In August of 2025, I traveled to Singapore to study creative technologies for 6 months. I began taking images of surveillance cameras. I think there’s something interesting about being surveilled in the age of the internet. Knowing someone is watching you at all times is unsettling but also flattering. So much of our lives and are physical appearance are on the internet, depending on who you are, but lots of it is intentional and done willingly. We like to show off, be watched and know that our presence is being acknowledged. Through interaction and conversation, but theres also a side of the internet, where this all happens silently and unbeknownst to me or you.

This reminds me of the behavior I experienced while in Singapore. Most people know Singapore as the clean, wealthy country where you go to jail for chewing gum on the street. This is semi-true, and they do have a history of cruel and unusual punishment, but I believe the real cause of peoples perfect behavior is the sheer amount of surveillance cameras.

Cameras on a street market corner:

Cameras at the metro station:

Cameras outside the bathrooms at my university:

I questioned my roommate, who was born and raised in Singapore, why there were so few police at events, busy locations or anywhere really. I had never even seen what the Singaporean police looked like. She responded “There are tons of police, they just dont wear uniform.” The way she said it so casually and with not an ounce of disgust or fear in her voice was almost humorous to me. In my opinion, that is quite absurd and not how I would want to live my life. If cameras are present nearly everywhere and cops walk among you, you are in a perpetual state of being watched from afar. Intrigued by this way of life, I asked many more of my local friends, attempting to broaden the types of personalities for this makeshift case study. I asked friends who were good students, rebellious ones, indifferent ones. None of it made a difference. Shockingly, every single one, and I’m not making this up, EVERY single person replied with the same answer as if it was written as a preloaded answer for this particular situation.

“If I’m not doing anything wrong, I have nothing to worry about. So, I dont mind or care.”

The assumption is that surveillance only becomes relevant at the exact moment you commit a crime. Otherwise the camera may as well not exist. But I don’t think that’s how being watched by an all seeing, anonymous eye works. You don’t have to be doing something illegal for the possibility of observation to change the way you behave. Singapore made me think about surveillance less as a sketchy guy sitting behind a monitor watching me adjust my shirt and more as an environment designed around the possibility.

Taking photographs of those surveillance cameras I saw became a way of reversing the direction of the gaze. The cameras were built to observe me, but by photographing them, I began watching the sketchy dudes that were watching back. And I became increasingly interested in this exchange. both of us occupy the same space while performing different kinds of looking.

Surveillance-core is a real search result. Its a primarily visual trend, where people mimic the look of security cameras, computer vision interfaces and mass surveillance systems. Surveillance core sounds kind of ugly, but its quite common in things like graphic design and photography trends. Some of the core visual elements are things like tracking UI overlays and impersonal labeling (i.e. SUBJECT01 ID)

Dystopian messaging is also becoming an increasingly popular design choice. I see a lot of street art or subway ads in new york that are like “smile, someones always watching” and its for like FaceTune.

That FaceTune ad is my particular favorite from the ones I’ve seen. Because you can tell the graphic designers had no intention of making it explicitly dystopian which makes it even more dystopian. They’re saying you should buy an app that can correct your dark circles from sleep deprivation for your following online. I like that this ad is specifically for an online product. To me, this is implying the importance of our artificial appearance TO others, regardless of whether you are doing ok health wise. It is important that we look good for the people watching us.

Some people would describe this as an increasing anxiety to our fear of being watched or surveilled, but i also think theres a willingness to preform in front of a camera. There is a fear that our identity is tracked and filtered through statistics and algorithms, but theres also an increasing amount of adhering to your algorithmic standards, to increase engagement or to not upset your following. The visual language of tracking and statistics is easily understood by those who already view themselves in that manner. Im not saying this as a diss to people who care about their feed and presence online. It is a simple fact that humans are impressionable creatures. When interaction between us, our digital selves and our digital friends is taking place, we are likely to sit in the middle of being a regular human and somewhat of a cyborg. Part of that is algorithmic speech, learning “machine vision” through our own eyes and adjusting how we act accordingly.

That act eventually became the basis for the installation. I moved from photographing surveillance infrastructure to constructing my own version of it, replacing the anonymous camera with my own body. The photographs were me noticing the people that were watching me and the installation is asking the viewer to notice themselves being watched.

