The First Full-Body VR Company Should Start as a Gym

The recurring dream

Virtual reality does not have one origin story. It has a sequence of returns.

Each generation rediscovers the same ambition: replace the rectangular screen with a world that surrounds the user. Each generation also arrives with a new technical bottleneck. Early systems lacked compact displays and real-time graphics. The 1990s produced commercial arcade machines such as Virtuality, but the headsets were heavy, the graphics coarse and the installations expensive.4 The 2012 Oculus Rift campaign revived consumer VR by showing that low-latency head tracking and commodity smartphone displays could produce presence at a price developers could reach; 9,522 backers pledged about $2.44 million.5

Modern VR is not a replay of 1991 with prettier graphics. A Quest 3 provides a standalone computer, six-degree-of-freedom head and hand tracking, color stereoscopic passthrough, a depth sensor, 2,064 by 2,208 pixels per eye and refresh rates up to 120 Hz.7 Apple’s current Vision Pro specification lists 23 million display pixels, micro-OLED panels, multiple world- and eye-tracking cameras, LiDAR, four inertial measurement units and 12-millisecond photon-to-photon latency through its R1 chip.6

These are remarkable machines. They are also evidence that display quality alone does not complete the medium.

High-resolution, low-latency, room-tracked computing now fits inside a compact consumer appliance. Yet VR has not become the default surface for work, social life or games. It has become a meaningful category—large enough to sustain platforms, developers, fitness applications and location-based venues—but not the universal computing transition repeatedly promised.

That mismatch should change the question. Instead of asking how many more pixels are needed, we should ask what kind of experience remains impossible even after the pixels become good.

The headset works

“VR never took off” is too simple. The category sustains users, developers, fitness applications, simulators, social worlds and location-based venues. The platform exists.

But a platform can be technically valid and economically incomplete.

Meta’s Reality Labs segment illustrates the distinction. From 2021 through 2025, annual segment revenue remained near $2 billion while annual operating losses grew from $10.2 billion to $19.2 billion. Those figures include augmented- and virtual-reality hardware, software and content, not just headsets, so they are not a clean measure of consumer VR profitability. They do show how much capital has been required to keep pushing an ecosystem whose commercial scale still lags its technical ambition.891011

There are several reasons VR remains bounded. Headsets still create friction. Some users experience cybersickness, with sensory integration and locomotion mismatch among the important mechanisms.30 Wearing a device over the face changes social behavior. Content is expensive to produce. A user must clear space, isolate from the room and accept an interface that often feels more ceremonial than opening a laptop or television.

But there is a deeper issue. Most VR interactions are visually rich and mechanically poor. The user sees a high-dimensional world but manipulates it through low-dimensional input. A controller can report position, buttons and grip. A vibration motor can signal contact. Neither can make a virtual object obey the same physical laws the user has spent a lifetime learning.

Virtual reality may therefore be waiting on a machine rather than another screen.

Arcades were hardware institutions

The arcade is often remembered as an aesthetic: black light, carpet, cabinets, attract screens and rows of joysticks. Economically, it was a way to share capital equipment.

Coin-operated amusement spaces predated video games by decades. When video arrived, the arcade became a distribution system for computing that households could not yet justify. A cabinet could combine a display, custom controls, specialized electronics, a durable enclosure and an operator who maintained it. The Strong Museum places the video arcade’s golden age from the late 1970s through the mid-1980s.12 The National Videogame Museum emphasizes that arcade machines were a locus of technical advancement before home systems absorbed more of the same capabilities.13

The home console did not kill public play. It killed the arcade’s hardware monopoly.

LAN centers followed the same pattern. In much of North America, the ritual became a hobby organized among friends and a niche business because gaming PCs, home networking and broadband improved. South Korean PC bangs persisted at a different scale because they became more than equipment rental: research describes them as “third places” embedded in social interaction and gaming culture.14

This distinction matters. A future arcade cannot rely only on nostalgia or social atmosphere. Those are valuable, but they are not enough to overcome the convenience of the home. It needs an expensive physical capability that improves when concentrated in a venue.

