We are watching the capital markets place a hard value on physical AI hardware. With Unitree's massive public debut establishing a multi-billion-dollar valuation anchor, the focus across the robotics ecosystem is immediately shifting to the simulation and data pipelines needed to make these platforms actually work in the real world.
Following the Unitree Robotics STAR Market IPO and 'Superman' prototype debut we've been tracking, final trading data reveals the offering was oversubscribed more than 8,000 times by retail investors—exceeding earlier indications of 5,000x. Backed by Tencent, Alibaba, and DeepSeek, the company also clarified its 2025 financial footprint, reporting 1.7 billion yuan ($252 million) in revenue driven by 5,500 humanoid units and 33,000 quadruped systems.
Why it matters
The massive retail oversubscription and revised revenue figures further cement Unitree's position as the commercial anchor for pure-play humanoid valuations. By demonstrating that high-volume manufacturing of quadrupeds can financially bridge the gap to humanoid commercialization, Unitree is establishing a viable hardware-first business model.
Unitree CEO Wang Xingxing emphasized that public listing proceeds will directly expand manufacturing scale and support physical AI world model R&D.
Startup Generalist AI introduced GEN-1.5 on Thursday, a multimodal robot foundation model capable of acquiring physical tasks from a single demonstration without gradient updates or model fine-tuning. The system evaluates 30 seconds of video context memory to output 100 Hz action trajectories, achieving a 59% average success rate across 10 evaluation tasks in one-shot mode, which increased to 83% following 10 gradient steps on 5 minutes of data. During laboratory demonstrations, the system improvised tool use, including utilizing a banana as an impromptu scraper to clear debris.
Why it matters
Enabling robots to adapt to unscripted physical tasks via in-context learning mirrors the prompt-based flexibility of large language models, eliminating the traditional requirement of training dedicated policies for every new SKU or workspace layout. This capability directly targets high-mix, low-volume manufacturing and domestic assistance where fine-tuning overhead is economically unviable. If in-context physical reasoning scales reliably, it shifts robot deployment from software engineering to visual prompting.
Generalist AI engineers argue that in-context prompting eliminates the need for massive task-specific fine-tuning datasets by relying on spatial world models. Outside observers note that while improvisational tool demonstrations highlight semantic flexibility, commercial deployment demands rigorous safety guarantees that unconstrained in-context execution cannot currently certify.
At the 2026 World Robot Conference in Beijing on Wednesday, Noitom Robotics released HiPHI, a 617.5-hour motion capture dataset available for free on Hugging Face. The open-access corpus features hybrid sensor fusion recorded at 90 Hz with sub-millimeter marker precision and FrameNet semantic indexing. It includes 245.7 hours of whole-body human-object interaction trajectories with complete 3D mesh tracking, validated directly on Unitree G1 humanoid hardware for loco-manipulation tasks.
Why it matters
High-fidelity kinematic data remains one of the rarest commodities in embodied AI training, as standard internet video lacks precise joint torque, contact force, and sub-millimeter 3D trajectory ground truth. By distributing over 600 hours of studio-grade motion capture indexed with semantic FrameNet structures, Noitom lowers the cost barrier for research labs training whole-body humanoid controllers. This release provides an immediate open-source asset for sim-to-real transfer and imitation learning without requiring expensive optical capture stages.
Noitom Robotics framed the release as the foundational dataset layer for its broader 'World Compiler' strategy, aimed at standardizing physical AI inputs. Community researchers welcomed the sub-millimeter marker accuracy and object mesh tracking, though maintainers pointed out that optical mocap data still requires careful retargeting to accommodate structural joint limits across differing humanoid topologies.
The Open Source Robotics Foundation issued a major package sync for the ROS Lyrical distribution on Wednesday, adding 104 new packages and 150 package updates for Ubuntu Resolute across x86 and ARM architectures. Key additions include ros-lyrical-chomp-motion-planner (2.15.0) and updated Nav2 navigation stacks. Concurrently, MoveIt maintainers confirmed during their August community session that MoveIt2 is finalizing its initial Lyrical release milestones, establishing compatibility with new C++20 compiler standards.
Why it matters
Regular ecosystem syncs ensure that industrial and academic developers running ROS 2 receive updated motion planning libraries, bug fixes, and hardware drivers. Upgrading core components like ros2-control and MoveIt2 to match Ubuntu Resolute and C++20 standards maintains buildfarm stability. Tracking these releases allows robotics software leads to plan dependency upgrades without breaking production interfaces.
