Andrew Peters @FlywheelRobotAI

Massive Upgrade for Roboticists Incoming!

Anthropic @AnthropicAI

2 days ago

Today, we're kicking off the first phase of the research preview for Model Hardware Standard (MHS): a new standard for AI agents to safely operate physical equipment in scientific research and advanced manufacturing. Read more: anthropic.com/news/model-har…

467 1K 11K 2.9M 4K

Humanoid robot builders by company age in 2026 • 🇨🇳 AgiBot — 3 years • 🇨🇳 EngineAI. — 3 years • 🇨🇳 Booster Robotics. — 3 years • 🇺🇸 Figure AI — 4 years • 🇩🇪 NEURA Robotics — 7 years • 🇺🇸 Apptronik. — 10 years • 🇨🇳 Unitree Robotics — 10 years • 🇨🇳 Leju Robotics. — 10 years • 🇺🇸 Agility Robotics. — 11 years • 🇨🇳 Fourier Intelligence. — 11 years • 🇺🇸 1X Technologies. — 12 years • 🇯🇵 SoftBank Robotics. — 12 years • 🇨🇳 UBTECH Robotics. — 14 years • 🇪🇸 PAL Robotics. — 22 years • 🇬🇧 Engineered Arts. — 22 years • 🇺🇸 Tesla Optimus — 23 years • 🇺🇸 Boston Dynamics — 34 years • 🇯🇵 Honda — 78 years • 🇯🇵 Toyota — 89 years • 🇯🇵 Kawasaki Heavy Industries — 130 years These figures represent the age of each company not the age of its humanoid robot program. Which younger company could eventually challenge the industry veterans? #HumanoidRobots #Robotics #PhysicalAI

Techniahqrobot | humanoid robots @techniahqrobot

3 weeks ago

GLOBAL HUMANOID ROBOT DEPLOYMENT 🇺🇸 UNITED STATES • About 1,950 humanoid robots • 13% of global installations 🇨🇳 CHINA • Over 12,750 humanoid robots • More than 85% of global installations 🇪🇺 EUROPE • Under 300 humanoid robots • Exact regional total not publicly

0 2 4 4K 2

A Harvard endowment manager hands MIT students a blank sheet of paper every year. No rules. No criteria. No guidance. Just write down where your money goes. Someone writes 100% Apple. Someone writes lottery tickets. Someone writes rare coins. Every year, confident picks. Every year, the same blind spot. Jake Xia collects the pages, plots every answer on one chart, and waits for the question nobody asks. Not one student asks how much goes in each position. They all ask what to buy. Nobody asks how to size it. That question is the entire difference between a portfolio and a guess. The answer won the Nobel Prize in 1990. It's called the efficient frontier, and Xia draws it on the board in under a minute. Five equations sit underneath it. Compound growth. Present value. The geometric mean. The Rule of 72. Real return. All of them older than any bank on earth. All fit on a napkin. None behind a paywall. The industry doesn't hide them. It just hopes you keep asking "what should I buy?" instead of "how much?" The first question feels like investing. The second one is the actual job. MIT filmed the whole thing. It has been free for years. Almost nobody switched from the first question to the second. The napkin is still free. The blank page is still the test.

Lumen @lumenxbt

a month ago

x.com/i/article/2082…

71 385 2K 10.5M 7K

太震撼了！一小时顶两年MBA！ 黄仁勋2011年斯坦福这场演讲，把创业、创新、公司重构的底层逻辑讲得清清楚楚，太值了 强烈推荐反复看

Robotics deep dive: THE DATA FLYWHEEL MYTH The investment case for humanoids is a loop: - deploy robots - collect data - train better models - sell more robots. KraneShares wrote it down in May: "the companies accumulating the most embodied training data through deployed fleets have self-reinforcing advantages that will compound." It is a coherent argument and it already worked once, which is why everybody believes it. Self-driving fleets did get better by driving. Look at what supplied the improvement though: a car in traffic is dragged through an environment nobody controls. Weather, roadworks, a plastic bag, a cyclist doing something stupid. Novelty arrives whether the operator wants it or not. Now the humanoid version of the same loop: one station, one fixture, one SKU, one pick, on a twenty-second cycle, for a year. A hundred robots on that station, twenty hours a day, 365 days: 730,000 robot-hours. Roughly forty-five times the entire public corpus of real, non-duplicate robot manipulation data I counted on Hugging Face last week (about 16,000 hours across 64,448 datasets). Enormous volume. One environment. One task. That census also measured what near-identical data is worth. About a third of the hours in the largest datasets were exact duplicates, the same file re-uploaded under a new name; one simulation benchmark appeared 54 separate times. Nobody counts 54 copies as 54 datasets. The ICLR 2025 scaling paper (arXiv 2410.18647) lands in the same place: 32 environments, one new object each, 50 demos apiece, and the policy generalises to a scene it has never seen. Past a threshold, more demos of the same environment do not enhance the model's ability to understand it. The signal at the end is murky too. A successful autonomous run is a recording of what the policy already does. Failures carry the information, but they are also the thing the customer is paying to eliminate. The one deployed flywheel I can find with numbers attached is honest about its conditions (arXiv 2511.19647): a robot working a university library for two weeks, 2,103 shelves, book identification from 32.0% to 71.8%. Thousands of different shelves, and what improved was perception. Data helps. The conversion from fleet-hours to manipulation capability is the unproven step, and the deployments generating the most hours look like the ones generating the least new information.

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-Weigh…

keyboard_arrow_left Previous keyboard_arrow_right Next

@JensenHuang @nvidia Welcome to X Jansen Huang ✨️✨️

Today, we're announcing that we’ve raised $152 million in our Series A at a $1.35 billion post-money valuation, becoming Europe's first pure-play humanoid robotics unicorn just two years after founding. This funding brings the total raised to date to $270 million. The round was led by @PrimeMoversLab, with participation from @SchaefflerGroup, @BoschGlobal, Fubon Financial Holding Venture Capital, and Aglaé Ventures. The capital will support development and launch of the next-generation humanoid robot platform, the roll-out of Beta version in Q4 2026, start of mass manufacturing of wheel-based robots and the continued development of KinetIQ, our AI brain. Humanoid's valuation reflects rapid execution and the technology's shift from research to commercial deployment, demonstrating that Europe can build and scale globally competitive Physical AI companies. Read more here: thehumanoid.ai/humanoid-raise…

@aryind_ @michellelsun My algo never showed up

Incredible stuff. The precision!

dar @radbackwards

2 months ago

Alien Technology

372 908 12K 2.1M 3K

End of an era: Decommissioning the original Model S & X assembly line in just 46 days

Shout out to the @SpaceX media team! Just such a badass look inside the lead up to flight 12. High five @jstyer and @BobbyPeden1 for the excellent narration and vibes!

SpaceX @SpaceX

2 months ago

The path to launch is filled with obstacles and success is only possible through the tireless efforts of many working together towards a common goal. “Critical Path” continues the ongoing Starship series, following SpaceX engineers through the final days before launch of the

2K 6K 39K 42.9M 8K

You may not like it, but this is what peak performance looks like.