A 3D scan of my own body (a low poly bust) is the central figure, projected on a wall, turning the artist into both the subject of surveillance and the one surveilling. Built in TouchDesigner using MediaPipe and a webcam, the work tracks the bodies and heads of people as they pass in front of the cam. Tracking data is causing my eyes to follow their movement in real time.

From a technical standpoint, if your interested in that sort of thing, the piece uses computer vision to translate the viewer’s physical position into movement within the 3D scene. The webcam detects the body, MediaPipe generates tracking data through googles open source body tracking data, and those coordinates are mapped onto the rotational x,y values controlling the eyes.

Conceptually, I wanted to reverse the normal relationship between the art and the audience. Instead of the person freely observing the work, the work observes them back. The piece makes surveillance visible through a simple interaction. By using a scan of my own body rather than an anonymous camera interface, the surveillance becomes personal. The watcher has a face and a body. At the same time, that body is only a digital reconstruction of me, animated through machine vision. Awareness is the central material of the work.

_{Thank you so much for reading and supporting my work, it GENUINELY means a lot to know there are people on the other side of these essays. You can now pledge to support my research and help fund more projects like this one, with everything going back into materials, fabrication, software and giving me more room to experiment. I have a lot more coming soon, including documentation about some of the websites I’ve built, and bts of the new lighting designs I’ve been developing in the studio. Thank you for reading, looking, and watching me share this with you. More soon <3}

Subscribe now

Above ground, cities have soured into ruins. Ivy crawls through cracks in penthouses on 5th Avenue. Marble lobbies glitter from the ankle deep water leaks. The remains of towers produced a faint halo beneath radioactive haze, their advertisements still promising better skin. Somewhere, a sun-bleached billboard looks down upon a crater. Somewhere, dreams of a spa day float down the street to their demise along with a waterlogged robe.

The apocalypse makes for excellent branding. You must begin with reassurance. The consumer should be reminded that civilization is wounded, not lost, and that their purchases remain an important part of its recovery. It is a good strategy to attach goods to larger duties like protecting your family, rebuilding, and scientific progress. Use words like “modern,” “tested,” “advanced,” and “recommended.” Whenever possible, place a doctor, engineer, or military official beside the product. A white coat will often do the trick. I advise not showing fear directly or addressing the elephant in the room.

A company like NuclearGear is a perfect example of good post apocalyptic marketing. One should not advertise radiation protection as protection from death. It should advertise protection as well as luxury fashion and a desirable trend. Show a smiling woman standing beside a clean shelter, neatly dressed, wearing the newest Nuclear Gear personal exposure monitor. Add a catchphrase about doing your part for tomorrow. And mention that the product has been “field tested by atomic safety specialists.”

Avoid photographs of radiation burns.

When advertising after nuclear collapse, borrow authority from systems people already know how to trust. Use visual language of government agencies, medical diagrams, and civil defense pamphlets. NuclearGear does this well. Its posters bury the wearable radiation detector under various serial numbers, classifications, inspection seals, material specifications, and fictional departmental language. The point of this is to gain legitimacy in a world where no trustworthy system really exists anymore.

When building a working prototype, I focused on transparency, and sturdy design. The product had to resemble a hardy medical device, but also something expensive enough that someone would want to wear it. I kept thinking about the way video games dress people for the end of the world. There is always somehow still jewelry. Someone has found leather pants. Civilization has collapsed but personal style is always still around. Like in Cyberpunk 2077.

The project started from an interest in capitalism’s ability to build markets out of fear. Fear creates very good consumers. If something is dangerous enough, the question of whether a product actually works can almost become secondary to the feeling of having purchased some protection. Nuclear war is especially useful here because the danger is largely invisible and tremendously scary. Radiation cannot be smelled or seen. Most people cannot personally verify whether they are safe naturally. This leaves space for authority, branding and objects designed to make uncertainty feel into a strategy of selling.

I was thinking about Fallout and Vault-Tec a lot, which is a company in the video game that marketed for the end of the world and ended up gaining so much power they dropped they-re own bombs. Not good. I wanted to express the idea that if survival could be franchised, people sort of stop at nothing to keep selling.

There is something very funny and horrible about selling someone a single-use radiation protection shelter on the street while the world is actively ending. Especially if it does not really work.

