Dynamic Creatures, a Boston-area startup that calls itself Boston Dynamics' official entertainment and hospitality partner, has shown off Hugo: a purple, furry, big-eyed humanoid built to greet people in theme parks, hotels and shops. The company says Hugo runs on the "first Vision-Language-Expression (VLE) model," which ties what the robot sees and hears to its gestures and facial expressions. The launch video came out on October 13, 2026.
The headline claim sounds like a leap in social robotics. The reporting suggests a more careful reading: the Boston Globe says Hugo is currently teleoperated by a human, and the autonomy is a goal. This post separates what the company announced, what independent outlets verified, and what nobody has shown yet.
TL;DR
| Question | Answer |
|---|---|
| What is it? | Hugo, a prototype character robot from Dynamic Creatures |
| Who is behind it? | Former leaders from Boston Dynamics, Google and the RAI Institute (Boston Dynamics); CEO Marc Theermann (Boston Globe) |
| Headline claim | A Vision-Language-Expression model links perception, language and expression in real time |
| Is it autonomous? | Not yet: the Boston Globe reports it is currently teleoperated |
| Hardware | 43 actuators (company claim); base platform unconfirmed, may resemble a modified Unitree G1 (Humanoids Daily's visual inference) |
| Customers | A large theme park and a retailer, unnamed (Boston Dynamics) |
| Price, funding, date | Not disclosed |
Who is Dynamic Creatures?
In its announcement from September 8, 2026, Boston Dynamics introduced Dynamic Creatures as a Boston-based company designing interactive character robots for entertainment and hospitality brands. It is led by former leaders from Boston Dynamics, Google and the Robotics and AI Institute. It is Boston Dynamics' official entertainment and hospitality partner, and it is already developing guest-facing character experiences with a large theme park and a retailer, though no rollout timeline or customer names are given. The page quotes Boston Dynamics' then-interim chief executive Amanda McMaster saying the company is combining "proven robotic platforms with character design."
Press coverage fills in one name. The Boston Globe, which ran a piece on October 13, identifies Marc Theermann, formerly Boston Dynamics' chief strategy officer, as CEO of the Cambridge-based startup.
What Hugo is, according to the company
Humanoids Daily describes Hugo as a purple, furry character with small horns, oversized hands and embroidered shorts. The company says Hugo pairs a robotic platform with 43 actuators, animatronic character design and expressive movement. The Hugo introduction video is on YouTube.
The software story is the Vision-Language-Expression model, which the company's reinforcement-learning team is said to be developing. The stated goal is to link perception and language with gestures, movement and emotional expression, so Hugo can interpret an encounter and respond to the person in front of it. The company also says it is working on navigation and safety capabilities for close interaction with people.
If you follow vision-language-action (VLA) models for robot arms, VLE is a close cousin with a different output: instead of only joint commands for a task, the model is meant to produce expressive behavior such as a head tilt, a wave or a facial reaction. We have covered the action side in Physical Intelligence's pi API beta and Reka's Rho-1 robot-action model.
What the Boston Globe adds: Hugo is teleoperated
The most important detail is not in the launch video description. The Boston Globe reports that Hugo is currently teleoperated, meaning a human controls it remotely. The company is developing AI that would let it learn to navigate new environments with little or no training, and software to answer people's questions in real time. The report says the system cannot yet be trusted with fully autonomous physical interaction with people, and that Theermann says it will improve over time. The video shows Hugo at the company's Boston headquarters high-fiving engineers and hugging children, according to the Globe.
Theermann's pitch, per the Globe, is to make people more comfortable with robots by making them "cute and cuddly," and he predicts that most people will have a robot at home within a decade.
Claimed vs. verified
| Claim | Status |
|---|---|
| Hugo exists and was shown in a launch video (October 13) | Verified by the video and Boston Globe, Humanoids Daily |
| Dynamic Creatures is Boston Dynamics' official hospitality partner | Stated in Boston Dynamics' own announcement |
| 43 actuators | Company claim only |
| A VLE model runs Hugo | Company claim; described as in development; no paper, benchmark or demo of autonomy |
| Real-time recognition of people and adaptation to room dynamics | Stated goal, not demonstrated independently |
| Autonomous operation around guests | Contradicted for now: Globe reports teleoperation |
| Modified Unitree G1 base | Visual inference by Humanoids Daily, unconfirmed |
| Theme park and retailer deployments | In development, unnamed, no date |
Note that the company's framing says the VLE model links vision, audio and motion in real time. The only independent report we found says the human operator is in the loop today. Both can be true if the model drives some expression while a person drives locomotion, but nobody outside the company has said so, and we found no technical documentation.
