Foundation
What changed, and what is the evidence.
ComputeControlHandsPhysical AIRuntimeVision
🗞️ Pulse first · 3 news posts
Coverage
- 🗓️ 2026-10-22 · ✍️ RoboterminalEka's Robot Picks 1,500 Things an Hour. Its Gripper Is Rated for About 139 Days of That.
- 🗓️ 2026-10-14 · ✍️ RoboterminalThe Pi API Beta Is a Four-Day Loop. The Demo Is Not the Rate.
- 🗓️ 2026-10-13 · ✍️ RoboterminalDeepSeek's First Robot-Brain Bet Is Priced at $400 Million. Skild AI Is Worth $14 Billion.
Related topics
Timeline
What changed, and what is the evidence.
Oct 22, 2026 · Funding · posted
| Headline |
|---|
4NE1 Gen 4 is already making its mark on the world stage. 🤖 Today, we unveiled the next generation of our humanoid robot on two major stages. First, at the opening of the TUM RoboGym powered by NEURA in Munich, built in collaboration with… |
Two stories broke the same week that tell you where AI security actually stands. Wiz, the Google-owned security shop, Calif, and OpenAI's own red team independently found two critical PHP flaws in August using GPT-5.6 Sol and Google's… |
BOOM 🆘 2027 is shaping up to be the brilliant year that Tesla's AI ambitions become real 🔥 On the Q3 earnings call, Elon and the team put dates on almost everything enthusiasts have been waiting for, and the biggest one is about the car… |
AI grabbed all the headlines, but the rest of $TSLA's Q3 deck tells a second story that is just as exciting: the core businesses funding all that AI is only getting stronger 🔥 💳 Tesla's services business just posted record profits.… |
Generative Bionics and AMD are building a Physical AI data factory Generative Bionics is partnering with AMD to build a 5,000㎡ Physical AI Data Factory in Italy. The first facility is scheduled to begin operations in Q4 2027. The goal is… |
Skild AI is building the foundation model for the world's robots. $1.4B raised since 2025. $100M+ in annual recurring revenue. Major industrial partnerships signed. Ahead of the DAO’s upcoming capital allocation proposal, we take a closer… |
Fal and Fireworks AI just put out a number worth sitting with: 961.6 tokens per second, one request, on a 27B model, four Nvidia B300s doing the work. That's the early-launch benchmark for EAGLE-4, the fourth generation of a speculative… |
Are VLAs/robot foundation models or agentic coding models for robotics the future? (Both) https://t.co/rXrSGHjlJ6 |
FieldAI CEO and Co-Founder Ali Agha spoke at @TechCrunch Disrupt 2026 on “Operating at the Edge: How AI Works When the Cloud Doesn’t,” sharing practical lessons from deploying autonomous robots in unstructured environments. Field… |
🚨 BREAKING: @origin_autonomy just emerged from stealth with $20M to automate the work that happens inside buildings, and it’s already running at three times human speed. Co-led by @Accel and @ZacuaVentures, with angels from Anduril, Skild… |
AMD just paid $8.2 billion to hire the godmother of AI 💸 @AMD is acquiring @theworldlabs, Fei-Fei Li’s spatial intelligence startup, in an all-stock deal. Li becomes AMD’s Chief Scientist, reporting directly to CEO Lisa Su. Two of the… |
What if robots could learn from millions of videos instead of relying on more limited teleoperation data? Today, much of robot learning still depends on humans remotely controlling robots to collect training data. It’s expensive,… |
This is the real deal! Trained in simulation. Deployed in reality. No post-training, just... pre-training. A humanoid robot executing soccer skills like dribbling at 2.3 m/s, passing, turning, and scoring on outdoor fields amid people,… |
We are proud to announce Neuron AI: our multimodal foundation model for Physical AI Neuron AI connects 3D vision, touch, audio, robot state and language with memory and turns that information into action. That’s the idea behind PLA, short… |
Jensen Huang flew to London this week and didn't fund a chip order. He funded five very specific bets: molecule prediction, infrastructure timelines, algorithm discovery, disease biology, and quantum error correction. The deployment sits… |
Physical Intelligence just opened its API beta. You upload data, fine-tune their latest robot foundation models on PI-managed compute, and run inference through a hosted gateway. CloudChef demonstrates workflow results:… |
Robot hands have always been the hard part. Picking something up is easy. Putting glasses into a case the right way around, while the other hand holds the case open, is not. @naturewill just released Natura-1, a foundation model built for… |
Users don't wait for the end of an answer. Build voice that handles it: local speech, interruptible by design, model placed where the job needs it. Build on a tested foundation → https://t.co/tHsiLSoyk9 #VoiceAI #EdgeAI |
