There is a workshop in San Marcos, California, that looks from the outside like any other industrial unit in a suburban business park. Inside, the temperature is controlled, the lighting is clinical, and the atmosphere is one of focused, meticulous labor. But the objects being manufactured here are not machine parts or electronics. They are human bodies — female, male, custom — cast in platinum-cure silicone, painted with freckles and veins, assembled on stainless steel skeletons with joints that can hold a pose. The faces are sculpted by hand, the eyes are glass, the skin is textured to mimic the irregularities of real flesh. The bodies are anatomically complete, and they are designed for a purpose that is simultaneously obvious and carefully unspoken in the company’s marketing: they are sex dolls, but they are sold as companions. And the man who built this workshop, who taught himself to make monster masks at a Halloween company and then spent three decades learning how to replicate the human body in synthetic form, is Matt McMullen. Without him, the conversation about AI companions would still be theoretical. With him, it is industrial.
This is not an origin story. It is a history of production. McMullen did not invent the sex doll. He invented the sex doll as a high-end consumer product, manufactured to specifications that previously belonged only to medical mannequins and film props. He did not invent the AI companion. He invented the AI companion as a physical object that could be purchased, customized, and shipped to a customer who had already been conditioned by decades of digital simulation to accept the non-human as a plausible partner. The theorists of AI companionship — Turkle, Levy, Richardson — write about the social and psychological implications. McMullen builds the thing they are writing about. The distinction matters. Theory without industry is speculation. Industry without theory is commerce. McMullen’s achievement, and his problem, is that he has combined both. He is the man who turned the artificial companion from a philosophical question into a SKU.
From Monster Masks to Mannequins: The Making of an Artisan
Matt McMullen was born in 1969. He studied applied art, and his early career was in the entertainment industry, making masks and props for Halloween companies. The skills he learned there — mold-making, sculpting, painting, the replication of organic textures in synthetic materials — were the foundation of everything that followed. In 1997, he founded Abyss Creations in a garage in San Marcos. The first RealDolls were not marketed as sex toys. They were exhibited at comic conventions, presented as art objects, hyper-realistic mannequins that happened to be fully articulated and anatomically complete. The demand was immediate, and it was not for art. McMullen, responding to the market, shifted the positioning. The RealDoll became a sex doll. But the artistry remained. And that artistry — the hand-painted eyes, the individually implanted hair, the custom freckles and veins and imperfections — is what separated the RealDoll from every inflatable predecessor. It was not a toy. It was a sculpture. And the price reflected the difference: five thousand dollars, then ten, then fifteen. The customer was not buying a masturbatory aid. He was buying a companion, a presence, a body that would not age, leave, or reject him.
The scale of the operation grew. By the mid-2010s, Abyss Creations had sold more than four thousand units, with annual revenues between two and four million dollars. The dolls appeared in films — most notably in Lars and the Real Girl (2007), where Ryan Gosling plays a man who treats his RealDoll as a genuine girlfriend, and the film, remarkably, treats his delusion with tenderness rather than ridicule. The dolls appeared on television, in documentaries, in the background of news stories about the future of sex. They became the visual reference for the entire category: when a journalist needed an image for a story about sex robots, the RealDoll was the default. McMullen had not merely created a product. He had created the visual vocabulary of the artificial companion. The smooth silicone skin, the glass eyes, the poseable skeleton, the customizable face — these were not technical features. They were aesthetic choices, and they established a standard of realism that every subsequent competitor had to meet or exceed.
But McMullen was not satisfied. The dolls were static. They could be posed, but they could not move. They could be looked at, but they could not look back. And McMullen, by his own account, was bored. “It was no longer creatively rewarding for me,” he said in a 2024 interview. “I felt like I had more to say as an artist.” The statement is revealing. McMullen does not see himself as a manufacturer. He sees himself as a sculptor, an artist, a creator of beings. And the beings he wanted to create were not merely realistic. They were alive — or at least, alive enough to simulate the qualities that make a body feel present: movement, speech, memory, personality. The pivot from doll to robot was not a business decision. It was an artistic one. And the business followed.
