Four meters tall. 1,200 kilograms. 500 kilograms of payload per arm. 41 joints. Hydraulic actuation.
Read the specification twice. The second pass is where the work happens, because the second pass is where you look for what the document refuses to say. There is no perception stack. No vision-language-action model. No autonomy claim. No navigation spec. No safety certification. No unit cost. No delivery date. No named customer.
A machine with 41 joints and no eyes is not a humanoid robot. It is a hydraulic crane wearing a costume.
Here is the part that belongs in a crypto publication. The launch of Neo1, the product of a firm calling itself NeodraDynamics, did not run through ICRA, CoRL, or NeurIPS. It ran through the China International Industry Expo, then a prediction-market post, then a crypto news outlet. That routing is the story. The hardware is the decoration. When a physical claim is distributed through speculation infrastructure rather than engineering infrastructure, the thing being sold is not the machine. It is the contract on the machine.
Context: a press release, a joke, and a settlement mechanism
Neo1's public record is thin and single-sourced. Every number above traces to one company release: four meters, 1.2 tonnes, 500 kilograms per arm, 41 hydraulic joints, "small hydraulic actuators" carrying the majority of the joint set. No independent teardown. No third-party measurement. No disclosed supplier list. The company name does not surface in any database I can query.
Set beside it, Tesla's Optimus. No production figures. No order book. No unit price. No delivery schedule. What the coverage offers instead is a joke from Elon Musk about Optimus building Optimus Prime, a Wedbush analyst predicting 2027 will be strong on the back of robotaxi and Optimus bundled into a single sentence, a Mark Cuban counter-prediction that the category fails inside a decade, a Unitree founder's estimate of a "ChatGPT moment" landing somewhere between two and ten years, and a passing mention of a flying Roadster that has no business in a robotics article.
Code does not lie, but it often omits the context. A press release is code with the error handling deleted. It states the happy path and nothing else.
I have spent years auditing exactly this class of document. In 2020 I reverse-engineered the price feed mechanisms of five lending protocols and found the same structural flaw in three of them: the median was fine, the source was not. Nobody had asked who reported the number, how stale it could get, and what happened when the reporter held a position. That is the correct lens for this story, and it is not a robotics lens.
Prediction markets are oracles with a user interface. A contract on "which humanoid robot reaches commercial deployment first" is a claim about the physical world wrapped in a settlement mechanism. Its integrity is bounded above by the integrity of its resolution source. If the resolution source is a press release from a company with a financial interest in the outcome, the market is not pricing robot capability. It is pricing that company's willingness to keep publishing.
Core: what the spec sheet specifies, and what it cannot
The actuator choice is a tell
Every serious humanoid program has moved to electric actuation. Optimus, Unitree, Figure, the new Atlas — all electric. Hydraulics convert energy at roughly 50 to 70 percent efficiency; electric drivetrains clear 90 percent. Hydraulics bring high-pressure lines, leak risk, fluid maintenance intervals, and response lag. The industry did not abandon hydraulics out of fashion. It abandoned them because the maintenance cost curve and the control bandwidth both lose.
Neo1 goes the other way, and the stated reason is load. 500 kilograms per arm. That number is real in heavy machinery and unremarkable there. KUKA and FANUC heavy-payload arms have handled tonne-class loads for years, bolted to a floor, with known kinematics and a service contract. The novelty claim is that this payload now sits inside a humanoid envelope.
Ask what the envelope buys. For a fixed work cell, nothing. A humanoid shell on a fixed task is a cost multiplier with no functional return — more joints to calibrate, more failure modes, worse stiffness, harder certification. For a mobile task, everything depends on the mobility system, and the specification is silent on whether Neo1 walks. Humanoid, in the release, appears to mean "tall and bipedal-looking." It does not appear to mean "capable of dynamic bipedal balance." Those are different claims and only one of them is testable.
The joint count is another tell, pointing the other way. Joint count in humanoids exists to serve dexterity — Optimus-class platforms run in the high twenties to low forties for fine manipulation. A four-meter machine doing port loading or mining does not need finger-grade articulation. It needs stiffness, payload margin, and stability under shock load. Forty-one hydraulic joints on a four-meter frame is not a capability statement. It is a number chosen because it sounds like a capability statement.
Then the absence. No perception. No planning. No decision layer. Without a described autonomy stack, the honest classification is teleoperated or preprogrammed heavy equipment. That is a legitimate product category. It is not the category the word "humanoid" invokes, and it is not the category the market is pricing.
| Parameter | Stated | Independently verifiable | |---|---|---| | Height / mass | 4 m / 1.2 t | No | | Payload | 500 kg per arm | No | | Joints | 41, mostly hydraulic | No | | Locomotion | Not specified | No | | Perception / VLA | Not specified | No | | Autonomy level | Not specified | No | | Unit cost / lead time | Not specified | No | | Safety certification | Not specified | No |
Eight rows. Two of them describe a machine. Six describe an absence.
The commercialization paradox
The stated target scenarios — port handling, mining, disaster response — already have mature automation. Ports run automated quay cranes and AGV fleets. Mining runs autonomous haul trucks and remote loaders. Disaster response runs heavy plant with operators who are already trained. Every one of those solutions beats a 1.2-tonne hydraulic humanoid on unit cost, and unit cost is the only procurement metric that matters in those verticals.
So the buyer does not exist at the price the humanoid shell implies, and the shell does not survive the price the buyer will pay. That is the paradox in one line: the capability is not enough to justify a humanoid premium, and the humanoid form is enough to destroy the cost case.
The Optimus side has a different problem, which is that it has no numbers at all. Tesla's valuation currently rests on four components: an auto business in a price war with decelerating growth, an energy business that grows but stays small, a robotaxi that has not shipped, and an Optimus that has not shipped. The analyst quoted in the coverage bundles robotaxi and Optimus into a single 2027 thesis. That bundling is the tell. Two unshipped programs priced as one correlated bet means either failure invalidates the whole position, and the sentence is constructed so the optimism never has to be decomposed.
The bull and bear cases are not symmetric in quality. Cuban's ten-year failure call is a cost-benefit argument: does generality justify the hardware cost. The 2027 call is a timing argument with no unit economics attached. One can be falsified by a spreadsheet. The other can only be falsified by a calendar.
The resolution layer is the actual product
This is where the crypto reader should stop skimming, because the resolution problem is ours. We built it.
Take a plausible contract: humanoid_robot_first_commercial_deployment. Write the settlement spec the way an auditor would have to write it.