This robot is making wontons. One hand cups the wrapper while the other adds the filling, then both hands fold it closed. It’s running Natura-1, @naturewill’s 50B-parameter mixture-of-experts model. The team says it was pretrained on… |
Physical AI feels like it's moving into a new stage. Physical Intelligence just opened the Pi API beta. Robot companies can upload their own data, fine-tune Pi's general-purpose foundation models, and run cloud inference to control their… |
Cool foundation models using wuji hand 2 https://t.co/IKie5FWBYH |
Super-impressed: Nature Will releases 50-billion parameter general foundation model for robot intelligence The big question: Is the robotics industry wasting billions on training as the frontier models & this foundation model get better… |
Can robots use their hands to do tasks they’ve never been trained for? @naturewill Natura-1 aims to combine three capabilities in one physical-world foundation model: * Zero-shot generalization * Dexterous control * Long-horizon execution… |
What if robots could learn to handle new situations without being trained for every single task? Nature Will ( @naturewill ) is introducing Natura-1, a new general foundation model for the physical world, bringing together dexterous… |
. @naturewill released Natura-1, a general foundation model for the physical world, trained on nearly one million hours of egocentric human video. Three capabilities in one model: dexterous control, zero-shot instruction following, and… |
Robots are getting closer to using their hands like we do. @naturewill is introducing Natura-1, a general foundation model for the physical world built around dexterous control, zero-shot generalization, and long-horizon execution. Here’s… |
Nature Will @naturewill just released its general foundation model for the physical world, Natura-1, trained on 1 million hours of egocentric human data. Dexterous manipulation is especially challenging. Objects move and rotate, contact… |
One of the biggest challenges in robotics is combining fine motor control with the ability to generalize to unfamiliar tasks. Nature Will's @naturewill new foundation model Natura-1 aims to tackle both, combining dexterous manipulation,… |
Trust matters when physical world data is collected by a distributed network. Buyers need usable 3D assets, while contributors deserve clarity about how their original footage is handled. @vangrid_io addresses this through a bounty… |
We're open-sourcing Menlo SDK so you can build your own applications for a humanoid robot. The SDK connects your code to Asimov 1, letting you control its movements and read its state. With the full source available, you can build on the… |
Before long, a humanoid will do your Saturday chores In the demo, a clothed Unitree G1 works through the chores on its own...Corn out of the fridge into the air fryer. Jeans folded. Duvet spread across the bed--It feel like a household… |
Excited to see our partners at @Ultraroboticsco announce their $50M Series A! Ultra is a great example of what's becoming possible when general-purpose robot foundation models meet real-world deployments! https://t.co/V3eHsCa2au |
Sabi designed its own non-contact EEG chip, had TSMC fab it, and put it in a cap that turns your thoughts into text! Custom ASICs, custom sensors, and a brain foundation model trained on their own neural data. $50M from Khosla, Accel, DST,… |
GeneralistのGEN-1.5、数秒の実演(お手本)から追加学習なしで新しい仕事に適応。 現場ごとの作業立ち上げ負担を減らす可能性。 https://t.co/jmKZHtoWEX #PhysicalAI #RobotLearning #foundation #model #ZeroShot #Generalist https://t.co/diw0EvQQU3 |
RP1 Takes on the IROS Challenge At IROS 2026 RoboParty’s RP1 took on the Kick Me interactive challenge. RoboParty is a full stack open source humanoid robotics company building an open technical foundation for developers worldwide from… |
🚨 BREAKING: @BostonDynamics appointed a new CEO! 😎 The new CEO is Rohit Prasad, former SVP and Head Scientist for Alexa and AGI at Amazon. An AI veteran, someone who spent 12 years turning AI into products at global scale, building Alexa… |
Capital only buys the cleanroom. Specialized production equipment only gets a facility to the starting line. Commercial output depends on daily process discipline on the floor. Critical toolmakers remain the true operational foundation of… |
GOOD NEWS 🎨 Tesla has officially been granted a design patent (US D1,151,323 S) covering the exterior body of its Cybercab robotaxi 🔥 The timing is telling. Tesla filed the application on October 9, 2024, just one day before unveiling… |
$2,599 buys a Windows laptop that Microsoft says renders AI images 4.3 times faster than a 16 inch MacBook Pro on M5 Pro. The marketing is aimed squarely at Apple. The architecture underneath is aimed at something bigger: deciding, task by… |