OSRF distribution maintainers emphasized that synchronization across amd64 and arm64 targets stabilizes deployments on edge compute modules. MoveIt2 maintainers encouraged community testing on the Lyrical release branch to catch API deprecations early.
Texas Instruments introduced the TCAN6062 on Wednesday, marketed as the first commercially available Controller Area Network extended data-field length (CAN XL) transceiver tailored for industrial and humanoid robots. The chip supports frame payloads up to 2,048 bytes and transmission speeds up to 20 Mbps while integrating signal improvement circuitry to reduce line ringing by up to 80%. Developed alongside standards bodies CAN in Automation and Bosch, the transceiver provides a deterministic, high-bandwidth internal bus.
Why it matters
Modern humanoid architectures with dozens of dynamic joint actuators, tactile sensors, and high-rate IMUs face internal wiring and communication bottlenecks under traditional CAN FD networks. Commercial CAN XL silicon bridges the gap toward high-speed industrial Ethernet without requiring complex point-to-point star topologies or heavy cabling harness builds. The expanded 2,048-byte payload allows sensor fusion packets to travel across joint actuators with lower latency and deterministic timing.
Bosch and CAN in Automation representatives praised the launch as a milestone that brings the CAN XL standard into production hardware. Embedded robotics engineers noted that while 20 Mbps CAN XL simplifies wiring compared to Ethernet switches, existing microcontrollers must update their controller IP to leverage the extended payload capabilities fully.
Toronto-based startup Veeda AI emerged from stealth on Wednesday after closing a $90 million seed funding round co-led by Khosla Ventures and Radical Ventures. Founded by former NVIDIA VP of AI research Sanja Fidler alongside Zan Gojcic and Huan Ling, the company is engineering multimodal foundation world models designed to simulate physical environments for training embodied agents. The capital raise represents one of the largest seed rounds for a Canadian AI firm and will fund compute infrastructure and simulation hires.
Why it matters
The massive valuation and seed backing for Veeda AI highlight how simulation infrastructure has become a primary investment priority in physical AI. Because collecting real-world teleoperation trajectories is slow and hardware-intensive, generative world models that accurately simulate physics and contact dynamics offer a scalable alternative to train robot brains. The high-profile spin-out of NVIDIA's Toronto research leadership underscores the fierce competition to build commercial simulation environments outside incumbent tech stacks.
Veeda AI co-founder Sanja Fidler stated that interactive world models are essential to bypass the safety risks and sample inefficiencies of physical trial and error. Venture backers at Khosla and Radical emphasized that scalable spatial simulation represents the core missing layer in bringing general-purpose robotics to commercial readiness.
FORT Robotics announced a definitive SPAC business combination on Thursday with Newbury Street II Acquisition Corp., valuing the combined entity at a pro forma enterprise value of $556.6 million. The transaction delivers $182 million in net cash proceeds while retaining existing shareholders at a 67% equity stake. Generating $11.6 million in 2025 revenue, the firm manufactures safety-certified hardware wireless stop systems and software safety architecture, following its May acquisition of Mapless AI teleoperation technology.
Why it matters
The transaction tests whether functional safety architectures and wireless e-stops can command high-margin software valuations rather than traditional industrial hardware pricing. As autonomous mobile robots and heavy equipment deploy alongside human workforces, certified hardware safety islands become mandatory for enterprise compliance. Securing public capital allows FORT Robotics to scale its machine-neutral safety layer across heterogeneous industrial fleets.
FORT Robotics executive management stated that the public transaction accelerates the deployment of certified safety layers across autonomous industrial machines. Financial analysts observed that converting hardware e-stops into recurring software safety subscriptions will be critical to justify the $556 million valuation multiple.
The U.S. FDA granted De Novo marketing authorization on Wednesday to Dutch medtech firm Vitestro for Aletta, establishing a novel regulatory classification for autonomous robotic phlebotomy. The standalone system combines near-infrared light and Doppler ultrasound to locate suitable arm veins, execute needle insertion, swap blood tubes, and apply bandages without direct human operator intervention. Clinical trial data across 10,000 patient procedures demonstrated draw success rates matching or exceeding human operators, allowing a single supervising phlebotomist to manage up to three units simultaneously.
