🤖 Warehouses Learn A New Routine
Good Morning, Roboticists!
The world is not a collection of images. It is a sequence, and robots are finally being trained accordingly.
Warehouses Learn A New Routine
TL;DR: Tutor Intelligence introduced second-generation Cassie and Sonny warehouse robots already working in U.S. production facilities. Cassie now drives between workstations, handles bulk material, and is headed for end-to-end case picking, while Sonny brings the company's 4.5-billion-parameter Ti0 model to e-commerce picking, navigating brownfield warehouses and delivering completed orders to packout without custom programming or fixed installations. Read more →
Construction Climbs The Wall
TL;DR: Raise Robotics' Autonomous Mobile Fabricator has moved beyond ground-level cells: after six months with Genie, its swappable workcell operated from a boom lift at roughly 65 feet, or six stories. The same system is designed to move among an AGV, scissor lift, and boom lift, initially for autonomous coating and surface maintenance on large structures. Read more →
Australia Gets A Robot Cleaning Crew
TL;DR: Pudu Robotics and Australian facilities-equipment company Conquest formed a partnership to expand autonomous cleaning across warehouses, industrial sites, and commercial facilities. Conquest will handle local site audits, demonstrations, deployment guidance, flexible purchasing, and national service support, while customers can assess Pudu robots during roadshows planned across Australia in October and November this year for prospective operators. Read more →
The Scaffold Gets Company
TL;DR: RobotPlusPlus closed a Series C worth hundreds of millions of RMB to expand its working-at-height robots and embodied-AI platform. The Beijing company says its ship-hull derusting machines already operate in more than 100 major shipyards, have serviced over 10,000 cargo vessels, and can cut repair cycles from nearly a week to roughly three days at scale. Read more →
Labs Stop Turning Knobs
TL;DR: MIT engineers built a reconfigurable robotic optics lab that can assemble, tune, dismantle, and rebuild laser experiments from standard components. In a demonstration, a seven-jointed arm completed 50 maneuvers to build a functional laser cavity within 30 minutes, then automatically realigned the system when researchers disturbed its components. The system uses QR-coded housings, cameras, and motorized adjustment tools. Read more →
The Robot Stops At Delivery
TL;DR: At a September 15 industry seminar, Unitree industrial solutions lead Xie Yipeng said the company will develop robots and core components rather than become an engineering contractor responsible for project delivery. The stance leaves third-party partners to adapt hardware, tools, sensors, and power systems into reliable factory applications for customers who need everyday throughput and support. Read more →
Time Finally Enters The Loop
TL;DR: TEMPO adds a video-derived motion summary and a short history of a robot's own commands to a pretrained vision-language-action model, giving it temporal context a single image cannot supply. Across four dynamic tasks, the researchers raised bottle-handover success from 44% to 74% and released TEMPO-Bench, containing more than 50,000 annotated frames for evaluating motion-aware robot perception. Read more →
The Next Move Comes First
TL;DR: ModAR makes robot policies predict point tracks, visual features, and depth before choosing an action, instead of forecasting only future RGB images. On three real-world bimanual tasks, its sequential world-action model beat baselines; human video also lifted the reported average success from 70% to 81.1%, tying useful robot learning to ordinary demonstrations rather than specialized robot-only data. Read more →
Human Demos Cross The Gap
TL;DR: UMI-Bridge uses a handheld Universal Manipulation Interface as the bridge between human demonstrations and robot data, aligning examples by action rather than visual similarity. Across three real-robot tasks, the researchers report 91.7% mean success versus 73.3% for naive co-training, and beat a full-data robot-only baseline using only one quarter of the robot demonstrations in tests. Read more →
One Command Moves Everything
TL;DR: A new reinforcement-learning controller maps a six-degree-of-freedom end-effector target straight into coordinated arm and leg actions for a bipedal mobile manipulator. Real-robot tests showed it reaching, shifting posture, and stepping under VR teleoperation, diffusion-policy, and scripted commands, without separately issuing base-velocity or footstep instructions. The same interface connected diverse high-level inputs to one whole-body policy. Read more →
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at mark@aisecret.us if the button fails.