“Learning the model is really only a tiny part of an overall robotic system. It’s an important part, but not the only part, and not the part that really makes the system work.” Dr. Sebastian Scherer, also known as “Basti,” is Director of… |
"Hey Alexa... uh, I mean Atlas, hand me the drill." Boston Dynamics has a new CEO: Rohit Prasad, the man who built Alexa. And it's a notable break from BD's past. BD's first two CEOs were hardcore roboticists who helped pioneer the field.… |
Yesterday OpenAI launched its Decisions API. You give GPT-6 Luna text or an image plus a short list of possible answers, and it picks one in a fraction of a second. In the launch demo, that is enough to aim a Hugging Face Microduck's head… |
You can start training a humanoid before you own one. Asimov 1's open-source training code gives you a foundation to train and evaluate locomotion policies in simulation. Train your own policy: https://t.co/Wv92FhjZvn… |
Crazy how far robot foundation models have come in terms of true fine motor skills: Without task-specific training, no sim-to-real transfer, and... fully autonomous ❗ The dual-arm robot places an O-ring onto a hydraulic rod and removes it… |
Kawasaki Targets a Fully Autonomous Humanoid by 2030 According to Nikkei Asia, Kawasaki Heavy Industries is targeting a fully autonomous humanoid robot by 2030. Its latest Kaleido 9 is the ninth generation of a humanoid program Kawasaki… |
Physical AI has a data problem that bigger models cannot solve alone @vangrid_io is interesting because the answer is not simply collecting more of the same data A robot or world model may need information from a specific street, building,… |
Foundation models make real to sim look easier than it is imo https://t.co/jyANfO7YgG |
Let's start with the robot news series! 😎 First of the week is out! @runwayml the company behind AI-generated video, just launched Praxis-1, a foundation model for robot control trained almost entirely on video. Robot training data is the… |
it’s honestly insane how many pieces of the robotics stack are finally becoming good enough at the same time we now have - cheap high torque actuators - compact motors and gearboxes - better batteries - force and tactile sensing - depth… |
Delta Intelligence has introduced Delta-0, a humanoid foundation model designed to help robots coordinate walking, balance, and manipulation. Trained on more than 10,000 hours of human data, Delta-0 can perform everyday tasks such as… |
What will it take for AI to move beyond generating pixels and start understanding worlds? At #NVIDIAGTC Berlin, hear from @bfl_ai, @ltx_io, and @vesoma_ai for a conversation about the future of world models. Explore whether traditional… |
How are thousands of components placed on a PCB? ⚙️ A PCBA can contain hundreds or even thousands of tiny components. How do they all get placed in the right position? SMT pick-and-place machines: ✅ Feed components into the system ✅… |
I can't remember the last time this many people from physical AI were booked into one room! The Long Horizon Physical AI Summit by @encord_team lineup is fire! 🔥 → Sebastian Thrun, founder of Google X, who led the self-driving project… |
Foundation models trained only on internet text and 2D images hit an immediate wall in industrial environments. Real-world physical AI requires spatial intelligence, force vectors, and multi-modal edge telemetry gathered directly where… |
Fast fashion fulfillment leaves little room for delay or error. ITOCHU Logistics serves one of the world’s Top 3 fast fashion brands, managing complex SKUs, seasonal peaks, and evolving omnichannel requirements. To build a more responsive… |
Building a mobile robot starts before the robot starts moving. We’re kicking off our Mobile Robot Full-stack Development In Practice series with the first step: setting up the robot’s computing and development foundation. Part 1 covers: •… |
[ ANALYSIS ] 🇪🇺 Countries Are Easy, Population Is Brutal: Can Tesla Actually Win Europe’s FSD Vote? 🔥 Croatia’s green light officially pushes Tesla FSD past the halfway mark toward the 15-country threshold needed for European Union… |
[ ANALYSIS ] 🇪🇺 Croatia’s green light officially pushes Tesla FSD past the halfway mark toward the 15-country threshold needed for European Union approval 🔥 The momentum looks electric on paper, but it also brings a hard institutional… |
if you are building: - actuators - dexterous hands - training data for physical ai - self-driving labs - magnets - the deployment layer of robotics - reindustrialization - space manufacturing - energy and battery innovation -… |
