wow.. the uncanny valley of motion is closing. 🤯 The humanoid robot by China's Xpeng @XPengMotors From 5 meters your brain says “person”, then remembers it is a robot. Weight shifts, small corrections, no stiffness. A walking beauty 🫡

Singapore’s scientists have developed a breakthrough so simple, yet so powerful: a paint that cools buildings without using a single watt of power. This “smart paint” reflects up to 98% of solar heat, keeping surfaces up to 5°C cooler than ambient air. Made from ceramic nanoparticles and reflective polymers, it forms a passive cooling layer that can last for decades — cutting air-conditioning costs dramatically. In tropical cities, where energy use surges from heat, this could mean millions of tons of carbon emissions avoided every year. The paint works on concrete, metal, and glass, and it’s already being tested on skyscrapers, buses, and even aircraft hangars.

And Robotic hands are also evolving faster than you think 👀

Nano Banana 2 by Google generated this page. But it didn't just generate the web page in the browser. The entire web browser is generated too, and the desktop as well. It's all just one generated image! Impressed?

glycine makes you 2x more beautiful… got pimples, dry skin or redness? you might just be missing this glycine: – reduces inflammation & redness – improves skin hydration & elasticity – builds collagen (your skins building-block) studies show up to 200% more collagen production with consistent glycine intake imagine spending a FORTUNE on skincare when you could pay a fraction for real results.

This is pretty amazing Scientists May Have Found a Way to Rejuvenate the Immune System. New pill is Reversing Immune Aging in Humans in just few weeks. 👀 Swiss biotech Amazentis SA just showed exciting clinical results for its longevity compound Mitopure, a purified form of urolithin A, the so called "youth molecule". In a 4 week human trial, participants taking Mitopure showed rejuvenated immune cells with a younger metabolic signature and stronger cellular activity. The supplement was safe, well tolerated, and previously shown to boost muscle strength in older adults. Scientists believe it may help counteract immune aging by promoting mitochondrial renewal, hinting at one of the most promising anti aging breakthroughs yet. Few days ago skin and heart now immune system. We will be able to rejuvenate all parts of human body by 2030s.

「数える」だけのAIと思うなかれ。 YOLOv12を活用して、錠剤の動きをリアルタイムで追跡。識別IDを保ったまま正確にカウントする。 人間の「数える精度」を上回る仕組みが、もう動いてる。コードは🧵

Grok 4 Fast has a massive 2M token context window that unlocks enterprise-scale reasoning It easily remembers ultra-long chats, huge codebases, or hundreds of pages in a single go The best part? xAI just updated it with massive performance gains: Reasoning mode: 77.5% ➝ 94.1% Non- Reasoning: 77.9% ➝ 97.9% Grok 4 Fast delivers near-flagship results with up to 98% lower cost

This looks fantastic, MicroFactory DevKit for Robots. The big idea is to standardize the physical workcell, then scale imitation learning across identical copies. A single, tightly controlled, exactly reproduced robot hardware cell removes the environmental randomness that usually breaks imitation learning. With the same fixtures, sensors, lighting, and calibration, demonstrations gathered in 1 cell can be reused, and the trained policy transfers to every cloned cell with little extra data. Sharply reduces domain shift and makes data and policies reusable across sites.

Great to see GLM-4.6 powering Cerebras Code. This is exactly why we open weights: so teams can combine their own infra and ideas with GLM capabilities, and bring more choices to developers worldwide. Huge welcome to all partners building on GLM. Let’s grow the ecosystem together.

動画生成AIの「速度の壁」が、またひとつ崩れた。 Bytedanceの新モデル「InfinityStar」は、空間×時間を統一的に扱う8Bパラメータの時空間オートレグレッシブモデル。5秒720p動画をわずか数秒で生成。 VBenchスコア83.74でSOTAを更新。DiT比で約10倍速。リンクは🧵