A full-body force machine has exactly that property. It requires floor area, high-power actuators, redundant sensing, inspection, calibration, staff, maintenance and controlled access. Even after the technology becomes possible, it may remain irrational to install in a house. The venue is not a compromise on the way to the home. It may be the correct permanent form.

The first arcade pooled scarce computation. The next arcade may pool scarce physical computation.

A gym is a room full of frozen programs

A conventional gym already contains a primitive version of physical software.

A chest-press machine specifies a path, a body position, a set of permitted adjustments and a resistance curve. A lat-pulldown specifies another. A leg curl, cable crossover and rowing machine each encode a different interaction into frames, pulleys, cams, weight stacks and upholstery. The program is real, but it is frozen into steel.

This is why a complete gym requires so many machines. Each device has a narrow mechanical vocabulary. It can be adjusted within limits, but it cannot become a different device through software.

The transition has already begun at smaller scale. Tonal uses electromagnetic digital resistance and adapts training through software.15 Proteus offers resistance across three-dimensional movement rather than along a single cable line.16 Black Box VR combines a headset with servo-based electromagnetic cable resistance and game mechanics; a small 12-week randomized trial reported several favorable fitness outcomes, although its sample is far too limited to establish general superiority over conventional training.1718

These products are not the machine proposed here. They are evidence that the boundary between software, strength equipment and immersive content is already dissolving.

Why now? Not because motors or force sensors were recently invented. It is because formerly specialist subsystems have become dependable modules. Servo drives, brushless motors, optical encoders, load cells, inertial sensors, depth cameras and real-time controllers can be integrated without inventing each component from first principles. Fitness products prove that software-defined resistance can survive commercial use. Modern headsets provide a precise spatial coordinate system. Robotics supplies decades of force-control methods. The remaining difficulty is not one miraculous actuator. It is a system architecture that makes many ordinary components behave like one trustworthy machine.

The larger step is to generalize resistance in space. A machine should be able to apply a vector of force to several attachment points while observing the user’s pose, motion and effort. It should change the force profile within a repetition; assist one phase and resist another; create eccentric overload without requiring a human spotter; simulate drag or a spring; constrain a joint to a safe path; and remove load faster than a stack of iron can fall.

This would not replace every valuable form of exercise. Running, jumping, throwing, balance work, free play and sport-specific practice each contain dynamics that may not fit inside a force cell. The initial claim can be narrower and still be consequential: one programmable machine could absorb a large share of the equipment devoted to machine- and cable-based strength training while making each interaction more measurable and adaptable.

The force room

The machine can be described without fixing its final shape.

Imagine a person standing inside a bounded frame. A sphere is an attractive diagram because it provides attachment directions around the body. A commercial product might instead use an octagonal gantry, a rectangular cage or a smaller hybrid structure that folds around the user. Heavy motors and gearboxes remain outside the human workspace. Lightweight cables, handles, cuffs, belts and compliant end effectors enter the space. An instrumented floor measures pressure and supports limited changes in incline, stability or motion.

Cable-driven robots are relevant because cables can transmit force while keeping much of the moving mass away from the person. Research systems such as Wireality have used multiple strings to render shapes and resistance in VR, while cable-based haptic interfaces have explored reconfigurable workspaces and force feedback over larger volumes.1920 Cable systems also have limitations: cables pull rather than push, can interfere with one another, create singularities and require careful routing. A useful machine would probably be hybrid rather than ideologically pure.

The machine’s essential design goal is not maximum force. It is high dynamic range. During free motion, the interface should disappear: low friction, low inertia, low cable drag. During a heavy movement, it should generate substantial, accurately controlled resistance. During an error, it should remove energy and become harmless. Series-elastic actuation is one relevant idea because intentional mechanical compliance can improve force control and absorb shocks rather than transmitting every control error directly to the person.21

A complete force room might combine four physical subsystems:

The general-purpose machine is therefore not a robot octopus that improvises around the user. It is closer to a physical rendering engine. Programs request forces and constraints using a limited vocabulary. The runtime maps those requests onto the available actuators while accounting for the user’s body and the safety envelope.