GPT-6 Astra is already showing up in more and more robot demos. KPI offers a glimpse of what that could mean. Astra handles the task, scene understanding and motion planning. KPI then turns those high-level intentions into real-time… |
A spatial intelligence network needs two things to scale: a reliable technical foundation and the ability to bring real-world data demand into the network. That is what makes the team behind @vangrid_io worth looking at. CEO Robert… |
Impressive level of coordination and whole body control. Picking up a record is nice because its a fairly fragile object, and placing it is quite precise. Very optimistic about robot foundation models https://t.co/nX1MppDKUE |
Humanoid robots are starting to generate their own experience? Delta-0 uses a real-to-sim-to-real loop: robots can autonomously execute tasks, retry after failures, and use the resulting feedback to continuously improve through real-world… |
Day one at IROS, and booth 1122 is open! Teams training VLAs, world models, and robotics foundation models are stopping by to see our demo and talk through what they're building. #IROS2026 @ieeeiros https://t.co/6KmDgOpq5P |
This demo is truly impressive A humanoid robot plays music for itself and then begins performing household chores autonomously. Of course, this is a cleverly engineered setup; honestly, if humanoid robots were ever programmed with a… |
Moving from foundation model benchmarks to factory floor cycle times reveals the real bottleneck. Precision harmonic reducers, planetary rollers, and rare-earth magnets cannot scale as quickly as compute clusters. https://t.co/yGGbBAiI4f |
TL;DR ⚡ If you don’t have time for the full deep dive, let’s skip straight to the core idea: Tesla turns ordinary surround-camera footage into a persistent overhead model of the road, so curbs, lanes, crosswalks, and junctions do not… |
The standard framing of Momenta as an "ADAS software company" understates what the firm is actually building. Momenta positions itself as a foundation model developer for physical AI. Its April 2026 R7 world model — deployed in mass… |
SLAM dunk your robotics knowledge! ⛹🏼♂️ Robots are getting smarter and more capable, hence SLAM is often considered one of the "holy grail" of robotics, enabling machines to map unknown environments while tracking their own position… |
𝐀𝐈 𝐡𝐚𝐬 𝐥𝐞𝐚𝐫𝐧𝐞𝐝 𝐭𝐨 𝐭𝐚𝐥𝐤. 𝐍𝐨𝐰 𝐢𝐭 𝐧𝐞𝐞𝐝𝐬 𝐚𝐧 𝐢𝐧𝐭𝐞𝐫𝐟𝐚𝐜𝐞 𝐭𝐨 𝐭𝐡𝐞 𝐩𝐡𝐲𝐬𝐢𝐜𝐚𝐥 𝐰𝐨𝐫𝐥𝐝. Humanoids will become that interface. Not because the human form is automatically the best machine for every… |
A Century of Innovation (1926–2026) 2026 — 🇺🇸 Agentic AI: Systems that don't just think, but act autonomously. 2026 — 🇨🇳 Humanoid robots: Humanoid robots move from research labs into workplaces and real-world deployments. 2025 — 🇺🇸… |
What makes Asimov 1 walk is now open-source, giving you a foundation to build new capabilities on a humanoid robot. https://t.co/E6AX8tUEz0 https://t.co/OEL4NhsKDI https://t.co/cQvUOPF6To |
Industrial physical AI does not stall on foundation models. It stalls at the contact point. When transient dynamics hit real actuators, cycle times drift. Delivering sub-millimeter positioning and force feedback across full shifts is where… |
Could the automotive supply chain become the foundation for America's drone industry? 🔗 https://t.co/O61F41Ok4P #DroneManufacturing #MadeInAmerica #MichiganManufacturing #AutomotiveManufacturing #Reshoring #DroneTech #SupplyChain… |
What if sim and reality were one? This system keeps them in sync… always! 👉 Project & Paper ⬇️ > uses 3D Gaussians and particles to predict and correct object states > creates a real-time digital twin that stays synced with the real world… |
Your next robotics project could start with these steps. Today, we're open-sourcing Asimov 1's locomotion policy and the training code behind it, giving developers a foundation to build on. Alongside the trained policy checkpoint, we're… |
We have plenty of human demonstration data for movement. We don’t have enough for touch. Without tactile feedback, foundation models can’t feel if an object is slipping, soft, or fragile during complex manipulation. Matrix Inno is bringing… |
AlphaGo, but for humanoid soccer. Skild AI brings self-play to humanoids. Their flagship robotics foundation model, S1, can now be post-trained with no human demonstrations by competing against itself in simulation. The testbed is soccer,… |
Why an in-person summit still matters when physical AI is moving this fast?! 👀 Look at the last few weeks alone. @SkildAI crossed a $100 million revenue run rate and shipped its S1 model. @wayve_ai started supervised autonomous rides in… |
Hai Robotics is strengthening its Americas leadership team to support the next stage of regional growth and customer success. We’re pleased to welcome six leaders across critical areas of our business: • Ottavio Saluzzi, VP of Software •… |