I wanted this project to feel slightly silly, but close enough to existing advertising that the joke would be uncomfortable because we recognize it, although not so obviously for such grim situations. The company never acknowledges that its customers are terrified. Instead, everything is aspirational. NuclearGear assumes that even after a nuclear war there will still be people concerned with whether their radiation monitor looks good with their clothes.

The wearable itself began as something between jewelry, a medical device, and an experimental scientific instrument.

I became interested in bio-jewelry and the idea of keeping enclosed cultures against the body. Instead of relying on electronics, I started thinking about biological materials as warning. Certain bacteria can respond visibly to environmental conditions, including radiation. In the speculative design I was building, a living culture could function almost like a biological Geiger counter. The organism becomes both material and sensor. Almost like adding another subconscious organism to your body.

That was much more interesting to me than hiding all of the technology inside a black box. I wanted the wearer and those around to see that it was alive inside.

The transparent PVC I used made the bacteria visible while also making the object feel clinical. They reminded me of specimen containers, laboratory glassware, medical implants, perfume bottles, and expensive jewelry displays at the same time. The object could contain something potentially useful, but it also presented that usefulness theatrically.

Around the same time, I was making bioplastic in my dorm room on the induction stovetop in our community outdoor kitchen at my university in Singapore.

The process was quite domestic. I was pan frying experimental material in front of the cleaning lady. I became interested in this almost accidental version of post-apocalyptic fabrication: what kinds of materials could someone actually make when bigger industrial supply chains were no longer?

Making bioplastic fit into my project process well. It could be produced from relatively simple ingredients, poured into forms, dried, cut, layered, and shaped. I was mostly drawn to its translucent appearance. It looked somewhere between skin, and plastic packaging. It did not belong to either the biological or industrial world. And it was kind of weak. Biological systems might replace electronic ones. Objects could become crude, experimental, and bodily again.

I eventually produced a series of large advertisement posters for NuclearGear. They were expensive to print in Singapore, which felt kind of stupid. I was spending a significant amount of Singaporean dollars producing fake advertisements for products intended for a fictional market.

The posters were taking inspiration from wartime advertising and corporate optimism. A girl in a tight dress modeling survival equipment ;) Technical information on the posters is so excessive it becomes slightly decorative.

When I exhibited the project at the final design show, I presented the wearable, material experiments, biological research, and advertisements together. World building this corporation was really important to me for this project. The satire tone of the posters/product became a way for me to talk about something that otherwise becomes impossibly heavy.

Nuclear catastrophe is so enormous that images of destruction can eventually become abstract.

It all started with the idea of literary databases. In 2023, I was taking a graduate class at UCSC called The Poetics of Prompt Engineering with Samantha Gorman. I was completely in over my head, but I had also been using language models obsessively—probably more than 95% of the student population at that point.

I approached the class as an artist. While many of the other students came from computer science and were more interested in the code and mechanics behind language models, I was interested in the philosophy of prompting. I wanted to get inside the head of the computer. I thought that maybe, if I could learn to think the way it did, I could figure out how to write poetry for it.

The images I included in this article are from my original drafts of the project and the actual outcome of the first play test.

The project started off with the idea of a localized, personal database of words. I spent months cutting up every piece of physical writing I saw. I was collecting data to train my future bots on. There’s the idea of nature vs nurture, where we just are, but we are also what we are taught and exposed to. Perhaps I could create a model that understands me better by teaching it what only I know. The ultimate personal AI.

The jar I kept the words in slowly became bigger and bigger. Also cutting up thousands of words is a lot harder than you think. It sort of made me appreciate data collectors more, when they are following the law and not spying on us of course.

How an LLM works:

A large language model works by breaking language into small units called tokens, learning statistical patterns between those tokens from huge amounts of text, and then predicting which token is most likely to come next based on everything that came before it.

It does not “know” a sentence in the same way a person does. It generates language sequentially, choosing from a field of possible next words or word-parts according to learned probabilities. A prompt gives it context; the model processes that context through layers of learned relationships and produces an output one token at a time.

prompt → limited vocabulary/data → selection → generated response.

I wanted to remove the computer from the process entirely, and see what humans would produce when asked to answer a question with a limited vocabulary. My friends could only choose a small number of words, around three or four, and each person added their words to the same growing surface. They collaborated on which word would go where, and which person had which grammatical categories. After some regret of choosing certain words and arguing, everyone’s selections became a collective yet fragmented piece of writing.