Why it matters
Establishing the first FDA De Novo pathway for fully autonomous vascular access creates a legal and clinical precedent for procedural healthcare robotics. By shifting the clinical supervision ratio from 1:1 to 1:3, the device directly mitigates acute phlebotomist staffing shortages in outpatient diagnostic networks. This decision indicates regulatory willingness to approve autonomous, contact-heavy medical hardware when backed by multi-modal ultrasound sensing and extensive clinical safety data.
Vitestro executive leadership highlighted that De Novo authorization validates years of clinical trial data and paves the way for immediate U.S. commercial deployment. Clinical lab directors noted that while automated blood collection increases throughput, successful adoption will depend on patient comfort and integration with existing electronic health record systems.
Wandercraft announced updated commercial details on Wednesday following U.S. FDA clearance for Eve, a hands-free personal self-balancing exoskeleton for adults with spinal cord injuries. Unlike clinical rehabilitation systems that require crutches or parallel bars, Eve utilizes dynamic internal balancing algorithms to allow hands-free walking on flat indoor and outdoor surfaces under companion supervision. Scheduled for U.S. commercial launch in September 2026, the company expects Medicare reimbursement determinations within 60 to 90 days following mandatory clinical fitting protocols.
Why it matters
Transitioning self-balancing exoskeleton technology from clinical physical therapy centers into personal daily use represents a major functional milestone for mobility assistance. Eliminating the need for upper-body crutch support allows wheelchair users to execute daily standing and walking tasks hands-free. Securing Medicare coding and coverage will serve as the crucial financial test determining broad commercial adoption in personal assistive robotics.
Wandercraft leadership emphasized that hands-free balance control restores everyday functional independence for personal mobility outside clinical walls. Physical rehabilitation specialists cautioned that compulsory clinical fitting and companion supervision remain necessary safety requirements before personal home deployment.
NVIDIA has released detailed performance benchmarks for the Cosmos 3 Edge 4-billion-parameter model running on its Jetson Thor modules, following their respective launches we've tracked over the past month. The architecture integrates a 2B Nemotron-based reasoning backbone trained on 76,000 teleoperated trajectories. In closed-loop RoboLab simulations across 120 manipulation tasks, the post-trained policy achieved a 22.9% success rate entirely on-device, generating 100 Hz action chunks at 15 Hz with a 1.53-second control latency.
Why it matters
Running multi-billion parameter world models directly on robot compute boards resolves the critical network latency and offline reliability barriers that prevent cloud-tethered VLAs from operating in industrial plants. The release validates Jetson Thor's Blackwell-based architecture as a viable host for real-time physical AI inference within tight thermal and power envelopes. This local execution capability lowers deployment costs for autonomous mobile manipulators and industrial arms by removing cloud subscription overhead. System architects can now evaluate closed-loop physical reasoning directly at the edge.
NVIDIA's robotics engineering team highlights that local execution eliminates wireless dropped-packet vulnerabilities in factory settings. Independent robotics developers on ROS Discourse note that while a 22.9% zero-shot success rate demonstrates progress for multi-task policies, enterprise deployments still require significantly higher task completion rates and faster chunk generation before replacing specialized deterministic control scripts.
Qualcomm announced at ModCon 2026 on Wednesday that it has fully open-sourced Modular's AI software stack and the Mojo 1.0 programming language specification under the Apache 2.0 license with LLVM exceptions. The release follows Qualcomm's $3.9 billion acquisition of Modular and includes the complete compiler codebase. The unified runtime supports multi-hardware cross-compilation across Qualcomm Snapdragon, AMD GPUs, NVIDIA GPUs, Apple silicon, AWS Trainium, and Google TPUs, demonstrated on stage alongside hardware partners.
Why it matters
Open-sourcing Mojo 1.0 under a permissive Apache 2.0 license provides an open alternative to proprietary vendor software lock-in like CUDA. For edge AI and robotics developers target-compiling custom neural operators across heterogeneous chips—such as Snapdragon SoCs and discrete accelerators—a unified open-source language drastically cuts software maintenance friction. Qualcomm's commitment to cross-hardware compatibility helps level the playing field for non-NVIDIA silicon in embedded robotics.