A developer might specify a virtual interaction as a combination of physical primitives:

Graphics engines made it unnecessary for every game studio to write a renderer. A force runtime would attempt the same abstraction for physical interaction. The largest technical moat may not be a patented motor. It may be the accumulated body model, safety kernel, calibration protocol, interaction vocabulary and test harness that make third-party physical programs possible without granting those programs raw authority over the machine.

The body becomes an account

A normal game account remembers identity, purchases, friends, achievements and progress. A force platform would need to remember what the user’s body can safely do.

Before the machine can produce convincing worlds, it must build a calibrated model of the person inside it. That model would include body dimensions, comfortable joint ranges, force–velocity relationships, movement asymmetries, preferred stance, attachment fit, prior injuries disclosed by the user, recent training load and uncertainty about every measurement.

The fastest signals would be mechanical: cable tension, actuator current, position, acceleration, repetition velocity, range of motion, foot pressure and left–right differences. Surface electromyography could contribute in controlled settings, but it is sensitive to electrode placement and motion artifacts. Heart rate and HRV are useful for session readiness and recovery context; they are less suitable as the primary millisecond-level controller of a heavy repetition.

The machine should therefore separate two kinds of adaptation.

The training policy can change over minutes, days and months. It decides which exercises to select, how much volume to prescribe and when to progress. Historical performance, sleep, heart rate, reported soreness and longer-term physiology may inform it.

The force controller operates within each movement. It watches position, velocity, tension and pose. It can maintain a target speed, alter eccentric resistance, identify a failed repetition and unload. Velocity-based training research supports the use of movement velocity and velocity loss as useful measures of performance and fatigue, although no single threshold is universally optimal.27

This creates a coherent data flywheel rather than a decorative “AI feature.” The gym measures strength and movement under controlled conditions. Games use those measurements to scale interactions. Games expose the user to varied movements that improve the model. The better the model becomes, the more confidently the machine can personalize both training and play.

The body account is also where the product becomes more than a universal cable machine. Two users can enter the same virtual world and encounter the same designed difficulty even when one is much stronger. A heavy virtual door is not defined by 300 newtons. It is defined by the fraction of a particular user’s safe capability, the direction of movement, the posture and the intended emotional effect.

That is how physical game design becomes possible. The designer specifies meaning. The platform compiles meaning into individualized force.

The safety kernel

A machine that can apply force to a person is a machine that can injure a person.

This is not a disclaimer attached to the thesis. It is the central engineering fact.

Programmable resistance can remove familiar hazards: a bar falling on the chest, an uncontrolled weight stack, a failed repetition without a spotter, or a load that cannot be reduced quickly. It introduces new hazards in exchange: sensor errors, control instability, cable entanglement, excessive force, bad pose estimation, software faults, unexpected power behavior and content that requests something the hardware should never perform.

The product is therefore not “AI-controlled exercise.” A learning system may suggest a workout or estimate fatigue. It should not possess direct, unconstrained authority over high-force actuators.

Several existing standards indicate the adjacent disciplines, although none can be treated as a turnkey certification route for this proposed system. ISO 20957-2:2024 covers additional safety requirements for stationary strength equipment, including electrical and magnetic resistance.22 ISO 13482 addresses personal-care robots and human–robot physical contact.23 ISO 13849-1 provides methods for designing safety-related control systems, including software, while ISO 13850 addresses emergency-stop principles.2425 The actual route would depend on product claims, jurisdiction, medical use, actuator architecture and whether the system is categorized as fitness equipment, robotics, machinery or something new.

The first facilities should therefore be supervised. A staff member can fit attachments, inspect wear, calibrate new users, observe unusual motion and stop the session. A controlled venue can collect fault data across thousands of uses before the company even considers an unsupervised product.

