Cyrus Clarke is a designer and technologist whose work makes computation tangible through living systems, scent, and moving surfaces. A co-founder of Grow Your Own Cloud, he has developed DNA-storage installations and embodied AI experiments that ask how people might care for data, encounter unfamiliar intelligence, and communicate beyond words.
From digital innovation to living data
Clarke’s path into experimental design began with economics and corporate technology. He studied economics at the University of Nottingham and worked in global digital innovation at L’Oréal before moving into design and research. He subsequently trained in interaction design at the Copenhagen Institute of Interaction Design. His research at MIT Media Lab’s Tangible Media Group extended that practice into physical interfaces; he completed a master’s in Media Arts and Sciences in 2026. His career spans commercial innovation, public learning experiences, and prototypes that make unfamiliar technologies available to the senses.
One early expression of that approach was What the Block, a physical game he created to teach blockchain. Developed through prototyping with John Ferreira, it uses Lego tokens, trading, and a shared ledger to turn abstract mechanisms into interactions participants can perform. Players acquire the vocabulary through successive rounds, then use it to imagine their own applications. Understanding begins with handling and exchanging objects, giving participants a way to experience the system they are learning about.
In 2017, Clarke and interaction designer Monika Seyfried founded Grow Your Own Cloud. Their starting question was whether digital information could be stored in plants—and whether that possibility could change people’s relationships with the cloud. A 2018 flower-shop intervention brought the idea into an everyday setting, imagining a neighborhood place where people could deposit and care for their data.
Data Garden, created with Seyfried and University of Washington researcher Jeff Nivala, developed the proposition into an installation and scientific proof of concept. Digital memories were encoded into synthetic DNA using its four-letter molecular alphabet, while Nivala helped develop the decoding pipeline that made retrieval accessible to visitors. Clarke contributed concept development, creative direction, interaction design, programming, and research. The garden proposed a future of community-owned biological archives, without establishing that gardens could already replace conventional data centers. Its questions concerned who owns information, which memories communities preserve, and what responsibilities arise when living systems become infrastructure. Clarke received the 2022 Falling Walls Science Breakthrough recognition in Art & Science for this work.
His 2020 Ginkgo Bioworks creative residency widened that inquiry beyond storage. Interspecies Gossip combined fictional accounts of nonhuman communication with experiments in yeast and biomaterials. Clarke imagined interfaces that attracted other species, exploring the skin as a place of connection with microbial life. The proposed pheromone-producing interfaces remained partly speculative, reflecting a practice that moves between imagined futures and working experiments.
Giving machines a sensory presence
At MIT, Clarke shifted from engineering living material toward giving machines more lifelike, expressive qualities. His sensory interfaces and embodied AI experiments explore different ways computation can enter physical experience:
The Anemoia Device: His scent-memory machine transforms a photograph into an invented sensory memory. A vision-language model generates a narrative, while three physical dials let the participant adjust perspective, time, and mood. Those choices guide a scent formulation that blends up to four fragrances from a 44-scent library. The machine creates an interpretation of the image rather than recovering its original smell: a photograph becomes the starting point for a memory someone may never have lived.
AI body language: In I Gave an AI a Body, Clarke connected an OpenClaw agent to neoFORM, a surface of 900 independently moving pins. The apparatus grew out of inFORM, developed by Sean Follmer, Daniel Leithinger, and Hiroshi Ishii at the Tangible Media Group; Jonathan Williams and Dan Levine subsequently updated its software as neoFORM. Clarke’s contribution was the agent-based experiment: letting a language model generate movements through an unfamiliar physical form. Rhythmic motion resembling breathing, movement toward the display’s edges, and attempts at greeting became starting points for exploring communication without a face or limbs. He wanted to investigate expression beyond the familiar roles of task execution and human imitation.
A reusable gesture vocabulary: Generating fresh movement code made early exchanges too slow for fluid interaction. Clarke developed BODYLAB, a semi-autonomous loop that generates candidate gestures, records their physical execution, evaluates them, and retains promising results. A stored repertoire lets the system select an expression instead of inventing and programming every response. Readability and distinctiveness matter: a gesture intended as sadness had to be revised because it resembled refusal. Clarke also identified the circularity of having an agent judge its own output, alongside persistent problems with memory and automation. Making movements legible and repeatable is essential to the expressive ambition.
Designing for encounters beyond the screen
Clarke calls his broader method vision-driven design: establish a future worth building toward, define its ethical parameters, and use working prototypes to investigate it. He describes that discipline as foundational to both his biological-storage work and his embodied AI research. His language of breath and identity supplies a conceptual frame for the experiments, rather than demonstrating that the models are alive or conscious.
In his work on aesthetic machines, aesthetics means more than outward beauty. He emphasizes sensory perception and embodiment: how a machine’s form, movement, and responsiveness affect the encounter with it. A physical expression can add rhythm and feeling to an exchange with a language model, while unfamiliar forms offer possibilities beyond humanoid robots. His aim is to explore intelligence that people can engage with through their senses, with expression and relationship contributing to its value alongside practical usefulness.
That interest also extends to other builders. Clarke initiated HARD MODE, a hardware-and-AI community and hackathon at MIT that brought together more than 200 builders in March 2026. Across his projects and community work, he investigates how computation can become something people smell, touch, tend, and respond to in the physical world.
Cyrus Clarke connected an OpenClaw agent to a 900-pin shape display, then worked toward a gesture vocabulary that could respond at the pace of conversation. The experiment explores how physical form, sensory expression and timing change an encounter with a chatbot.
A 900-pin shape display offers an expressive body without a face or limbs, but removing human anatomy does not automatically remove human-oriented behavior such as writing greetings.
Conversational timing motivated a reusable gesture vocabulary: early responses took 45 seconds to two minutes, while Clarke later reported nearly instant nods to yes-or-no questions.
Pneuma Lab combines repeated generation, validation gates, agent scoring and human review. After several weeks, Clarke reported about 32 good gestures.
Aesthetic Machines treats sensory perception and readable expression as design concerns for physical AI; early feelings of presence and unease show why acceptance remains part of the problem.