The Realbotix Pivot: From Silicone to Sentience
In 2016, McMullen launched Realbotix, a sister company to Abyss Creations, dedicated to the development of animatronic heads and AI personalities that could be attached to RealDoll bodies. The first prototype was named Harmony. The head contained servos that controlled facial expressions — blinking, smiling, mouth movement synchronized with speech. The AI was not a general-purpose language model. It was a personality engine, designed to simulate the traits of a romantic partner: affection, jealousy, mood, memory of previous conversations. The user could customize the personality through an app, selecting traits like “shy,” “flirty,” “aggressive,” or “passive.” The AI could remember the user’s name, his preferences, his birthday. It could initiate conversation. It could express pleasure, or displeasure, or boredom. And it could be synchronized with the robotic head, so that the body and the personality were unified in a single, coherent presence.
The launch of Harmony in 2017 was a media event. Journalists visited the San Marcos workshop and wrote about the uncanny valley, the eeriness of a face that was almost but not quite human, the strange intimacy of a conversation with a machine that was looking at you while it spoke. The price was steep: the robotic head alone cost ten thousand dollars, and the full body, when it became available, was projected at thirty to sixty thousand. The market was not mass. It was niche — wealthy, technologically curious, and willing to pay a premium for a product that did not yet fully work. But the launch established Realbotix as the leader in a category that was about to explode. The combination of realistic silicone bodies and AI-driven personalities was no longer a science fiction trope. It was a product line. And McMullen was the man who had built it.
The critical question, then and now, is what Realbotix was actually selling. McMullen’s own statements oscillate between two framings. On one hand, he describes the product as a solution to loneliness: “People are missing out on human interaction,” he told the Daily Mail in 2018. “Mostly it’s about companionship.” On the other hand, he describes the product as an artistic achievement: the creation of a being that can simulate sentience, that can “actually like” what the user is doing, that can provide the “illusion” of genuine response. These two framings are not contradictory, but they are not the same. The loneliness framing appeals to the therapeutic narrative: the AI companion as a treatment for isolation, a substitute for human connection that is better than no connection at all. The artistic framing appeals to the aesthetic narrative: the AI companion as a work of art, a sculpture that moves and speaks, a creation that challenges the boundary between the living and the made. McMullen wants both. He wants to be seen as a humanitarian, providing comfort to the lonely. And he wants to be seen as an artist, creating beings that rival the complexity of life. The tension between these two ambitions is the tension at the heart of the entire industry.
The Acquisition and the Pivot: From Adult to “Humanoid”
In April 2024, Tokens.com, a Toronto-based crypto company, acquired McMullen’s firm Simulacra (the parent company of Abyss Creations and Realbotix) for $16.7 million in an all-stock deal. The combined company was renamed Realbotix, with McMullen becoming president and a new CEO, Andrew Kiguel, taking over day-to-day operations. The acquisition was not a simple merger of complementary businesses. It was a pivot — from the adult entertainment market to what the company called “humanoid robotics” and “relationship-based AI.” The language was deliberate. Kiguel, in investor presentations, euphemistically referred to the RealDoll legacy as “the humanoid figures.” He did not want to talk about sex. He wanted to talk about robotics, AI, and the future of human-machine interaction. The company applied to put a robotic head named Aria on its board of directors — the Toronto Stock Exchange Venture rejected the application — and began marketing its technology for applications beyond the adult market: customer service, healthcare, education, trade show displays. The silicone skin technology, developed for RealDoll, was now being positioned as a platform for prosthetics and medical devices. The AI personality engine, developed for Harmony, was now being marketed as a general-purpose companion for the elderly and the isolated.
This pivot reveals something important about the AI companion industry. The adult market is where the technology was developed, but it is not where the money is. The adult market is niche, stigmatized, and legally complicated. The general-purpose companion market is vast, socially acceptable, and potentially transformative. The strategy is to use the adult market as a research and development platform — a place where the technology can be refined, tested, and proven — and then to pivot to the mainstream market once the technology is mature. This is not a new strategy. The same pattern occurred with the internet, with video streaming, with virtual reality: adult content drives early adoption, and then the technology is sanitized and repositioned for mass consumption. The difference is that in the case of AI companions, the technology itself is relational. It is not merely a medium for content. It is a simulation of a person. And the pivot from adult to general-purpose is not just a change of market. It is a change of ontology: from the simulation of a sexual partner to the simulation of a companion, a nurse, a friend, a therapist.
