Amazon's robot army gets chatty

PLUS: This humanoid is summiting Earth's toughest peaks

Good morning, robotics enthusiasts. Amazon’s newest warehouse robot ditches the code — you just tell it what to do, and it sorts out the priority, the route, and the timing.

Unveiled in London alongside an $11.6B European logistics blitz, the new Proteus can parse plain-language orders, haul heavy carts, and decide what needs doing first. Amazon is dangling tens of thousands of new logistics jobs, too, even as its robot fleet passes 1M and the layoffs keep coming for white-collar workers.

In today’s robotics rundown:

  • Amazon’s new robot takes verbal orders

  • Humanoid climbing Earth’s extreme peaks

  • Europe’s robotaxi era finally begins

  • 1X doubles down on world models

  • Quick hits on other robotics news

LATEST DEVELOPMENTS

AMAZON

Image source: Amazon

The Rundown: Amazon’s next-gen Proteus warehouse robot now takes orders in plain conversational language — a friendlier face for automation that the company rolled out in London this week alongside a €10B (~$11.6B) European expansion.

The details:

  • The new Proteus is an autonomous mobile robot that can navigate warehouses, move heavy carts, and prioritize tasks based on conversational instructions.

  • Amazon showcased the robot in London as part of an €10B European logistics push, with plans to deploy the new model starting in 2027.

  • The reveal lands against Amazon’s automation-driven white-collar cuts, with the total reportedly reaching 30K.

  • As a counterweight, Amazon says the spend will add 25K European logistics jobs and expand a $1B worker-training program.

Why it matters: Amazon runs the world’s largest robot fleet — over 1M machines closing in on its 1.56M workers — and Proteus is one of its most prominent bots. Robotics now backs 75% of Amazon’s deliveries; the question is whether 25K promised European jobs can outlast the same efficiency drive that just cut 14K office roles.

ROBOTICS RESEARCH

Image source: X @pabloberlangab

The Rundown: A modified Unitree G1 humanoid named “Pemba” just reached the top of Chimborazo, Ecuador’s highest volcano — the first challenge in a stunt-meets-research bid to walk a bipedal robot up three of the world’s most punishing peaks.

The details:

  • Pemba just reached the 6,200-meter summit of Chimborazo, the point on Earth’s surface farthest from the planet’s center, after a 16‑hour push.

  • The robot walked autonomously on sections below 30 degrees of incline and was carried over steeper, more technical terrain by human climbers.

  • Project Pemba is testing whether humanoids can replace dense networks of stationary cameras for wildlife monitoring and remote sensing.

  • Next up is Hawaii’s Mauna Kea, followed by an Everest attempt that is currently stuck in limbo because Nepal lacks any legal framework for robotic expeditions.

Why it matters: Marching a humanoid up high-altitude volcanoes and, eventually, Everest turns today's stunt bots into testbeds for real-world deployment in disaster zones, polar stations, and even off-world missions, demonstrating their ability to operate where conditions are too harsh or risky for humans.

ROBOTAXIS

Image source: Waymo

The Rundown: Self-driving taxis are beginning to roll out across Europe after significant growth in the U.S. and China, where private robotaxi fleets doubled last year to about 8K vehicles operating across more than two dozen major cities.

The details:

  • The EU is rolling out a new “test bed” framework to fast-track approvals and let companies run trials across borders.

  • Europe’s first commercial-style trial launched in April in Zagreb, where Pony.ai is running about 10 robotaxis alongside Uber and Croatian startup Verne.

  • More will arrive in 2026: London (Waymo; Wayve with Uber; Baidu’s Apollo Go), Madrid (WeRide with Uber), and Munich (Uber with Momenta).

  • By 2035, forecasts put the global robotaxi count from 700K to 6M, with paid service in cities like London and Madrid possibly arriving as soon as 2027.

Why it matters: Robotaxis are rolling onto European streets, pushing cities to rethink how people move and who sits behind the wheel. Uptake is slower here — stricter safety rules, strong public transit, and a warier public — so every new service becomes a live test of whether the driverless model actually holds up.

1X

Image source: 1X

The Rundown: Palo Alto’s 1X is setting up a dedicated World Model Lab to build physics-grounded “world models” from scratch and pipe them into its NEO humanoid — right as the startup’s California factory ramps up work for U.S. home shipments.

The details:

  • The lab centers on embodied world models that learn space, motion, and object dynamics before any task-specific fine-tuning kicks in.

  • CEO Bernt Børnich tapped ex‑Luma AI researcher Samarth Sinha to lead the lab, bringing large multimodal models expertise into humanoid policy learning.

  • It builds on the world model 1X rolled out in January, which feeds on everything from human video to teleoperated and on-policy NEO data.

  • The Hayward, CA, plant is tooled for 10K NEO units a year, with first-year capacity already sold out and consumer deliveries slated for late 2026.

Why it matters: 1X is betting that owning the whole stack is the only way to crack general-purpose home robots before Tesla and Figure. If the world-model approach delivers real generalization, NEO could teach itself new chores instead of waiting on engineers to script each one.

QUICK HITS

Nvidia CEO Jensen Huang said the company is deepening its partnership with South Korea’s LG Group to co-develop humanoids and architect next-gen AI data centers.

Hyundai and Nvidia are expanding their alliance to industrialize AI robotics across Hyundai’s factories, vehicles, and $6.5B Saemangeum AI innovation hub.

Robotics startup Generalist AI raised $400M in a Radical Ventures–led round at a $2B valuation to scale its GEN-1 embodied robotics foundation model.

Uber opened a waitlist in its app for London riders to get early, no-extra-cost access to Wayve-powered robotaxis with safety drivers on board.

A Unitree humanoid performing a martial‑arts routine at a Xinjiang tourist site accidentally kicked a young boy in the stomach, knocking him down.

Three UBC students built an AI air-hockey robot trained entirely in a randomized simulation, letting it play competitively in the real world with no physical practice.

Luma AI opened the Open Physical AI Lab, an open-science program letting outside researchers and companies build on its world-model software.

Deep Robotics released a new demo of its upgraded DR02 humanoid that pushes the all‑weather industrial platform closer to real‑world deployment.

Hello Robot launched Stretch 4, a $30K home-assistance robot with a telescoping arm, an open-source stack, and pilot deployments for users with mobility impairments.

Australian robotics firm Luyten unveiled Ascend, a tower crane–based 3D concrete printer that uses AI‑driven construction to build multi‑story structures.

Boston Dynamics used human motion capture, sim-to-real reinforcement learning, and zero‑shot transfer to teach Atlas a Rabona-style soccer kick.

A smart autonomous robot nicknamed Raggy is about to begin final trials on Dorset farmland, where it uses computer vision to identify and remove toxic ragwort plants.

COMMUNITY

That's it for today!

Before you go we'd love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

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