Data is robotics’ bottleneck — NVIDIA just open-sourced its data factory.
AI NEWS
Past AI News
Chinese and English are the same day. This page lists every item, each with the original link, the author, and the engagement.
This day
October 23, 2026
9 items for 2026-10-23, ranked by combined engagement.
Top 3
Eka Robotics raises a $150M Series A for its Vision-Force-Action model
Robots need touch, not just sight — and investors are paying for it.
The coding-agent productivity playbook comes to the factory floor.
All
NVIDIA open-sources Flora, a robotics data engine labeling 200K hours of video a day at 125x lower cost
Data is robotics’ bottleneck — NVIDIA just open-sourced its data factory.
CollapseNVIDIA open-sources Flora, a robotics data engine labeling 200K hours of video a day at 125x lower cost
NVIDIA’s Jim Fan released and open-sourced Flora, a suite of models that parse physical-world footage into training signals for robots. He says Flora already powers the team’s in-house data engine, labeling 200K hours of video a day on about 500 GPUs, with Astra-level performance at 125x lower cost. It provides timestamped captions to curate human egocentric and robot data, dense progress rewards for quality control and RL post-training, and rubric scoring for real-world benchmarking. Model cards limit use to research and development.
Eka Robotics raises a $150M Series A for its Vision-Force-Action model
Robots need touch, not just sight — and investors are paying for it.
CollapseEka Robotics raises a $150M Series A for its Vision-Force-Action model
MIT professor and Eka Robotics co-founder Pulkit Agrawal announced a $150M Series A. Eka is built around its Vision-Force-Action (VFA) model and three convictions: force is the language of manipulation, reinforcement learning is how dexterous manipulation scales, and fast, reliable physical execution is the bottleneck. The money will scale the model and speed customer deployments. Reports say a16z and F-Prime led at a $1.45B post-money valuation.
Ex-OpenAI research chief Bob McGrew’s Arda, valued at $1.5B, aims to be Cursor for manufacturing engineers
The coding-agent productivity playbook comes to the factory floor.
CollapseEx-OpenAI research chief Bob McGrew’s Arda, valued at $1.5B, aims to be Cursor for manufacturing engineers
Arda, the manufacturing-AI company of former OpenAI chief research officer Bob McGrew, emerged from stealth with a $160M Series A at a $1.5B valuation. Its software reads product CAD files and generates assembly sequences, work instructions and workstation layouts, updating them as designs change, and is pitched as Claude/Cursor for manufacturing engineers. Anduril is already a user; reportedly, basic work instructions dropped from about 10 hours to 15 minutes.
Skyfall AI unveils SKY-001, pitched as the first engineering world model for CAD and BIM
World models move from games and robots into engineering design.
CollapseSkyfall AI unveils SKY-001, pitched as the first engineering world model for CAD and BIM
Skyfall AI introduced SKY-001, which it calls the first world model for engineering, arguing that design needs long-horizon planning and latent reasoning that LLMs weren’t built for. In the company’s own comparison, SKY-001 was 100x faster and 150x cheaper than top frontier LLMs on CAD and BIM tasks, with higher precision. It explores many designs in parallel, breaks each into its smallest components and reassembles them in 3D for tools like Autodesk Fusion. It is a waitlisted research preview.
Agile Robots unveils Agile ONE SN reference humanoid on NVIDIA GR00T as Isaac expands
A European robot maker and NVIDIA ship a standard chassis for humanoid R&D.
CollapseAgile Robots unveils Agile ONE SN reference humanoid on NVIDIA GR00T as Isaac expands
At NVIDIA GTC Berlin, Agile Robots unveiled Agile ONE SN, a reference humanoid for robotics research. Built on NVIDIA Isaac GR00T and powered by Jetson Thor, it gives researchers and developers an integrated platform to develop, train and validate humanoid skills. NVIDIA also expanded Isaac with new tools for agents, datasets, a runtime and robot-learning framework updates to speed the path from data to deployment.
NVIDIA releases Isaac Lab 3.0 with a modular architecture, multi-physics and real-hardware deployment
A major upgrade to a widely used open framework for robot training in simulation.
CollapseNVIDIA releases Isaac Lab 3.0 with a modular architecture, multi-physics and real-hardware deployment
NVIDIA officially released Isaac Lab 3.0, its robot-learning framework, with a new modular, lightweight architecture. It adds multi-physics and multi-rendering support, training at scale, and streamlined deployment on real hardware. Highlights include Newton integration, the new OVRTX renderer, RL post-training and LEAPP deployment.
Agility Robotics signs a commercial deal with AGCO to bring Digit into a farm-equipment plant
Another real factory contract for humanoids, now in farm machinery.
CollapseAgility Robotics signs a commercial deal with AGCO to bring Digit into a farm-equipment plant
Agility Robotics signed a commercial agreement with AGCO for a Digit humanoid pilot at AGCO’s Hesston, Kansas facility, expected to begin in 2027. Initial tasks are parts sorting and stacking and placing totes for storage. Agility says its robots now span automotive, logistics and agricultural machinery. Contract value and robot count were not disclosed.
Siemens and NVIDIA push AI-driven autonomous manufacturing at Siemens’ Erlangen electronics plant
An industrial giant uses its own plant to show real AI-factory gains.
CollapseSiemens and NVIDIA push AI-driven autonomous manufacturing at Siemens’ Erlangen electronics plant
Siemens is teaming up with NVIDIA at its Electronics Factory Erlangen to advance AI-driven, autonomous manufacturing. The plant already runs more than 100 AI algorithms in production, across logistics, quality inspection and maintenance. Siemens cites up to 40% faster time-to-market and up to 20% savings in production planning using AI agents. NVIDIA Cosmos generates PCB inspection training data, and Siemens digital twins validate inspection and production plans before deployment.
China’s MIIT is drafting new-generation smart-manufacturing policies to deepen AI in industry
The next policy signal for AI in Chinese manufacturing.
CollapseChina’s MIIT is drafting new-generation smart-manufacturing policies to deepen AI in industry
At a State Council Information Office briefing on Oct 20 on Q1–Q3 industrial performance, MIIT chief engineer Wang Weiming said September’s national advanced-manufacturing conference set new-generation smart manufacturing as the main direction for making industry smarter, greener and more integrated. MIIT is drafting new-generation smart-manufacturing policy measures to deepen the integration of AI and manufacturing, per Securities Times. The full policy and timing have not been released.