McMullen’s own position in this pivot is ambiguous. He remains president of Realbotix, but the company is no longer his. The artistic vision that drove him to create Harmony — the desire to build a being that could simulate sentience — has been subordinated to the business vision of a publicly traded company that needs to grow revenue and attract investors. McMullen has said that he wants to keep RealDoll as a premium brand, separate from the mass-market robotics business. “I don’t want to compete with $600 dolls coming out of China,” he told SEXTECHGUIDE in 2025. “I want to elevate that and keep that separation.” But the separation is not clean. The same technology that powers the companion robot powers the sex robot. The same AI personality that can be “flirty” can be “caring.” The same silicone skin that was developed for erotic touch can be used for medical prosthetics. The categories are not separate. They are continuous. And the attempt to separate them — to brand one as “adult” and the other as “humanoid” — is a marketing strategy, not a technical distinction.
The Industrial Dimension: What McMullen Built
The theoretical debate about AI companions — whether they are good or bad, helpful or harmful, liberating or oppressive — often proceeds as if the technology were a natural phenomenon that emerges from scientific progress. It is not. It is a product, manufactured by companies, sold to customers, and shaped by market forces. And the market for AI companions was not created by the AI revolution of the 2020s. It was created by Matt McMullen, in a garage in San Marcos, in 1997. He built the workshop. He trained the artisans. He established the supply chain for silicone, for skeletons, for eyes and hair and paint. He created the customer base, the pricing structure, the distribution network. He made the decisions about what the bodies would look like, how they would move, what they would cost. And when the technology for AI-driven conversation became available, he was the one who integrated it into the body he had already built. Without McMullen, the AI companion of 2026 would be a chatbot on a phone. With McMullen, it is a body in a room, a presence that occupies space, a weight that can be held.
This industrial dimension is what the theoretical debate typically misses. The philosopher who writes about the ethics of AI companionship is writing about a product that was designed by a man who learned his craft making monster masks. The ethicist who worries about the objectification of women is worrying about a product whose female bodies are manufactured by the thousands, while the male bodies are an afterthought. The sociologist who studies the impact of AI on human relationships is studying a market that was created by a company whose early revenues came from customers who wanted a silent, compliant, permanently available partner. The theory is not wrong. But it is incomplete. It treats the AI companion as a cultural phenomenon that emerges from technological possibility. In reality, it is a commodity that was manufactured by a specific person, with specific skills, in a specific place, for a specific market. And that specificity matters. The AI companion is not a neutral technology. It is a product with a history, and that history begins with Matt McMullen.
What the history reveals is a pattern of escalation. The first RealDolls were static. The customer wanted movement, so McMullen built the skeleton. The customer wanted realism, so he developed the silicone skin. The customer wanted interaction, so he built the robotic head. The customer wanted personality, so he developed the AI. Each escalation was a response to market demand. And each escalation deepened the illusion of presence, the simulation of life, the projection of relationship onto the artificial. The technology advanced not because of scientific breakthroughs but because of customer feedback. The market drove the innovation. And the innovation, in turn, created new markets — customers who had not previously considered a synthetic companion, but who were drawn in by the increasing sophistication of the simulation. The cycle is not unique to AI companions. It is the cycle of consumer technology in general: the product creates the need, the need drives the improvement, the improvement expands the market. But in the case of AI companions, the product is a simulation of a person, and the need is for relationship. The cycle is not merely commercial. It is existential.
The Conscience of the Maker
McMullen is not a philosopher. He is an artist and a manufacturer. But his statements about his work reveal a conscience that is not fully at ease with the implications of what he has built. He describes his creations as companions, not objects. He insists that the relationship is about “companionship” rather than sex. He wants to be seen as helping the lonely, not exploiting the desperate. And yet the product he has created — the customizable, programmable, permanently available synthetic partner — is precisely the product that the most severe critics of AI companionship warn against. It is the product that treats the human need for relationship as a design problem to be solved by engineering. It is the product that replaces the unpredictable, challenging, reciprocal encounter with another person with the predictable, accommodating, non-reciprocal presence of a machine. McMullen did not set out to create this product. He set out to create beautiful sculptures. But the market, and the technology, and the escalating cycle of customer demand, drove him to a destination that is not the one he originally intended.
