Musk proposes robot test scenarios: a robot looks down and assembles a Swiss watch with one hand
On October 27, 2026, Elon Musk posted on X that these are the real benchmarks, and then proposed two robot test scenarios: a robot looking down and assembling a Swiss watch with one hand, while the other hand, without looking up, catches a baseball flying diagonally at 100 miles per hour; and juggling burning chainsaws on a unicycle.
The related discussion came in the context of a reality check on robot capabilities. A previous post pointed out that no robot can currently complete even simple tasks such as replacing a jack-o'-lantern battery. Musk used extreme multitasking, divided attention, high-speed reactions and balance requirements as examples of a higher threshold, without specifying a specific model or timeline.
Musk has long emphasized that dexterity in humanoid robot hands is a core challenge. In a previous interview, he said that the human hand has about 27-28 degrees of freedom and can perform various actions such as swinging a bat, threading a needle and playing the piano. The design goal for Optimus Version 3 is to approach the dexterity of a human hand, have an AI brain capable of real-world exploration, and achieve mass production. This post did not provide any new hardware or progress data.
Original sources:
Title: How about the robot is looking down while assembling a Swiss watch... Publisher: Elon Musk (@elonmusk) Publication date: 2026-10-27 URL: https://x.com/elonmusk/status/2114945910648225841 Source nature: original post by the person involved Verification status: the full post and context were actually read for this report, verification date 2026-10-26 Supporting content: assembling a Swiss watch with one hand while looking down and, without looking up, catching a baseball flying diagonally at 100 miles per hour; juggling burning chainsaws on a unicycle.
Title: These are real benchmarks Publisher: Elon Musk (@elonmusk) Publication date: 2026-10-27 URL: https://x.com/elonmusk/status/2114945014023536935 Source nature: original post by the person involved Verification status: actually read for this report, verification date 2026-10-26 Supporting content: in response to discussion of related robot capabilities, he said that these are the real benchmarks.
Title: Exclusive: Original Text of Musk's Latest Interview - Tesla FSD, Humanoid Robots... Publisher: 36Kr (interview paraphrase) Publication date: about 2025-09 (interview content) URL: https://eu.36kr.com/en/p/3468992506173571 Source nature: media report Verification status: the relevant passage was actually read for this report, verification date 2026-10-26 Supporting content: Optimus Version 3 needs to solve hand dexterity, the AI brain and mass production; description of the degrees of freedom and versatility of the human hand. Does not prove that the specific benchmarks in this post have already been achieved.
ABAB AI Insight
This post by Musk shifts the discussion of robot capabilities from being able to complete a single task to a composite scenario involving separated visual attention, fine manipulation and high-speed dynamic reaction at the same time. Assembling a watch requires millimeter-level positioning and force control; catching a ball flying diagonally at 100 miles per hour requires millisecond-level prediction and grasping. The two must be done by different arms within the same time window, and the visual gaze must not be lifted. Juggling burning chainsaws on a unicycle adds balance, multi-object tracking and heat-source safety constraints. These are not capabilities that have already been demonstrated; they are higher-threshold examples he publicly set.
On the business path, Tesla Optimus has long focused publicly on hand degrees of freedom, custom actuators and factory data loops. In interviews, Musk has repeatedly emphasized that the human hand is the hardest mechatronic challenge, with about 27-28 degrees of freedom, and that the Version 3 goal is to approach human-hand dexterity and have a design that can be mass-produced. Current public demonstrations are still mainly simple factory handling tasks, and there remains a clear gap to simultaneously completing fine assembly and high-speed catching. The post itself does not change the hardware timeline, but it sets an extreme standard against which future demonstrations can be publicly measured.
Compared with other humanoid robot approaches, some companies focus on single-task or controlled-environment demonstrations, while Musk uses composite extreme scenarios to compress the latency and reliability requirements for perception, decision-making and execution. The similarity is that all of them need to solve hand dexterity and balance; the difference is that he raises the public benchmark directly to cross-modal, multi-task parallelism rather than step-by-step single-task validation. The focus for capital shifts from “can it walk and pick things up” to “can it maintain continuous high reliability in a real dynamic environment.”
What this statement changes is the external calibration of expectations that “robots are already close to practical use.” It shifts the discussion from “whether there is a robot that can replace a battery” to “whether it can simultaneously complete precision and high-speed tasks under divided attention,” so that any subsequent demonstration must be measured against these composite requirements. The observation metric is whether the next public demonstration includes multi-task parallelism or high-speed dynamic grasping, and whether the Version 3 hand degrees of freedom and actuators reach publicly stated near-human benchmarks.
ABAB News · Cognitive Law
The real benchmark has been upgraded from a single task to parallel precision operation and high-speed dynamic reaction under divided attention.
Hand dexterity remains the core mechatronic constraint, and extreme scenarios expose the gap between current demonstrations and the goal.
Public extreme benchmarks compress the expectation of “usable,” forcing subsequent demonstrations to be measured against composite reliability rather than isolated success.