TacnIQ ai just secured $1.5M from In Group Holdings, rounding out a $3M pre-seed round. What the Singapore based team is building is really interesting tactile foundation models trained on real physical interaction data across logistics,… |
4D spatial reconstruction is the foundation for next-gen physical AI. True world models need to operate across both space and time, unlocking more robust autonomous systems in automotive. https://t.co/faVS27GRiO |
The scalability of physical AI is fundamentally tied to safety validation. Lessons from autonomous vehicle development are directly applicable to robotics, emphasizing the need for rigorous design, testing, and deployment protocols to… |
Excited to see @NVIDIARobotics #Isaac #ROS 5.0 bring #agentic workflows to the ROS community. 🤖 We’re glad to contribute with #reBot Arm, connecting open-source ROS 2 development with NVIDIA Isaac ROS and #Jetson platforms. With… |
ForgeAI raised an angel round in the RMB 10M range, with TusStar Ventures among the investors. The Beijing team is building a robot foundation-model stack designed to move learned skills across different robot bodies. It is already testing… |
Brett Adcock’s framework for humanoid robotics is one of the clearest ways to look at where the industry is actually going. He breaks it into four stages: Great humanoid hardware Whole-body, AI-first architecture Scaling intelligence… |
UBTECH began deliveries of its globally popular ultra-bionic humanoid robot, UWORLD U1, on September 16. The first deliveries include Shanghai Yongda Group(a luxury car retailer), Beijing Boshi Group(a major Beijing Mercedes-Benz dealer),… |
UnifoLM-X2-1.0 addresses limitations in world-action foundation models around rapid planning, decision-making, and real-time interaction. The system is designed to predict future states, adapt to changing environments, and execute actions… |
The cross-embodiment foundation model is a critical unlock. If intelligence can truly be abstracted from specific hardware, it changes the economics of robotics development. This is about skill transfer and generalization, which is far… |
A robotics company may be shutting down but its data could live on. 🤯🤖 EIDON AI has released its robotics dataset publicly on Hugging Face after winding down operations. ( https://t.co/e5OqbX54y4 ) The dataset contains: 🎥 1,274 hours of… |
The shift I keep waiting for is local inference leaving the hobbyist corner, and this week I think I watched it happen. What used to mean a terminal window and a quantized model file on your laptop just got Apple's blessing: MLX-Serve, an… |
Apple shipped something small this cycle that I think matters more than another point release of a frontier model: macOS 27 now includes an fm command in the terminal, putting Apple's on-device model one shell command away from every… |
What Exactly Is Physical AI ? The term "Physical AI" is being used quite broadly by management, so before we evaluate Sona’s opportunity, it is important to understand what it actually means. Vivek Vikram Singh has positioned Physical AI… |
This is the actual moat in physical robotics. Software and foundation models scale instantly across compute clusters. Qualifying harmonic reducers, planetary rollers, and sub-micron grinding tolerances across Tier 2 supplier networks takes… |
Enterprise diffusion moves at the speed of systems integration, not API availability. On industrial plant floors, deploying existing foundation models requires connecting legacy PLCs, fieldbus protocols, and deterministic cycle… |
a real-world test of robot control! And tellingly its being done with fable and astra, not (insert robotics foundation model here) time for some more soul searching on the part of my field i think https://t.co/eoojAxP3m4 |
The harder problem for Physical AI is not simply collecting more data. It is knowing where that data came from, how it was captured, and whether it can be trusted. That is why the infrastructure behind @vangrid_io caught my attention.… |
Zero-shot fully autonomous drone flight from our scaled embodied foundation models (internal world model, embodied reasoning, direct control all in one) https://t.co/RXUqnt0Yvw |
One robot model. Different robot bodies. 🤖 Reward AI has introduced OM-1, a robot foundation model designed to transfer manipulation skills across tabletop arms, industrial arms, mobile robots and humanoids. The interesting part is how it… |
Announcing our commercial partnership with @FeagineRobotics Feagine builds soft embodied intelligence infrastructure—from bionic flexible robot arms to embodiment-free data and the Fi0 cross-embodiment foundation model. Through our… |