PRESENTED BY
The pictured robot is Fourier’s GR-3, not the reported Max prototype. AI-assisted editorial collage; source imagery: Fourier / PR Newswire and Meta.
Meta’s humanoid ambitions now have a reported hardware partner and a prototype name: Fourier and Max. The Information describes a custom robot built in three months, with a successor already underway. Meta says the work is for internal research.
Elsewhere, the industry is getting more specific about what robots should do: Audi is testing parts sequencing, Xiaomi reports an eight-hour sorting task, and Agility is one step from its expected public-market debut. We also spoke with Lightberry’s CEO about Lumi’s conditional FCC exemption and what still comes before sales.
IN THIS EDITION
Meta and Fourier: what the reported Max collaboration tells us.
Money and milestones: PI’s funding talks, Unitree’s results and Agility’s merger vote.
Robot intelligence and work: Argon + Genie 4, Audi, Xiaomi and Reward AI’s demo.
Also: our Lightberry interview, Tesla’s December rumors, AGIBOT’s Worken proposal, batteries and robots beside a marathon route.
LEAD STORY
The Information reports that Meta and Fourier developed Max in three months, demonstrated it internally in August and began work on Max 2 in September. The custom prototype reportedly stands about 1.5 meters tall, weighs 48 kg and uses Jetson Thor.
The design brief puts interaction at the center: recognizing voices, interpreting gestures, maintaining eye contact and showing when the robot is listening, thinking or uncertain. Fourier contributes hardware experience from multiple robot generations; Max is described as a customized platform, rather than a renamed GR-3.
Meta says the collaboration is for internal research and experimentation and that it has no plans to sell these units. The news is a clearer picture of the physical machines behind its robotics effort—not a consumer launch announcement.
ROBOT INTELLIGENCE
DeepMind paired Gemini 4 Argon with Genie 4 in a demonstration depicting Apollo playing checkers. Argon breaks a user’s task into steps and sends text instructions to Genie as it generates the scene, allowing the imagined scenario to develop as the task progresses.
For robotics, the promise is more varied, steerable environments for training and evaluation. This is a simulation demonstration, distinct from the physical Apollo cornhole trial we covered earlier. Genie 4 is entering research preview; physical accuracy and consistency over longer sequences remain challenges.
CAPITAL & COMPANY RESULTS
Bloomberg reports that Physical Intelligence is in funding talks at a $20 billion valuation, with existing backers Avenir and Lux discussing leading the round. The financing has not been announced as completed.
The talks follow PI’s API beta, which lets developers fine-tune models and use hosted inference. Its broader bet is a shared intelligence layer that learns across robot bodies and applications. The API gives that idea a practical route to outside developers; the reported valuation reflects expectations for where it could lead.
Unitree reported 665 million yuan in Q3 revenue, up roughly 70% year on year. Attributable net profit fell 17.4% to 112.7 million yuan, while R&D spending rose 249.6% to 127 million yuan—about 19% of revenue.
That is substantial growth alongside a bigger investment in models, compute and engineering. The revenue covers Unitree’s wider business, including quadrupeds. Its much larger nine-month profit increase also benefits from a one-time expense in the prior-year comparison.
Churchill XI shareholders approved the merger on November 2, with more than 84% of votes cast in favor. Closing is expected November 4, followed by Nasdaq trading as AGLT on November 5, subject to remaining conditions.
The approximately $465 million gross capital total includes $164 million in trust cash, a $201 million Foxconn-led private investment and $100 million Agility has already received. It is not all fresh cash arriving at closing. Our financing breakdown explains why the total differs from the earlier $620M-plus estimate.
PRESENTED BY ROBOSTRATEGY
Most of the companies we usually cover in this newsletter are still private, which puts them out of reach for public-market investors. RoboStrategy, Inc. ($BOT) is built to give you access to several of them.
Investing involves risk, including possible loss of principal. Read the prospectus before investing.
FACTORY WORK
Agile Robots’ mobile humanoid. Image: Audi press release.
Audi’s proof of concept with Agile Robots targets picking components from containers and arranging them for assembly. Dassault Systèmes and NVIDIA support virtual testing of the workflow. Audi’s existing material supply involves around a million picks daily; that describes the opportunity, not robot throughput.
