Signals InboxJuly 16, 2026Humanoid Robotics
Humanoids: what surgical tasks can they do today?
Humanoids can manipulate instruments, retract tissue and hold an endoscope during realistic surgical procedures today. But every meaningful task still depends on direct human control, and no humanoid has operated on a human patient or completed an autonomous surgical step inside a living subject.
We track humanoid robotics daily. Want the market signals in your inbox?
Send me the signals →Humanoids can perform real surgical tasks today, but only as teleoperated tools. Their strongest results are two live porcine gallbladder procedures and a cadaveric sinus operation in which Unitree G1 robots manipulated standard instruments under continuous human control.
The evidence looks broader than it really is. Eleven of the 16 highest-ranked tasks come from those experiments, but most reduce to three physical abilities: moving an instrument, manipulating visible tissue and maintaining a useful camera view.
The sharp divide appears around autonomy. Humanoids can reproduce a surgeon’s movements inside the body, but they cannot independently identify anatomy, select an intervention, manage bleeding or change the surgical plan.
The best argument for a surgical humanoid is compatibility. It can stand beside an ordinary operating table and use tools designed for human hands, potentially allowing one platform to assist in surgery and perform other hospital work.
For now, though, the surgeon remains the intelligence. The humanoid is an unusually flexible pair of remote-controlled arms.
Interested in humanoid robotics?We can send you all the signals
Send me the signals → Delivered straight to your inboxHumanoid surgical tasks ranked by current capability
| # | Surgical task | Capability score | Evidence stage | Control mode | What has actually been proven |
|---|---|---|---|---|---|
| 1 | Manipulating laparoscopic instruments during live abdominal surgery | 72 | Preclinical | Teleoperated | Unitree G1 humanoids completed two porcine cholecystectomies using general-purpose laparoscopic instruments. Surgeons controlled every movement remotely. |
| 2 | Moving a laparoscopic grasper through a fixed abdominal port | 70 | Preclinical | Teleoperated | The LapSurgie system maintained the required remote center of motion while moving instruments inside living pigs. |
| 3 | Grasping visible tissue with a laparoscopic instrument | 69 | Preclinical | Teleoperated | Humanoid-held instruments grasped and manipulated tissue during live gallbladder procedures. The surgeon selected every grasp. |
| 4 | Retracting tissue to expose the operative field | 68 | Preclinical | Teleoperated | The live procedures included assistant-style tissue manipulation and exposure. Autonomous force selection was not tested. |
| 5 | Maintaining endoscopic visualization throughout a complete procedure | 67 | Preclinical | Teleoperated | A seated Unitree G1 held and repositioned a rigid endoscope throughout a cadaveric sphenoidectomy. A trained researcher controlled it continuously. |
| 6 | Positioning an endoscope at a surgeon-requested viewing angle | 66 | Preclinical | Teleoperated | The humanoid changed the endoscope’s position and orientation through a Meta Quest 3 control interface while coordinating verbally with the surgeon. |
| 7 | Holding a surgical instrument steadily for extended periods | 64 | Preclinical | Teleoperated | Stable camera holding was maintained throughout the cadaveric procedure, supporting a realistic first-assistant role. |
| 8 | Operating two instruments during a laparoscopic workflow | 63 | Preclinical | Teleoperated | The humanoid framework supports bimanual instrument control, and the live study tested configurations involving two humanoid platforms. |
| 9 | Following a surgeon’s hand movements through remote control | 62 | Preclinical | Teleoperated | Operator commands were mapped into robot wrist and instrument-tip movements while respecting laparoscopic port constraints. |
| 10 | Working beside a human surgeon at a standard operating table | 60 | Preclinical | Teleoperated | Humanoids were positioned around conventional surgical tables and used standard handheld instruments rather than a dedicated surgical console robot. |
| 11 | Working beside another humanoid during the same operation | 58 | Preclinical | Teleoperated | One porcine procedure tested a multi-humanoid setup. This proved physical coordination, not autonomous teamwork. |
| 12 | Completing standard laparoscopic training-board exercises | 56 | Repeated | Teleoperated | Users with different surgical experience levels completed dry-lab tasks through the humanoid system and were compared with manual and specialized robotic approaches. |
| 13 | Transferring small objects between laparoscopic graspers | 54 | Repeated | Teleoperated | Object-transfer exercises were completed during controlled laparoscopic user studies. Living tissue and bleeding were absent. |
| 14 | Placing an object accurately at a visible target | 53 | Repeated | Teleoperated | Benchtop testing quantified instrument accuracy and workspace reach using the G1 platform. |
| 15 | Rotating a wristed laparoscopic instrument inside the body | 52 | Repeated | Teleoperated | LapSurgie mapped operator hand and finger movements into commercially available wristed instrument control. |