Safety also determines the machine’s content model. A normal game can ship a bug that causes a character to fly through a wall. A physical program cannot ship a bug that causes a cable to accelerate a human limb through an unsafe range. Every interaction primitive would need tests, limits and a clear physical semantics. Content might be signed, versioned and replayable against a simulator of users, poses, sensor faults and emergency transitions.

The safety system is not overhead that delays the “real” product. It is the product. If a company can build a trustworthy, testable layer that safely renders forces to different bodies, the gym and arcade applications become software markets above it.

Why start as a gym

The gym is not merely the easiest demonstration. It is a unusually good bootstrap market for this machine.

Strength training already sells controlled resistance. Users expect to grip handles, wear attachments, follow defined movements and tolerate instruction. Progress can be measured. Sessions recur several times a week. A premium studio can employ staff. The machine can begin without a headset, which removes an entire category of discomfort and lets the company prove mechanics before adding illusion.

The first commercial promise should be conservative. It should not be “every exercise in one sphere.” It might be a high-quality 25-minute full-body strength session with automatic setup, controlled eccentric loading, personalized progression and fast unloading. The machine could remember the user’s ranges and handle positions, change resistance continuously and give a coach a clean record of what actually happened.

The premium market is a reasonable beginning because the machine will be expensive, unfamiliar and operationally demanding. Luxury gyms already sell time, attention, design and privacy—not only access to iron. A private or semi-private force cell could justify higher revenue per square meter while the equipment is still scarce.

But the long-term case does not depend on permanent exclusivity. Society did not wait for every household to afford a mainframe before computing created value. Shared infrastructure can expose expensive capability early, improve it through use and gradually lower cost. The relevant question is whether the cell creates enough value per occupied hour to finance a learning curve.

That produces a staged development path:

This order also disciplines the technology. A game demo can hide weak hardware behind novelty. A gym cannot. If resistance feels rough, calibration takes too long, attachments fail, uptime is poor or users do not become stronger, the product has failed in measurable ways.

The gym forces the company to build a machine rather than a spectacle.

From training to play

The transition from gym to arcade should not occur as a sudden pivot. It should occur as training programs gradually acquire narrative and games gradually acquire physical consequence.

At first, a row is still a row. The user may see a boat, a river and teammates, but the force profile remains a controlled exercise. A cable press may become the act of moving a gate. A squat may charge a jump. A rotational pull may draw a machine, open a valve or steer a vehicle. The world supplies meaning while the movement remains legible to the safety system.

This is more promising than grafting exercise onto an unrelated game after the fact. The physical primitive becomes part of game design from the beginning. A designer works with the body rather than treating movement as a calorie tax attached to entertainment.

Over time, the vocabulary can expand. A compliant surface can become a shield. Drag can become water. A directional spring can become a bow. Controlled recoil can make a tool feel energetic. Multiple users can pull against a shared simulated object without ever receiving each other’s raw force directly. The runtime mediates the interaction and preserves each body’s limits.

The crucial conceptual fusion is this:

An exercise is a force program with a physiological objective. A game is a force program with a narrative objective.

The same software object can serve both. A virtual climb might be designed to improve pulling endurance, but it can also be a cooperative expedition. A rowing session can target a cardiovascular zone while functioning as a race. A boss encounter can produce interval training without forcing every player to generate the same absolute power.

This does not mean every game should disguise exercise. The arcade becomes a true medium only when some experiences are worth doing for their own sake. But the early overlap is strategically valuable because it lets content arrive before the hardware is general enough to simulate arbitrary worlds.

Existing location-based VR shows that people will visit venues for an experience beyond what they own at home. Sandbox VR uses group full-body tracking and haptic equipment, while Zero Latency operates free-roam multiplayer systems across a venue network.2829 Their value comes from space, setup, group coordination and specialized equipment. A force arcade would add something more difficult to reproduce at home: sustained programmable interaction with the body.