Qualcomm software leads emphasized that open-sourcing the full stack guarantees ecosystem longevity and prevents vendor lock-in for enterprise AI developers. Industry partners including AMD publicly supported the move on stage, while open-source maintainers cautioned that building a vibrant contributor community around a newly open-sourced compiler requires sustained engineering transparency.
Marvell Technology filed an SEC 8-K disclosure on Wednesday announcing a major commercial agreement with Google LLC to co-design custom AI semiconductors for Google's Tensor Processing Unit ecosystem. Under the deal, Marvell will supply custom inference accelerators, CXL memory controllers, and optical interconnects. In connection with the agreement, Marvell issued Google a warrant to purchase up to 58.9 million shares valued at $12.2 billion, with vesting tied to $500 million revenue tranches through fiscal 2033.
Why it matters
Securing Google alongside existing custom silicon deals with Amazon and Microsoft cements Marvell as the primary independent co-design competitor to Broadcom in hyperscale AI silicon. The integration of high-bandwidth CXL memory controllers and custom interconnects highlights how ASIC architecture is shifting to overcome memory-bandwidth bottlenecks during large-model inference. The revenue-vested warrant structure directly aligns Google's infrastructure buildout with Marvell's silicon roadmap.
Financial analysts noted that the agreement provides multi-year revenue visibility for Marvell's custom ASIC division while boosting share prices by 12%. Semiconductor industry observers pointed out that hyperscalers are increasingly bypassing off-the-shelf GPU architectures in favor of tailored ASIC hardware to drive down per-token inference costs.
Engineers at the Swiss Federal Institute of Technology Lausanne (EPFL) published research on Wednesday detailing an acoustic propulsion system for miniature robots using 3D-printed Helmholtz resonators. By converting focused ultrasound frequencies into directed air jets, the team propelled microfliers weighing between 150 and 184 micrograms, reaching rotor speeds up to 13,000 RPM and achieving a thrust-to-weight ratio of 4.9. The motorless design uses acoustic frequency shifts as control signals to adjust lift and direction remotely.
Why it matters
Bypassing electromagnetic motors and onboard battery packs removes the primary weight and scaling constraints that prevent sub-milligram flying microrobots from achieving sustained flight. Utilizing acoustic resonance fields allows external power beam steering, enabling untethered micro-swarms to operate in enclosed or hazardous environments. Overcoming viscous drag forces at micro-scales opens new hardware design avenues for targeted environmental sensing and non-invasive medical micro-tools.
EPFL lead researchers highlighted that using sound frequencies as a multiplexed control signal enables selective steering of individual microfliers within a shared acoustic field. Independent microrobotics experts noted that while the thrust-to-weight ratio is impressive, practical outdoor deployment remains limited by acoustic attenuation and ambient wind disturbances.
Researchers at The Hong Kong University of Science and Technology (HKUST), led by Professor Richard GU Hongri, introduced an automated robotic nanoprobe in Science Advances on Thursday. The system integrates nanoscale tips with nanoelectrodes to detect reactive oxygen and nitrogen species (ROS/RNS) surges in real time, automatically activating dielectrophoretic nanotweezers to isolate individual mitochondria from living cells within 100 nanometers. Extracted mitochondria successfully underwent fusion and fission after transplantation into recipient cells, confirming structural viability.
Why it matters
Traditional single-organelle research relies on invasive fluorescent dyes and manual micromanipulators that frequently damage cellular structures. Combining real-time electrochemical sensing with automated dielectrophoretic nanotweezers provides a standardized, label-free workflow for sub-cellular surgery. This automated precision opens new diagnostic and therapeutic avenues for studying neurodegenerative conditions and metabolic disorders at the single-organelle level.
HKUST lead researcher Richard GU Hongri stated that label-free extraction preserves native organelle health for functional transplantation research. Biomedical engineers praised the automated closed-loop detection, noting that sub-100nm mechanical accuracy reduces operator error during live-cell manipulation.
Researchers at Julius-Maximilians-Universität Würzburg, led by Professor Bert Hecht, published findings on Wednesday detailing sub-micrometer nanorobots capable of capturing and transporting individual bacteria. Equipped with plasmonic nanoantennas, the microdrones convert redirected laser light into photon recoil thrust, executing sharp 90-degree turns when light polarization is adjusted. In lab trials, the light-driven devices successfully navigated dense biological fluid to capture, transport, and release targeted bacterial cells.
