New articles on Wevolver

Help Shape the 2027 Edge AI Technology Report Edge AI is evolving rapidly, from compact language models and multimodal systems to robotics, AI accelerators, and real-world deployment. As we develop the 2027 Edge AI Technology Report, we want to hear from engineers, developers, and researchers about the topics that matter most. Complete our 5-minute anonymous survey to help shape coverage of the technologies, challenges, case studies, and practical guides featured in the report. No contact details are required, and responses will be reviewed in aggregate. Thank you for sharing your perspective. Take the survey here: https://wevlv.co/edgeaisurveygrps

Falcon 9’s first stage lands on the Of Course I Still Love You droneship

An interactive exhibit showing how a computer processor does math. Video Credit: ProofofMaro

This video showcases the advanced engineering of the ABENICS spherical gear, highlighting its ability to transmit power smoothly under high stress. Using Finite Element Method (FEM) analysis, it demonstrates that each aluminum alloy tooth can endure forces between 2,000 and 7,600 Newtons before yielding. A color-coded safety factor distribution illustrates how stress is effectively spread across multiple teeth, emphasizing the gear’s durability and efficient load distribution. Video Credit: Garage Folks

Ultra-stable bidirectional self-balancing unicycle. Video Credit: XIXI Wang

Developed by researchers from Ant Group and Stanford University, UME is a wearable upper-limb exoskeleton designed for robot teleoperation and learning. When the operator moves an arm, the robot follows. But unlike many conventional teleoperation systems, the communication also runs in the opposite direction. When the robot encounters resistance, UME feeds the measured joint torque back into the operator’s arm in real time. If force-aware demonstration systems become scalable, robot foundation models could increasingly learn not only visual motion, but also the physical judgment required for safe, contact-rich work. Learn more: https://ume-exo.github.io/

A French startup, Tornyol, has developed a 40-gram autonomous microdrone that hunts flying insects using sonar rather than cameras. By analysing wingbeat Doppler signatures, it can identify and intercept specific insects mid-air. The technology could enable chemical-free, targeted pest control, protecting beneficial species while eliminating unwanted pests—a potential shift from traditional agricultural chemicals to robotics and signal processing. Video Credit: Tornyol

Inspired by limbless animals, researchers developed a soft robot that crawls across complex terrain using inflatable actuators, kirigami skin, and asymmetric friction. Precise inflation sequences enable forward movement, while asymmetric gaits provide steering and rotation. Proximity sensors support adaptive obstacle avoidance, highlighting potential for search-and-rescue, monitoring, and inspection applications. Collaborators: Tirado J., Parvaresh A., Seyidoğlu B., Bedford D.A., Jørgensen J. & Rafsanjani A. Credit: SDU Soft Robotics

Embedded systems have become harder to debug as event-driven code, multicore architectures, real-time constraints, and power sensitivity add complexity. This on-demand webinar examines debugging techniques that go beyond basic breakpoints and printf statements. Learn how conditional breakpoints, trace and stack analysis, log breakpoints, skip counts, power debugging, automated testing, and static and dynamic analysis can help engineers identify defects and understand system behaviour. Watch the full session to explore practical debugging strategies for modern embedded development: https://wevlv.co/4fukyxq

Falcon 9’s first stage has landed on the A Shortfall of Gravitas droneship Video Credit: SpaceX #space #falcon9

Neuralink has been exploring how clinical trial participants can control a powered wheelchair using only their minds. The Neuralink implant decodes movement intentions in real time to control an on-screen cursor. The cursor operates a custom app displaying a live camera feed of what is in front of the participant. Moving the cursor up drives the wheelchair forward, down moves it backward, while left and right control steering. The farther the cursor moves in a direction, the faster the wheelchair travels. The video offers an early look at Neuralink’s work on telepathic wheelchair control, an important step toward restoring independent mobility for people with paralysis. Video Credit: Neuralink

NUS researchers developed OstraBot, a biohybrid swimming robot powered by self-trained lab-grown muscle. By coupling muscle tissues so they continuously exercise each other, researchers achieved record strength without external stimulation. OstraBot reached 467 millimetres per minute, demonstrating potential for biodegradable robots, environmental monitoring and minimally invasive medical devices. Video Credit: National University of Singapore

This demo shows how AI-powered computer vision can help restaurants measure drive-thru performance in real time. Plainsight’s visual AI detects vehicles, tracks queue activity, measures wait times, and identifies abandonment events when customers leave before completing an order. These insights can help quick service restaurants and fast casual brands better understand speed of service, lost sales, operational bottlenecks, staffing needs, and customer experience across the drive-thru. Instead of relying only on manual observation or transaction data, computer vision can provide real-time visibility into what is happening outside the restaurant, including how long cars wait, where delays occur, and how drive-thru congestion may impact revenue. Video Credit: Plainsight

Landing burn and splashdown of Starship on Flight 13. Video Credit: SpaceX

Xynova demonstrated the teleoperation of its Flex 2 dexterous hand using MANUS Metagloves Pro Haptic. The demonstration showed how human hand motion, captured in real time, can be transferred to a robotic platform through a control pipeline grounded in continuous finger articulation data. Video Credit: MANUS™

Columbia Engineering researchers have developed "Robot Metabolism," a system that enables robots to grow, repair themselves, and become more capable by incorporating parts from other robots or their surroundings. Using modular Truss Links, the robots self-assemble into increasingly complex structures and adapt to new challenges. The research aims to create autonomous robot ecosystems for applications such as disaster response and space exploration, where machines can sustain and improve themselves without human intervention. More details: https://www.engineering.columbia.edu/about/news/robots-grow-consuming-other-robots Video Credit: Columbia Engineering

Formnext Asia Shenzhen 2026 brings together manufacturers, engineers, technology providers, and business leaders from across the additive manufacturing industry to focus on innovation, industry collaboration, and the latest developments in digital manufacturing and precision production. On August 26–28, 2026, professional visitors from around the world will attend Formnext Asia Shenzhen to collaborate with more than 320 exhibitors and learn more about the rapid expansion of industrial and consumer 3D printing technologies. Join exhibitors like Bambu Lab, Creality, Farsoon Technologies, Raise3D, HBD, Intamsys, UnionTech, Han's Laser, Meshy AI, and Tripo AI to discover how to print beyond limits. Learn more about the event: https://wevlv.co/4wNODhh

Engineer and designer Will Cogley controls a robot head with his own face.

Researchers at ETH Zürich demonstrated a 6-DOF electromagnetic levitator that catches a thrown ball mid-flight. Using neural networks instead of traditional analytical models, it predicts trajectories and controls position and orientation without symmetry assumptions or extensive calibration. Video credit: Jasan Zughaibi