The content economics may eventually become more attractive than the hardware economics. A cell is capital equipment. A new physical world is software that can run on the installed base. Once the platform has enough standardized primitives, content can increase utilization without rebuilding the venue.

That possibility is what turns a clever gym machine into a platform thesis.

The arcade returns

The modern arcade should not compete with the living room on screen size, graphics or game libraries. The living room has won those categories.

It should offer a class of machine the living room cannot safely contain.

That venue could also occupy a healthier place in everyday life than the classic arcade.

Adults are advised to perform regular aerobic activity and muscle-strengthening work; the World Health Organization recommends at least 150 minutes of moderate aerobic activity per week or the vigorous equivalent, with strengthening work on two or more days.26 A physical arcade does not automatically satisfy those recommendations. A game can be exhausting without being well-designed training, and a venue can market “fitness” while delivering little useful load. But the hardware makes it possible to measure what actually happened.

A person might visit three times a week with friends. The session could combine strength work, interval movement, skill, competition and cooperative play. Progress would be partly physiological and partly ludic. The social group would create a recurring appointment rather than a vague intention to exercise alone. The game would create anticipation. The machine would ensure that physical demand remains individualized.

The result could resemble neither a gym nor an old arcade. It may feel closer to a club organized around embodied games: teams, seasons, leagues, coaching, progression and new worlds released onto a stable physical platform.

The arcade’s golden age was partly a youth institution because young people had limited access to advanced hardware. The next era could be broader. Adults already pay for boutique fitness, escape rooms, climbing gyms, padel courts, golf simulators and group entertainment. A physical arcade would sit at the intersection: more repeatable than an escape room, more varied than a single sport, more social than a home workout and more materially consequential than ordinary VR.

That is a better ambition than the metaverse as a collection of virtual rooms. The problem was never only that virtual rooms looked insufficiently real. It was that entering them did not give the body enough to do.

What would have to be true

A thesis this physical should end with failure modes rather than inevitability.

The machine could be technically possible and commercially wrong. It could become a beautiful prototype that is too slow to fit, too expensive to maintain and too frightening to insure. It could produce useful resistance but never convincing contact. It could be safe only at forces too low to matter. It could create an impressive first session and weak long-term retention. The hardware could be general while the content remains narrow.

There is also a philosophical limit. No control system can make all physical interaction safe. Contact sports, falls, high-speed locomotion and sharp impacts involve energy that may be incompatible with a broadly accessible venue. A successful force platform may remain deliberately stylized. It can create convincing resistance and constraint without reproducing every dangerous property of reality.

This is not a weakness. Games have always abstracted. A joystick does not reproduce driving; it preserves the decisions that matter. A force runtime should preserve the physical meaning of an interaction while discarding energy that does not improve the experience.

The machine does not need to simulate reality perfectly. It needs to create a new, coherent physics that the body can learn.

The thesis

Virtual reality has repeatedly been introduced as the next display. That framing made sense when displays were the bottleneck. It is less useful now.

The visual stack is good enough to reveal the absence of the physical stack. The missing layer is a machine that can apply programmable force to a person across useful movements, understand the body inside it, and refuse unsafe commands regardless of what the content requests.

The gym is the right first market because it asks the machine to solve a real physical problem before it is allowed to become a dream machine. It supplies repeat use, measurable outcomes, supervision and willingness to pay. It teaches the system what each user can do. It finances better actuators and safer controls.

Once the machine can safely render weight, drag, springs, constraints, surfaces and limited impacts, the distinction between exercise equipment and virtual reality begins to disappear. The same body account can personalize a workout and a world. The same force primitive can produce a repetition or an encounter. The same venue can become a gym in the morning, a league at night and an arcade on the weekend.

The resulting company would not begin with the metaverse. It would begin with a person, a machine and a controlled load.

It would learn to move the body before it attempted to move society into a virtual world.

The gym is not adjacent to the arcade. It is how the arcade earns a body.