Meanwhile, Xiaomi says its humanoid completed an eight-hour side-cover sorting task in its auto factory, with anomaly detection and correction. The video shows parts moving between containers and racks, but gives no throughput or intervention figures.
These projects make the job concrete: identify the right part, handle it without damage and put it in the correct place repeatedly. The next useful comparison is output across a shift, rather than duration alone.
DEMO TO WATCH
Reward AI’s OM-1 delivery demonstration. Screenshot: Reward AI.
Reward AI’s latest video moves from delivery to picking up packages and socks on stairs, mopping and opening a refrigerator. The company presents the footage as fully autonomous and running at normal speed. OM-1 learns manipulation from human demonstrations captured with its wearable Omnibody Hand, while a separate controller trained in simulation translates those actions to the robot. It’s an engaging look at handling very different objects and contact conditions.
OUR INTERVIEW
Lightberry’s Lumi has received a Covered List exemption, tied to its onshoring plan and continued product vetting. CEO Ali Attar told us independent FCC certification remains the next step before off-the-shelf sales.
Lightberry says it replaces Unitree’s sensors, computer and software, and plans U.S. manufacturing beginning with the body, head and hands. Attar expects shipping “very soon,” without giving a date. Our interview separates the exemption from certification and explores the manufacturing commitments behind it.
TESLA WATCH
Humanoids Summit organizer Modar Alaoui told The Standard that Tesla said reservations would open around the December 1–2 event or shortly afterward. The publication also reports plans for an Optimus V3 demonstration. Tesla did not respond to its request for comment, and reservation fees and consumer delivery terms remain unclear.
Separately, Robert Scoble’s account of elaborate Optimus data-collection suits prompted us to examine Tesla’s own job listings. They confirm motion-capture work and multi-shift collection operations, but do not establish his claimed 40-camera setup. It’s a useful look at the human work behind robot learning.
FROM OUR SPONSORED AGIBOT COVERAGE
Our AGIBOT Innovation Day coverage was sponsored by AGIBOT.
AGIBOT’s proposed Worken measure combines “Work” and “Token” into “Effective Intelligent Work Hours.” Its formula multiplies active robots × autonomous working hours × output efficiency relative to a task benchmark.
For example, 10 robots working autonomously for six hours at an efficiency ratio of 0.8 produce 48 Worken. That is an illustrative calculation, not a deployment result. The useful idea is to measure productive output after installation; the difficult part is agreeing what work qualifies and how to compare it. Worken remains a proposed framework, not an industry standard or a universal price for robot labor.
THE RUNDOWN
Bleu Robotics raises €5M. The startup aims to let operators teach changing industrial tasks through human demonstrations. Its claimed workflow is roughly 20 minutes collecting examples and another 20 before execution. The funding supports turning the research into a product for industrial customers.
LG looks toward home humanoids around 2030. It plans to build experience in research, public-sector and industrial markets first. One interesting advantage: LG can design appliances for robots as well as robots for appliances. Its data factory has begun partial operations.
A projected $850M humanoid battery market by 2037. IDTechEx’s forecast includes spare packs for swapping alongside batteries for new robots. Our report examines why charging, replacement packs and useful working time matter as much as headline runtime.
XPENG says IRON’s battery is in small-batch production. The exact chemistry remains unclear: XPENG promotes all-solid-state technology, while earlier reporting described a semi-solid alternative. Numerical performance results and production volumes were not disclosed.
AND FINALLY…
Humanoid traffic-support robots at the Hangzhou Marathon. Screenshot: DEEP Robotics.
DEEP Robotics’ Hangxing No.2 humanoids appeared beside the route of the November 1 Hangzhou Marathon, which drew 36,000 runners. Based on DR02, they supported roadside operations with traffic-control gestures, safety announcements and public information.
DR02’s outdoor design includes advertised IP66 protection and a −20°C to 55°C operating range. The footage offers a different setting from the factory and living room: robots working alongside officers at a crowded public event.
This edition of Humanoids Daily was presented by RoboStrategy.
Seeing something before I do? Just hit reply — that's where a lot of these stories start.
P.A
Humanoids Daily
Know someone following the robotics industry? Forward them this briefing—it helps Humanoids Daily reach more people.
Received this from a friend? Subscribe to the free weekly briefing.