Robotics

Today's Stories

1. 1

   LessWrong: AI 2027 scenario underestimates miniaturization incentives

   AI researcher Daniel Kokotajlo and his team published an AI 2027 scenario that models a robot economy where human-scale robots build robot factories. A commenter on LessWrong argues the scenario is not science-fictional enough and prioritizes respectability over accuracy. The commenter contends the scenario misses extreme incentives for miniaturization to physical limits and overlooks the strategic power of self-replicating systems that depend only on environmental inputs rather than human-supplied resources. This suggests mainstream AI forecasts may systematically underestimate certain technological trajectories.

   The critique highlights a gap between plausible near-term AI scenarios (like manufacturing integration) and more speculative but potentially high-impact outcomes (like miniaturized self-replicators). How the AI forecasting community responds to these concerns may shape future scenario-building.

2. 2

   Amazon builds robotic Texas warehouse, opens Trainium chips to AWS customers

   Amazon is constructing a large, highly automated robotic fulfillment center in Texas and expanding access to its in-house Trainium AI chips to external AWS customers, signaling a push to deepen control over both physical logistics and AI infrastructure. The Texas warehouse automation and broader Trainium rollout reflect Amazon's strategy to lower unit handling costs, support faster delivery, and offer customers an alternative to Nvidia hardware while keeping more of the AI stack in-house—though this requires substantial upfront capital spending and dependence on bond market appetite.

   The execution of the 248,687 square foot robotic facility and customer adoption of Trainium chips will determine whether Amazon can achieve the efficiency gains and higher-margin services it is betting on, particularly as fuel and freight rates remain elevated.

3. 3

   Chinese robot startup X Square bets on unified stack for general-purpose machines

   X Square Robot, a Chinese embodied-AI company, has unveiled an integrated foundation stack for robotics spanning data collection, a world model (WALL-WM), and an action model (Wall-OSS-0.5). The company is releasing the code openly and argues this layered approach—organized around physical events rather than fixed time slices—solves robotics' core problem: how to build capability that transfers across tasks and machines. The company's valuation has climbed above 20 billion yuan (about US $2.9 billion（約4600億円）). Robotics has long assembled separate perception, planning, and control modules that don't generalize, unlike large language models which scale predictably. X Square's bet is that data quality and infrastructure, not model size, are the real bottleneck for general-purpose robots. The company reports reaching performance comparable to an all-robot dataset at roughly a 20-fold lower cost by pretraining on robot-free human demonstrations captured via wearable rigs, then anchoring to real-robot data. If validated independently, this cost reduction could reshape robot training economics.

   The company's strongest results are currently measured on its own robots and benchmarks. With code now released, the broader robotics community will test and reproduce these capabilities across different robots, tasks, and settings—a crucial step to confirm whether the stack's principles hold beyond X Square's controlled environment.

4. 4

   Digital and Physical AI follow identical growth patterns, analysis shows

   A consulting firm presented a global labor automation map for Physical AI (humanoid robots) at a Tokyo summit in May 2026. When compared with the firm's existing research on Digital AI job functions, the two maps revealed the same underlying growth logic. The parallel patterns suggest that automation—whether through software AI or physical robots—follows predictable pathways across industries and regions. This may help businesses and policymakers anticipate where labor disruption will occur and plan workforce transitions accordingly.

   The analysis was conducted at a humanoid robotics summit in Tokyo in May 2026, indicating that physical robot capabilities have advanced enough to warrant serious global labor-impact mapping alongside digital AI.

5. 5

   São Paulo Launches First Citywide Drone Detection Subscription Service

   R2 Wireless and Ôguen Tecnologias announced a partnership to deploy a subscription-based drone detection network across São Paulo, Brazil, making it the first city to implement a municipal-scale Drone Detection as a Service (DDaaS) network. The service provides continuous monitoring of urban airspace, geolocation, and real-time airspace awareness to public agencies and private organizations. Brazil's officials face growing concerns about modified and improvised drones used by organized criminal groups. The companies say their subscription model offers lower cost and reduced complexity compared to traditional counter-drone deployments, since customers avoid purchasing and maintaining their own hardware while receiving ongoing software updates and threat intelligence.

   The platform is designed to detect and geolocate commercial drones as well as modified aircraft, DIY first-person-view (FPV) drones, spoofed systems, and anonymized platforms—not just standard communication protocols. The companies say the service can scale to support thousands of customers and is being discussed with public security authorities for deployment across the São Paulo metropolitan region.

6. 6

   Mantis Robotics unveils dual-arm robot, plans 2027 rollout

   Mantis Robotics, a US startup backed by Agility Robotics, unveiled its MR-X dual-arm industrial robot and said it will target manufacturing, logistics, warehousing, and smart factories, with a planned rollout in 2027. The move expands Agility Robotics' portfolio beyond its existing robotics offerings. Industrial dual-arm robots that can handle complex manipulation tasks are in demand across factories and warehouses as manufacturers seek automation solutions.

   The company is targeting a 2027 rollout; availability in specific regions and pricing details have not been announced in this report.