Good morning, robotics enthusiasts. Amazon is reportedly taking its Prime promise to the next level, quietly developing AI-powered humanoids that could soon handle last-mile deliveries with unprecedented speed and efficiency.

As these futuristic couriers threaten to upend the traditional delivery workforce, it’s worth asking: How far should we go to make Prime even more prime?

In today’s robotics rundown:

• Amazon testing humanoid delivery bots

• Hugging Face's MacBook-ready robotics AI

• Unitree, Reborn team up to fast-track humanoids

• Aerones nabs $62M for wind turbine–cleaning bots

• Quick hits on other robotics news

LATEST DEVELOPMENTS

AMAZON

📦 Amazon testing humanoid delivery bots

Image source: Ideogram/The Rundown

The Rundown: Amazon is reportedly accelerating the development of AI-powered software for humanoids, designed to tackle the last-mile delivery challenge — a move that could disrupt the jobs of human delivery workers.

The details:

• Amazon is developing proprietary AI software for humanoid deliveries and nearing completion of a “humanoid park” in San Francisco for testing.

• The Information reports that Amazon is working with Chinese robotics company Unitree, with trials involving robots exiting Rivian electric vans.

• Amazon’s warehouses already feature thousands of robots, from mobile robot Proteus to advanced picking arms such as Sparrow and Cardinal.

• The project is raising concerns about job losses among delivery drivers and gig workers in Amazon’s vast logistics network.

Why it matters: While Amazon has not officially commented on the project, the implications are huge. The company’s relentless drive for efficiency relies increasingly on AI and robotics to optimize every step of its supply chain—and now that potentially extends to robot-powered deliveries at your doorstep.

TOGETHER WITH ENCORE

🤖 Turn multi-sensor data Into Physical AI

Image source: Encord

The Rundown: Encord enables Physical AI teams to unify multi-sensor data curation and annotation on one platform. Use Encord to accelerate iteration cycles and achieve safe real-world deployments with balanced, annotated, high-quality datasets.

Use Encord to transform multimodal data:

• Build VLA datasets with synchronized multi-sensor views

• Ensure safe real-world operation by supervising edge-case behavior

• Reduce labeling time by 95% through model-assisted workflows

Try Encord now.

HUGGING FACE

Hugging Face’s MacBook-ready robotics AI

Image source: Hugging Face

The Rundown: Hugging Face just unveiled SmolVLA, a compact open-source robotics model that the AI startup says dramatically lowers the barrier for sophisticated robotics projects at home, and is small enough to run on a MacBook.

The details:

• At just 450M parameters, SmolVLA is small enough to run on a single consumer GPU, yet matches or outperforms much larger models.

• The model is trained entirely on open, compatibly licensed datasets shared by the robotics community via Hugging Face’s platform.

• Its architecture consists of a compact vision-language model (SmolVLM-2) that processes RGB images, sensorimotor states, and natural language.

• There’s also a lightweight “action expert” transformer that outputs robot control commands in real time.

Why it matters: SmolVLA is fully open source, with downloadable code, model weights, and datasets. The AI is central to Hugging Face’s growing robotics initiative, which includes the LeRobot library for robotics models and datasets, the buyout of Pollen Robotics, and the release of low-priced hardware kits like the SO-101 arm.

UNITREE/REBORN

🤝 Unitree, Reborn team up to fast-track humanoids

Image source: Reborn/X

The Rundown: Chinese robotics giant Unitree Robotics announced a long-term collaboration with UK startup Reborn to deploy Reborn's open-source framework, Roboverse, to accelerate learning and capabilities of Unitree’s humanoids.

The details:

• Roboverse is a comprehensive framework that combines high-fidelity simulation environments, vast human motion datasets, and unified benchmarks.

• Reborn says it enables simulation-based training up to 30x faster than traditional methods, allowing robots to learn complex, human-like tasks.

• By bridging the “sim-to-real” gap, Roboverse enables skills learned in simulation to carry over more effectively to real-world robotic applications.

• This partnership allows Unitree’s robots to move beyond pre-programmed routines, enabling them to learn fine manipulation and adaptive movement.

Why it matters: With Reborn’s training technologies, Unitree is positioned to make intelligent, adaptable robots accessible at scale. Both companies are also hosting hackathons and fostering a collaborative developer community, aiming to deploy over 1K humanoids in China by 2026.

AERONES

💨 Aerones nabs $62M for wind turbine–cleaning bots

Image source: Aerones

The Rundown: Latvian startup Aerones snagged $62M to launch its AI-powered robots onto wind farms across the globe — in a move aimed to disrupt an industry still reliant on hands-on labor and costly downtime.

The details:

• Aerones’ robots, used by GE and Enel, can inspect, clean, and repair turbine blades in half the time of human crews, minimizing downtime.

• The company says this not only slashes operational costs but also boosts turbine uptime, contributing to an additional 400,000 MWh of clean electricity.

• It aims to scale operations to service thousands of turbines in over 30 countries, focusing on key wind markets in North America, Europe, and Asia.

• Aerones, which just opened a new hub in Dallas, also offers internal blade crawlers, drone inspections, and anomaly detection systems.

Why it matters: Major players like Aerones and Clobotics have established themselves in the wind turbine robot market with cutting-edge tech for onshore and offshore needs. As regulatory standards tighten and wind farms scale up globally, the race for autonomous maintenance solutions is intensifying, making this a sector to watch.

QUICK HITS

📰 Everything else in robotics today

Tesla submitted new trademark applications for the term “Tesla Robotaxi,” following earlier setbacks, as it ramps up its upcoming self-driving vehicle service.

UK-based startup Humanoid released a teaser video highlighting its upcoming general-purpose humanoid, HMND 01, set to debut later this year.

China’s Agibot partnered with Pepsi to launch a Pepsi-branded humanoid dubbed “Fizz Bot,” introduced on stage with David Beckham at the launch event.

Pony.ai announced a strategic alliance with Shenzhen Xihu Corp., Shenzhen’s largest taxi operator, to roll out a fleet of over 1K Pony.ai robotaxis across the city.

California startup Impossible Metals developed autonomous robots that harvest critical battery metals from the seabed with minimal ecological disruption.

Researchers developed a new technique that enables robots to more accurately recognize human facial expressions, making them more emotionally aware.

Whale Dynamic partnered with Noodoe to create an end-to-end ecosystem linking fully autonomous delivery vans with AI-powered charging management.

The Robero NEO—now crowdfunding on Kickstarter—is an AI-powered golf caddy that autonomously follows you around the course for eight hours on one charge.

U.S. researchers created an innovative control system that enables aerial robots to handle flexible materials such as cables and hoses in real time.

Aldebaran, a pioneering French robotics company, has been placed in judicial liquidation and is ceasing operations after failing to find a buyer.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, June 6th at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll confidently be able to incorporate Flux Kontext into your creative workflows for consistent, high-quality results.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

See you soon,

Rowan, Jennifer, and Joey—The Rundown’s editorial team