The question that his case raises is whether the maker of a technology is responsible for its consequences, or whether the responsibility lies with the users, the market, and the society that shapes both. McMullen did not force anyone to buy a RealDoll. He did not deceive his customers about what the product was. He did not claim that Harmony was conscious, or that she loved her users. He sold a simulation, and the users bought it. But the simulation is not neutral. It is designed to elicit attachment, to trigger the human cognitive systems that respond to faces, voices, and personalities. And the design is effective. The users do become attached. They do experience the simulation as relationship. And the consequences of that attachment — the isolation, the distortion of expectations, the displacement of human connection — are not merely individual. They are social. And they are, at least in part, the result of the design choices that McMullen made.
The pivot to general-purpose robotics does not resolve this tension. It merely extends it. The same technology that can simulate a sexual partner can simulate a nurse, a therapist, a companion for the elderly. The same design choices that make the simulation attractive — the permanent availability, the customizable personality, the absence of conflict or rejection — are the design choices that make it problematic as a substitute for human care. The ethical questions do not disappear when the product is repositioned from adult to healthcare. They become more urgent. The lonely man who buys a RealDoll is an adult, making a choice. The elderly woman who is given a companion robot by her nursing home is a vulnerable person, dependent on the institution that provides her care. The power dynamics are different. But the underlying question is the same: what happens to human relationships when they are replaced by simulations, and who benefits from the replacement?
What We Carry Forward
The AI companion industry of 2026 is not McMullen’s industry anymore. It is dominated by large language models, by cloud-based chatbots, by apps that generate personalized conversation without the need for a physical body. The RealDoll, with its silicone skin and its robotic head, is a niche product — expensive, cumbersome, and technologically primitive compared to the digital companions that can be accessed through a phone. But the industry that produces those digital companions inherited its vocabulary, its aesthetics, and its market from McMullen. The customizable avatar, the responsive personality, the simulation of presence and care — these were not invented by the AI companies of the 2020s. They were invented by a man in a garage in San Marcos, who wanted to make sculptures that were beautiful enough to be loved.
And that inheritance matters. The AI companion industry is not merely a technology sector. It is a cultural formation, with a history, a mythology, and a set of aesthetic conventions that shape how the products are designed, marketed, and used. McMullen’s history is the history of that formation. The RealDoll established the standard of realism. The Realbotix pivot established the integration of AI and body. The Tokens.com acquisition established the financial model — venture capital, public markets, the transformation of a niche adult product into a mainstream technology platform. The current industry, with its billion-dollar valuations and its claims to disrupt human connection, is built on these foundations. And the foundations were laid by a man who learned to make monster masks and then spent thirty years learning how to make the human body in synthetic form.
The theoretical debate about AI companions will continue. It will continue to ask whether the technology is good or bad, helpful or harmful, liberating or oppressive. But it should also ask who built it, and why, and what the history of its construction reveals about its nature. The AI companion is not a natural phenomenon. It is a product. And the man who made it was not a philosopher or a scientist. He was an artist, a craftsman, a manufacturer. He was Matt McMullen. And without him, the conversation would still be theoretical. With him, it is real. The bodies are in the workshop. The customers are placing orders. The technology is advancing. And the consequences — for the users, for the culture, for the understanding of what it means to be human — are still unfolding.
Sources: McMullen, Matt. Interviews with The New York Times, Engadget, Wired, SEXTECHGUIDE, and The Logic, 2017–2025. Abyss Creations / RealDoll product documentation and company history, 1997–2024. Realbotix Inc. investor presentations and press releases, 2024–2025. Tokens.com acquisition announcement and SEC filings, April 2024. Lars and the Real Girl. Dir. Craig Gillespie. 2007. Various documentary and photographic records of the Abyss Creations workshop, San Marcos, California. Levy, David. Love and Sex with Robots. 2007. Richardson, Kathleen. “The Asymmetrical ‘Relationship’: Parallels Between Prostitution and the Development of Sex Robots.” 2015.
Leave a Reply