Why this matters
Character robots are a different bet from warehouse humanoids. The measure of success is whether a guest feels something, not whether a pallet is moved. That makes teleoperation a common bridge: operators drive the robot while data is collected to train autonomy later. The approach resembles what other embodied-AI firms do, and it is why demos should always be read with the question "who is driving?" We applied the same lens to Tesla's Digital Optimus Diablo demo and to the AA Robotics benchmark of frontier models on robot arms.
The safety question is the hardest part. A robot that hugs children needs force limits, fall behavior and fail-safes that a theme park insurer will accept. The company says it is developing "navigation and safety capabilities for close interaction," which is a roadmap item, not a certification. For a sense of how seriously humanoid safety can go wrong, see our report on Figure's F.02 decommission.
How to judge a character-robot demo
Because expressive robots are judged on feel, polished video can hide a lot. A few practical questions help when you watch the next one:
- Who is driving? Teleoperation, scripted animation and learned behavior look similar on camera. Ask for an unedited, operator-free run.
- How long is the continuous take? A single cut of several minutes with unplanned guests tells you more than a montage.
- What happens on an error? Watch for how the robot recovers when someone steps into its path or speaks over it.
- Is latency shown? Conversational expression depends on response time. A reaction that arrives a second late feels uncanny, whatever the model behind it.
- Are failures published? Real autonomy programs report intervention rates, such as how often an operator had to step in per hour.
Teleoperation is not a flaw in itself. Operators generate the demonstration data that later trains an autonomous policy, and many robotics programs start this way. The honest framing is "prototype driven by a person today, autonomy in development," which is what the independent reporting supports.
Until Dynamic Creatures publishes measurements of this kind, Hugo is best understood as a design and branding statement backed by serious robotics pedigree, not yet as evidence that a general expression model exists.
What this means for builders and buyers
For a theme park or retailer evaluating a character robot, the practical takeaway is to separate three layers: the hardware platform, the character design, and the autonomy software. Dynamic Creatures says it combines existing robotic platforms with character design, so the differentiator is the third layer, and that is the layer with the least public evidence today. A buyer should ask for the operator-to-robot ratio, because a robot that needs one human per unit has very different economics from one that runs unattended.
For researchers, the VLE label is a useful prompt rather than a result. Until there is a paper, a benchmark or an unscripted demo, it is fair to treat Vision-Language-Expression as a product direction: a model whose outputs include expression, not only motion.
What to watch next
- Autonomy evidence. A live, unscripted demo with no operator, or a technical report on the VLE model with measured latency and failure rates.
- Named customers. The theme park and retailer mentioned by Boston Dynamics.
- Hardware identification. Whether the base is a third-party humanoid or Boston Dynamics hardware.
- Safety standards. Any certification or third-party review for contact with guests.
- Funding and pricing. Neither has been disclosed.
What people are asking
Is Hugo a companion robot for the home? The Globe quotes the CEO predicting home robots within ten years, but the stated near-term market is theme parks, hotels and retail. Today's Hugo is a prototype.
Is this Boston Dynamics' product? No. Boston Dynamics calls Dynamic Creatures its official partner, and the Globe says the startup uses much of its technology. It is a separate company.
How does this compare with frontier-lab robotics? It is an expression-first product. For the broader debate about how far robot intelligence really is, our summary of Yann LeCun's view of human-level AI and robots is a useful sceptic's counterweight.
Related reading
- Physical Intelligence pi API beta
- Reka Rho-1: text, video and robot actions in one network
- Tesla Digital Optimus Diablo claim, checked
- AA Robotics v0.1 benchmark, claimed vs. verified
- Figure F.02 decommission
- Yann LeCun on human-level AI and robots
- Gemini 4 Argon for robotics
- Sources: Boston Dynamics announcement, Humanoids Daily, Boston Globe
Details are accurate as of October 16, 2026 and come from the cited reports; the company had published no technical paper at that date.