| 16 | Reaching surgical targets at different depths and angles | 50 | Repeated | Teleoperated | Workspace testing showed that trocar placement strongly affects which internal targets a humanoid can reach safely. |
| 17 | Picking up a known surgical instrument from a fixed holder | 45 | Demonstrated | Teleoperated | Humanoids can grasp and control mounted medical tools, but complete sterile tray handling has not been proved in surgery. |
| 18 | Returning an instrument to a predefined holder | 44 | Demonstrated | Teleoperated | The required rigid-object placement is within demonstrated humanoid manipulation, but no complete sterile workflow has been published. |
| 19 | Exchanging wrist-mounted surgical instruments between tasks | 42 | Demonstrated | Human-assisted | The live-surgery framework included a custom tool-swapping attachment. Autonomous sterile exchange was not demonstrated. |
| 20 | Applying controlled pressure with a handheld medical tool | 40 | Demonstrated | Teleoperated | Medical-task studies used impedance control for safer tool contact, but reported force and sensor limitations. |
| 21 | Identifying a predefined anatomical target from video | 34 | Demonstrated | Human interpreted | Humanoids transmitted surgical video to operators, but humans still identified the anatomy. Specialized robots have demonstrated autonomous recognition in ex vivo procedures. |
| 22 | Cutting tissue along a surgeon-selected path | 32 | Demonstrated | Teleoperated | A humanoid can physically move a cutting instrument, but no published humanoid study has isolated and validated cutting performance autonomously. |
| 23 | Applying a surgical clip to an exposed structure | 29 | Claimed | Not proved | Specialized SRT-H placed clips autonomously on ex vivo porcine tissue. No humanoid has publicly reproduced that result. |
| 24 | Severing a clipped duct with laparoscopic scissors | 27 | Claimed | Not proved | SRT-H completed this step autonomously ex vivo, but humanoid evidence remains limited to teleoperated instrument manipulation. |
| 25 | Adjusting autonomously after tissue appearance changes | 25 | Claimed | Not proved | SRT-H handled altered starting positions and blood-like visual changes. No humanoid has shown equivalent autonomous recovery. |
| 26 | Manipulating and reorienting a surgical needle | 23 | Claimed | Not proved | Specialized surgical robots have demonstrated needle manipulation. Humanoids have not shown it in a surgical procedure. |
| 27 | Driving a curved needle through soft tissue | 20 | Claimed | Not proved | This requires much finer force and pose control than current humanoid surgical studies have demonstrated. |
| 28 | Placing a sequence of evenly spaced sutures | 18 | Claimed | Not proved | STAR performed autonomous intestinal suturing on living pigs, but it was a purpose-built surgical robot using a controlled plan and marked tissue. |
| 29 | Tying a secure knot after placing a suture | 16 | Claimed | Not proved | No humanoid has demonstrated a complete needle pickup, insertion and knot-tying sequence in surgery. |
| 30 | Controlling unexpected bleeding during an operation | 13 | Claimed | Not proved | The robot would need to locate the source, choose an intervention and act quickly under changing visual conditions. |
| 31 | Selecting the correct instrument as the procedure changes | 11 | Claimed | Not proved | Current humanoids execute operator commands. They do not independently understand the surgical stage and select tools. |
| 32 | Diagnosing an unexpected surgical complication | 8 | Claimed | Not proved | No humanoid has combined anatomy, patient data and intraoperative events to make an independent surgical diagnosis. |
| 33 | Changing the surgical plan after discovering abnormal anatomy | 6 | Claimed | Not proved | Current humanoid studies transfer human movements rather than human surgical judgment. |
| 34 | Completing an entire operation autonomously on a living patient | 3 | Claimed | Not proved | No humanoid has autonomously operated on a living animal or human. The live humanoid procedures were fully teleoperated. |
| 35 | Operating independently across several types of human surgery | 1 | Claimed | Not proved | No humanoid has entered human clinical surgery, even under continuous teleoperation. |
Interested in humanoid robotics?We can send you all the signals
Send me the signals → Delivered straight to your inboxFigure AI launches Index with 16 million videos already uploaded
Chinese robot cut its 100m record from 9.39 to 8.86 seconds
XPeng’s robot unit raises $900M before selling a single robot
Unitree’s robot IPO jumps 460% in Shanghai
Unitree CEO sees robots hitting "a ChatGPT moment" by 2028
Unitree’s Superman robot hits 12.66 m/s, faster than Bolt’s peak
FCC blocks Unitree’s next robots, opening America’s humanoid race
San Francisco residents can now hire a humanoid cleaner for $30
Google DeepMind gives robots a Gemini brain that acts
Musk: it will be "America’s fastest industrial scale-up since WWII"
BotQ just built Figure’s 1,000th F.03 humanoid robot\
Tesla is building its first Optimus lines
Q1Are humanoid robots actually performing surgery today?
