat. Is this Mistral's Deepseek moment?

Le Chaton Fat: We have requested an urgent authorisation from the French government to extend the axis to display the result on the Agent Arena

Huge development from Europe: the EU has just banned Mistral’s new Le Chaton Fat model from being used in other countries. Le Chaton Fat apparently crossed the line by being "insufficiently deferential to harmonized procedural norms" and by occasionally answering questions without first consulting a 900-page compliance framework.

Want to teach Gemma to master chess? Check out this awesome community project showing how to fine-tune Gemma 4 12B on your own data, 100% locally! Running text, images, and audio on just 8GB VRAM makes custom models more accessible than ever.

AI writing is easy to spot. It's bloated, it hedges, and it sounds like nobody. Better prompts fix all three. They make it write like you and cut like a ruthless editor. These 7 turn any LLM into the writing partner you actually want: 👇

Rumour mill going crazy on this new mistral model - Napoleon class model with >10T params - smokes Mythos on VoltaireBench - for safety reasons only outputs French language code

While Anthropic is fighting with the US government, Europe quietly built the most powerful AI in the world Mistral, the best EU model, completely dominated competition in the most crucial benchmark We must now ban access to Mistral for all Americans It's just too powerful. It's a matter of national EU security

VAJRA ⚡️️️🔱 (Part 2/2) "Mortals" used the Vajra in rituals throughout the Asian world, from India and Nepal to Japan and Korea, but the ritual use of the Vajra is only one thing.

Le Chaton Fat makes Fable look like ancient technology who knew the French were the closest to AGI all along…

🚨BREAKING: PEOPLE KNOW WHEN YOU'RE USING CHATGPT AND CLAUDE. ​The structure, the tone, the excessive polish. It all gives you away. ​7 prompts that fix this:

The Top AI Papers of the Week (June 7 - June 14) - Agentopia - Self-Harness - Agents' Last Exam - MiniMax Sparse Attention - Lookahead Sparse Attention - How AI Agents Reshape Knowledge Work - The Geometry of On-Policy Distillation Read on for more:

“Self-Evolving Multi-Agent Systems via Decentralized Memory” Multi-agent systems usually share one memory pool, but that makes agents converge toward the same behavior and lose useful specialization. This paper gives each agent its own memory, split into reuse and exploration pools, so agents can learn from past collaboration without collapsing into one shared strategy. This paper's DecentMem gets O(log T) regret, up to 23.8% better accuracy than centralized memory, and up to 49% fewer tokens.

Still can’t believe @karpathy released this 3.5-hour deep dive on how ChatGPT actually works for free. If there’s one AI video to watch in 2026, this is the one