Why it matters
Harnessing the momentum of individual photons for mechanical propulsion eliminates onboard electrical circuitry and magnetic coils in sub-micron robotics. Operating at scales 50 times smaller than a human hair enables direct physical manipulation of individual cells and pathogens without tethered tools. This optical control mechanism opens new pathways for targeted micro-cleaning, precise biobanking, and localized cell sorting.
Professor Bert Hecht's team highlighted that photon recoil propulsion provides precise directional control without relying on chemical fuels or external magnetic arrays. Biophysicists noted that scaling the technology from single-cell lab dishes to complex, opaque in-vivo environments will require overcoming light scattering constraints.
A research team led by Jackson K. Wilt published details on Thursday of a multi-material 3D printing technique for soft robots using photopolymerizable polyurethane-acrylate resin paired with a sacrificial fugitive ink. Extruded through a custom co-axial nozzle, the Pluronic F-127 fugitive ink forms intricate internal fluidic channels that are dissolved in water post-UV curing. The process allowed researchers to print seamless artificial muscle actuators, compliant hinges, and articulated human-like hands with internal pneumatic routing in a single manufacturing pass.
Why it matters
Conventional soft pneumatic actuators require labor-intensive manual molding, layering, and adhesive bonding, which frequently leads to delamination leaks under high pressure. Direct multi-material 3D printing with sacrificial fugitive inks enables complex, non-linear internal air channels to be manufactured monolithically. This lowers fabrication barriers for custom bio-inspired grippers, soft exosuits, and prosthetics requiring intricate pneumatic routing.
The research team emphasized that eliminating manual assembly drastically improves structural reliability and repeatability in soft actuator production. Materials scientists noted that while water-soluble fugitive inks streamline channel creation, scaling production speed requires faster UV-curing photopolymers that maintain high elastic strain.
A research team led by Assistant Professor Tan Yu Jun at the National University of Singapore (NUS) published details on Thursday of a self-healing magnetoelectric sensory skin (SMES) for underwater soft robots. Constructed from a self-healing elastomer embedded with liquid-metal conductors, the skin recovers nearly 100% of its electrical function within 10 days after underwater punctures. Utilizing electromagnetic induction via micro-magnets and liquid-metal coils, the self-powered sensor responds in 41 milliseconds and survived 10,000 cyclic loading tests.
Why it matters
Subsea soft manipulators face continuous physical abrasion, punctures, and water ingress that rapidly destroy conventional electronics. Combining self-healing elastomers with liquid-metal wiring allows soft robotic skins to repair physical cuts autonomously without human intervention. Eliminating external battery requirements via self-powered induction extends the operational lifespan of underwater inspection and bio-sampling manipulators.
The NUS engineering team demonstrated smart diving gloves and puncture-resistant soft robotic hands operating continuously underwater. Marine robotics researchers noted that while self-healing recovery times of 10 days are suitable for long-duration deployments, faster active healing mechanisms are desirable for immediate field repairs.
Pony.ai released its Q2 2026 financial results on Wednesday, reporting total revenue of $36.2 million (up 68.8% year-over-year) and robotaxi-specific revenue surging 691.2% to $12.1 million. Direct passenger fare income increased 849%, driven by seventh-generation vehicle rollouts in Guangzhou and Shenzhen. Concurrently, Pony.ai launched commercial rides in Zagreb, Croatia alongside Uber and Verne—marking its first European market—while expanding its international pipeline to over 4,000 planned robotaxi units.
Why it matters
The dramatic rise in direct C2C passenger fare revenue confirms that urban riders are actively paying commercial rates for driverless transit. By partnering with fleet operators like Verne and platforms like Uber, Pony.ai limits direct capital expenditure while expanding into international markets. However, company leadership highlighted that regulatory permitting timelines—rather than vehicle technology—have become the primary bottleneck constraining global scaling velocity.
Pony.ai CEO James Peng stated that the company's asset-light partnership strategy enables rapid scaling across international cities using its PonyWorld 2.0 world model. Financial analysts noted that despite surging revenues, widening GAAP operating losses underscore the high continuous R&D burn required to maintain driverless fleets.