🚨 Everyone’s obsessed with “new AI regulation.” Oxford just dropped a bomb: we don’t need new rules we already have them. Here’s the uncomfortable truth the world doesn’t need another 500-page “AI safety” manifesto. We just need to use the standards that already keep planes from falling, nukes from melting, and banks from collapsing. Oxford’s latest study says AI isn’t some alien life form that needs an entirely new playbook. It’s just another high-risk technology and we already have global systems for that. ISO 31000. ISO/IEC 23894. The same frameworks that manage billion-dollar risks in energy, aviation, and finance. Translation: instead of reinventing regulation, we should plug AI into the safety infrastructure that already works. Because right now, Big AI labs like OpenAI, Anthropic, and DeepMind are running their own private “safety frameworks.” They decide when a model is too powerful. They decide when to pause. They decide when to release. That’s like Boeing regulating itself after building the 737 MAX. Oxford’s point? Don’t burn the system down — evolve it. Merge AI’s fast-moving internal guardrails with the slow, proven machinery of global safety standards. Make AI auditable, comparable, and accountable like every other life-critical industry. When an AI model suddenly shows new reasoning abilities or multi-domain behavior, that should automatically trigger a formal review not a PR tweet. That’s how you scale trust. That’s how you make AI safety operational, not theoretical. Oxford’s stance is clear: AI doesn’t need prophets. It needs process. 🔥 So here’s the real question do you think governments and AI labs can actually build a shared language of risk, or will they keep talking past each other until it’s too late?

🚨 This might be the most mind-bending AI breakthrough yet. China just built an AI that doesn’t just explain how the universe works it actually understands why. Most of science is a black box of conclusions. We know the “what,” but the logic that connects everything the “why” is missing. Researchers call it the dark matter of knowledge: the invisible reasoning web linking every idea in existence. Their solution is unreal. A Socrates AI Agent that generates 3 million first-principles questions across 200 fields, each one answered by multiple LLMs and cross-checked for logical accuracy. What it creates is a Long Chain-of-Thought (LCoT) knowledge base—where every concept in science can be traced back to its fundamental truths. Then they went further. They built a Brainstorm Search Engine for inverse knowledge discovery. Instead of asking “What is an Instanton?”, you can explore how it emerges through reasoning paths that connect quantum tunneling, Hawking radiation, and 4D manifolds. They call it: “The dark matter of knowledge, finally made visible.” The project SciencePedia already maps 200K verified scientific entries across physics, math, chemistry, and biology. It shows 50% fewer hallucinations, denser reasoning than GPT-4, and traceable logic behind every answer. This isn’t just better search. It’s the first AI that exposes the hidden logic of science itself. Comment “Send” and I’ll DM you the paper.

Woah, this is cool! Thank you for building in the open with the community!

We're already deep in the uncanny valley now.

🚨BREAKING: Elon Musk has dropped a bombshell, publicly calling for the arrest of politicians involved in a massive money laundering scheme of taxpayer money through non-governmental organizations. He calls for Attorney General Pam Bondi to "put the hammer down" on the culprits. Do you support this? IF you support this, Give me a THUMBS-UP👍!

We improved our ARC AGI2 score to 29.72%, which is the new SOTA for the ARC AGI2. This score is not legit for the arc prize competition because we got it after the deadline. It is legit for the ARC AGI leaderboard because the test data is the same. Our solution costs $0.21% per task, when the previous SOTA of 29.40% cost $30 per task. We'll disclose our method on Dec 5th, when the arc prize competition results are unveiled.

the mysterious model on OpenRouter named "polaris alpha" is likely from OpenAI they're following a similar pattern to the GPT-4.1 summer pre-release called "quasar alpha" it's probably a non-reasoning model with a large context window but more leaks include: GPT-5.1 reasoning GPT-5.1 pro

Meta made an acquisition offer to Reseracher for $1 billion – Google offered almost $3 billion. Google’s $2.7 billion acquisition of CharacterAI brought back AI pioneer Noam Shazeer - but his outspoken posts on gender, Gaza, and politics have ignited internal strife. Shazeer’s comments have repeatedly been deleted by moderators, drawing rebukes from top leaders like Jeff Dean and fueling debates over free speech, workplace inclusion, and Google’s evolving culture, says TheInformation. As the architect of major Gemini model breakthroughs, Shazeer’s brilliance now clashes with Google’s push to limit political and divisive speech at work.