When they see you start something they'll tell you DON'T DO IT!, when they see you doing it they'll try to tell you HOW TO DO IT!, and when you succeed they'll tell you that THEY ALWAYS BELIEVED IN YOU. This is the best example and congratulation @elonmusk

Well today is the WorldCup 26 and at the same time the NBA Final are active. Which one bring more fans around the world (Onsite and online)?

Destroy yours legs and thank me tomorrow

The last definition of a master class, will see what happen in the next days

Monaco must be a mandatory place for visit once in your life

30 minutes of video. Robot learns the task. Open-source, end-to-end. An open-source framework for training robot policies from only 30 minutes of human egocentric videos captured via Meta Aria glasses: Achieving zero-shot transfer to robots without any robot data collection. The method relies on Interaction-Centric Tokens that encode hand-object spatial relationships invariant to embodiment and viewpoint, supplemented by auxiliary objectives like object motion prediction and latent consistency to extract richer supervision signals from the same data. HumanEgo demonstrates strong cross-embodiment, cross-environment performance on bimanual tasks, outperforming baselines like ACT and teleop data while being trainable on a single RTX 4090 GPU. Thanks for sharing, @TX_Leo_Wang. 📌 Website: humanego-ai.github.io Paper: arxiv.org/pdf/2605.24934 Code: github.com/TX-Leo/HumanEgo Video: youtu.be/pdL46diijuY ——- Weekly robotics and AI insights. Subscribe free: 22astronauts.com

Weekly Robotics 361: 🤖 DIY Disney's inspired BD-X droid, 🤖 state estimation in contact-rich scenarios (very cool!), 🤖 reverse engineering Unitree Go2 motors, 🤖🤖 ROS 2 Lyrical released, ... and more! weeklyrobotics.com/weekly-ro… #robotics #ros

If you want to get into training robots badly: You can start training real robots at home for under $300... Thread 👇 (with links) Here’s the exact low-barrier roadmap the community is using right now to go from zero to running ACT policies on a real arm in weeks. Thread 👇 1/ Simulation First (Zero Cost) Start in MuJoCo... honestly: the gold-standard physics engine used by DeepMind and every serious robotics lab. Train policies, test behaviors, and iterate insanely fast before touching hardware. → Official site: mujoco.org/ → GitHub (install in minutes): github.com/google-deepmind/m… Pro tip: MuJoCo runs perfectly on a normal laptop. 2/ Grab the $200-300 Hardware Everyone’s Using The SO-101 (also called SO-ARM101): the official open-source arm designed by Hugging Face’s LeRobot team. 3D-printed, easy to assemble, and built specifically for LeRobot ACT training. Buy options: → WowRobo DIY/Assembled Kit (~$199): shop.wowrobo.com/products/so… → Full Hugging Face SO-101 docs parts list: huggingface.co/docs/lerobot/… (Yes, real people are training on this exact arm for a few hundred bucks.) 3/ Install LeRobot Train ACT Policies LeRobot is Hugging Face’s open-source robotics library (PyTorch-native). It ships state-of-the-art policies like ACT (Action Chunking Transformers); perfect for precise manipulation tasks. → Main repo (star it): github.com/huggingface/lerob… → ACT policy guide training examples: huggingface.co/docs/lerobot/… One-command training examples are in the repo. You’ll be running policies the same day. 4/ Add Multimodal LLMs for Spatial Intelligence Hook up a cheap webcam multimodal LLM (like LLaVA, Qwen-VL, or even Claude-3.5) to: • Understand your workspace in real time • Select the right ACT policy based on what it sees • Give natural-language commands Exactly what @KuphDev is doing in his livestreams (color-sorting ducks with SO-101 LeRobot multimodal vision). This is the “spatially aware” leap everyone’s excited about. 5/ Sim2Real Transfer (The Magic Step) Train in MuJoCo → fine-tune on the real SO-101. LeRobot makes domain randomization and sim-to-real dead simple. Start with basic pick-and-place, then level up to multi-step tasks. 6/ Next-Level Moves (Once You’re Hooked) • Add a second arm for bimanual tasks • Train with human demos (teleoperate the leader arm) • Scale to mobile bases or even low-cost humanoids later Total starter cost: simulation = free SO-101 ≈ $200-300 webcam laptop you already own. The barrier is gone. Hardware access was the only thing stopping most people... now it’s solved. Who’s actually starting their robot journey this week? Drop 🔥 if you’re in (or building right now). Photo: Google Gemini ——- Weekly robotics and AI insights. Subscribe free: 22astronauts.com

¿Cómo se llama en tu país un grupo de ratas? 🐀

Buen ejemplo para el mundo

How Drake froze the CN Tower using 75 projectors worth over $15 million (Via: CBC News)

If the AP & Swatch at the end is not a limited edition make not sense to me rush to the store to get one. My peace not need to be altered

This is the way

I was really hoping to buy a Swatch & AP next weekend, assuming I'd be able to wear it on my wrist. But now that I see it's going to be a pocket watch, I'm having second thoughts...

In a modern era with a lots of technology why all the UFO videos have a poor/bad/low quality...make not sense at all

Paint Bowser with only 3 colors!🥦

Paciencia y genética: la fórmula de Quico Fernández para producir un jamón de bellota en Salta que compite con los grandes de España. 🌳🍖 70 años de árboles y 40 años de selección de raza para llegar a un nivel de altísima calidad internacional. Si están por Salta, pasen por Cerdo Negro y comprueben cómo un productor argentino se puso a la par de los mejores del mundo. Podemos decir que es un Jamón Argentino Tipo Ibérico.

Effective back exercises to reduce back fat.. Build a defined back at home.

Sometimes people just want attention...

Blender Beginner VS👊 Professional