Amazon announced a major expansion of its Prime Air drone delivery service on Wednesday, adding 11 new operating hubs across Arizona, Florida, Kansas, Michigan, and Texas. The rollout expands Prime Air's footprint sixfold, targeting nearly 500 U.S. cities by year-end 2026. Operating under FAA Part 135 certification, the service offers free ultra-fast delivery for Prime members on eligible items under 5 pounds, relying on onboard vision-based Detect-and-Avoid (DAA) hardware.
Why it matters
Scaling drone delivery into hundreds of cities marks a transition from localized municipal pilots to nationwide commercial aerial logistics. Relying on FAA Part 135 beyond-visual-line-of-sight (BVLOS) approvals enables Amazon to compete directly with rival drone networks operated by Alphabet's Wing and Walmart. The expansion puts real-world operational pressure on onboard vision systems to maintain airspace safety at scale.
Amazon Prime Air leadership stated that onboard Detect-and-Avoid sensors provide the necessary redundant safety layers for dense suburban flights. Aviation safety advocates emphasized that scaling to 500 cities will test air traffic management coordination as commercial drone density increases.
Ecovacs released detailed performance specs on Wednesday for its GOAT O1200 LiDAR Pro robotic lawn mower, priced at £949. The consumer platform combines 360-degree LiDAR and camera-based visual navigation to eliminate physical boundary wires across properties up to 1,200 m². A key hardware addition is its TrueEdge side-trimmer mechanism, which extends beyond the chassis wheel line to cut lawn borders alongside electronic height adjustments ranging from 3cm to 8cm.
Why it matters
Incomplete perimeter edge-cutting remains one of the primary consumer complaints regarding autonomous lawn mowers, forcing homeowners to manually trim borders. Integrating active side-extending trimmers directly into wire-free LiDAR navigation chassis addresses this usability gap. This design shift illustrates how consumer outdoor robotics is moving past basic lawn coverage toward full-service yard maintenance.
Product reviewers noted that the TrueEdge mechanical trimmer significantly reduces manual edging, though it generates higher acoustic noise during edge passes. Industry analysts pointed out that wire-free LiDAR mowers under £1,000 are intensifying price competition for traditional RTK GPS-only manufacturers.
On-Device Inference Stack Shifts to Local World Models Deployments like NVIDIA's Cosmos 3 Edge on Jetson Thor demonstrate that multi-billion parameter foundation architectures are successfully moving directly onto edge hardware, eliminating cloud latency and offline vulnerabilities for real-time manipulation.
Kinematic Motion Capture Replaces Synthetic Teleoperation Open-source dataset releases featuring studio-grade motion capture and FrameNet semantic indexing provide physically grounded trajectories that allow developers to bypass fragile teleoperation pipelines for whole-body humanoid training.
Public Market Capitalization Secures Capital for Hardware Scale Blockbuster public debuts on Shanghai's STAR Market demonstrate that pure-play humanoid and quadruped manufacturers can command strong public valuations, providing liquidity to scale manufacturing capacity.
Decentralized Micro-Actuation Bypasses Traditional Mechanical Transmission Innovations in axial flux micro-actuators, piezoelectric drives, and acoustic Helmholtz resonators are reducing reliance on heavy gearboxes, unlocking higher power density for dexterous manipulators and micro-scale robotics.
Clinical Autonomy Gains Ground Through Regulatory Pathways De Novo authorizations and FDA clearances for self-balancing exoskeletons and automated blood draw systems mark a transition from human-operated medical robotics to semi-autonomous clinical systems.
What to Expect
2026-08-22—World Humanoid Robot Games open in Beijing, featuring competitive domestic service and industrial robot trials.
2026-09-02—SEMICON Taiwan 2026 kicks off, featuring AAEON's Jetson Thor edge AI wafer inspection and pick-and-place systems.
2026-10-19—FDA public comment period closes for docket FDA-2026-N-7874 on Generative AI Medical Device frameworks.
How We Built This Briefing
Every story, researched.
Every story verified across multiple sources before publication.
🔍
Scanned
Across multiple search engines and news databases
500
📖
Read in full
Every article opened, read, and evaluated
149
⭐
Published today
Ranked by importance and verified across sources
20
— The Robot Beat
🎙 Listen as a podcast
Subscribe in your favorite podcast app to get each new briefing delivered automatically as audio.
Apple Podcasts
Library tab → ••• menu → Follow a Show by URL → paste