Humanoid robots are performing surgical actions today, but surgeons still perform the surgery through them. The strongest study involved two live porcine gallbladder removals, with every humanoid movement controlled remotely by a human. A separate G1 held an endoscope during cadaveric sinus surgery.
Eleven of our top 16 tasks come from these two experiments, but they reduce to three core abilities: move an instrument, hold a view and manipulate tissue. That is meaningful physical progress, not broad surgical competence.
Humanoids have crossed the line into preclinical surgery. They have not crossed into human clinical use, autonomous surgery or independent surgical decision-making.
Q2Are surgical humanoids autonomous or still teleoperated?
Surgical humanoids are still teleoperated. All 20 tasks in our ranking with direct humanoid evidence rely on a human operator, human interpretation or human-assisted setup.
There is a large gap between physical and cognitive capability. The humanoid can reproduce instrument movements inside the body, but the surgeon still recognizes anatomy, chooses the target, controls force and decides what happens next. Even the cadaveric endoscope task required continuous control by a trained researcher.
Specialized robots are further ahead. SRT-H autonomously completed a 17-step gallbladder phase on ex vivo tissue, while STAR performed autonomous intestinal anastomosis on living pigs under controlled conditions. Humanoids have not yet autonomously completed one surgical step in vivo.
Q3Why use a humanoid when surgical robots already exist?
A humanoid’s advantage is compatibility. It can stand beside a normal table, use handheld instruments and fit inside rooms designed for people. Dedicated systems offer greater rigidity, precision and maturity but require specialized equipment, setup and operating-room workflows.
The live porcine study showed that this compatibility is real: commercially available humanoids were adapted to laparoscopic surgery without becoming purpose-built surgical machines. But the researchers also identified precision, control and safety gaps before clinical deployment.
The humanoid case depends on versatility. A robot used only for surgery will struggle to outperform a specialized surgical platform. A robot that can assist during surgery and then perform other hospital work has a more believable reason to exist.
We track humanoid robotics daily. Want the market signals in your inbox?
Send me the signals →Q4Which surgical tasks are humanoids still nowhere near doing?
Humanoids remain furthest from tasks that require them to decide rather than merely move. The bottom 15 tasks involve autonomous anatomy recognition, needle control, suturing, bleeding management, complication diagnosis or changes to the surgical plan.
The gap is visible in our ranking. The top 20 tasks are mostly positioning and teleoperation problems. Every task below them adds independent perception, force judgment or decision-making. None has been demonstrated by a humanoid in a realistic surgical procedure.
Holding surgical tools is no longer the main question. Understanding what is happening, recognizing when something has gone wrong and safely choosing an irreversible action is the wall.
We broke humanoid surgical capability into 35 concrete tasks and assessed each one across five dimensions: successful execution, precision, repeatability, autonomy and clinical realism. The ranking measures what humanoids themselves have proved, not what da Vinci or other specialized surgical robots can do.
Demonstrated means a task was completed in a controlled experiment. Repeated means it was reproduced across trials or operators. Preclinical means it was performed on a cadaver or living animal. Clinical requires use during human care, while scaled requires routine use across medical institutions.
We prioritized direct experiments over company claims and general predictions. The evidence includes benchtop tests, repeated laparoscopic training exercises, cadaver procedures and live-animal surgery.
Several high-ranked tasks come from the same procedures because one operation tests multiple abilities. We counted those abilities separately, but we did not treat them as independent proof of broad surgical competence. This is why the analysis also groups the results into underlying capabilities such as instrument movement, tissue manipulation and camera positioning.
Specialized surgical robots such as STAR and SRT-H are used only as comparison points. They show what autonomous recognition, suturing, adaptation and multi-step execution can look like under controlled conditions, but their results are never counted as evidence that a humanoid can perform the same task.
The final scores reflect the evidence available today rather than a prediction of which capabilities will arrive next. Tasks rank higher when they have been completed under more realistic conditions, reproduced more consistently and performed with less human assistance.
Key sources used for this analysis include: the study of in vivo humanoid laparoscopic surgery and clinical readiness, the LapSurgie humanoid teleoperation framework, the study of a humanoid first assistant during cadaveric endoscopic surgery, research on Unitree G1 teleoperation across medical procedures, the SRT-H autonomous cholecystectomy study, the STAR autonomous intestinal anastomosis study, the related Science Robotics publication, the FDA overview of robotically assisted surgical devices, and Unitree’s technical information for the G1 platform.
Building or investing in humanoid robotics?We can send you all the signals
Send me the signals → Delivered straight to your inbox