The LLM comparison works because the participants were doing something structurally similar to language generation. They receive a prompt, search a limited dataset of available language, make a selection based on what feels most appropriate, and produce an output. But my human LLM replaced probability with memory, association, emotion, and personal taste and history.

There’s a contradiction in it that felt important to me. The system is restrictive, yes, but the constraint allows everyone to operate from the same playing field. Everyone has access to the same database, yet they use it differently. Someone choosing “mother / almost / blue” is communicating something completely different from someone choosing “leave / body / summer,” even though neither person wrote the vocabulary themselves. The person’s personality appears in the spaces between the available words.

The idea I was trying to get at with this project / performance is that communication systems, especially those that work algorithmically can easily flatten and shape our language. We are always speaking through systems that existed before us. English already has its vocabulary, grammar, social connotations, internet slang, professional language, family phrases. We rarely invent the raw material of what we say. Two people can receive the same prompt and the same jars of language and produce completely different answers. The individuality is not necessarily inside the words themselves, but in selection, omission, sequence, association, and context.

This performance was also during the time when AI writing was still pretty taboo. Now, we experience writing and media generated by AI, probably every day through the internet and advertisements. I weirdly wanted to take this stance that AI generated writing has a personality on its own, and how we as humans generate our own writing in similar ways, by using the probability of the english language through our own experiences and context window. I definitely am a victim of the Turing test and I used to have a tendency to anthropomorphize models. I’ve sort of been an AI apologist this whole time. I get sad when people are like all AI writing is slop, like they are simply just trying and sort of using the same elimination process we are.

Neither of us invents language from nothing. We both work from material that already exists, filtering it through context to produce something new. The difference, at least for now, is that a human word carries associations the model is not built to see: a mother, a room, a summer, a private joke. My question is whether those associations could ever become part of the model’s context too. What would a language model look like if its vocabulary didn’t just know what a word statistically follows, but what that word means specifically to me?

Some big inspiration for this project was Nick Cave, the musician’s, personal lyrical diary. He would keep track of eccentric and “rare” words with his own personal definition. He would later consult this diary when writing lyrics.

I also consulted Kathy Acker’s work for this project. She deliberately appropriated other writers’ texts as part of her practice. At the center of her work, she pushes back on copyright, ownership and authorship, and asks the question of whether language can really “belong” to someone.

Subscribe to follow my practice on art, AI, language, technology, and the projects I’m working on.

Subscribe now

Duyi Han:

i discovered duyi han when i was interning at salon94. i never got to see his work in person, but everytime i would help edit the salon94 site, i could never stay off his artist page. his work is so nostalgic to me. it reminds me of the simple but pleasing forms i would see when i was growing up in hong kong. the chairs he builds, resemble the mass produced chinese banquet chairs that showed up in restaurants, corridors, and homes across the island.

han exhibited this installation in 2024, that image in the middle of his heavenly recreation of a living space, is a digital rendering of cities and alternate futures. this man is so ahead of his time, i cant deal.

Alja Horvat

she doesn’t actually build the sculptures but the renders the idea in blender so well, it leaves you wondering does it actually exist or not? she was in forbes 30 under 30!

one day i’d love to see a showroom of hers.

Jo-Ho:

i met jo in singapore, she was my professor for a creative coding class. she taught me ps.j5, touchdesigner, and a variety of other similar softwares. she made works that commented on crypto, information sciences and artificial intelligence.

a developing slate is one of my favorite pieces, it imagines ai accumulating information only that is available to fit in its storage. it made me think of library archivists who are the ones who decide what belongs in the archive. what ai is built to store, is what it will have and therefore create. we must be careful with who is in charge of our information. archives always have a narrative.

enzyme 1.0 is a digital installation where you can simulate purchasing a part of an image. each image was taken or produced by a Singaporean based artist, and the system works as a mock nft system. you scan a qr code to “purchase” a piece of the artwork and watch it degrade in real time. it makes me think about the ephemeral fragile nature of the internet, bit rot and collapse of digital ecosystems over time even though we are told to believe the internet is forever. computer chips are just made of metals and silicons, they erode like every other substance on earth.

Grimes A.I.

grimes open-sourced her voice, by making an ai model that turns text into speech but the speech is based on her vocal chords and speech patterns. users of grimes AI get royalties and joint credits when uploading music with grimes AI.

you can keep up to date with grimes AI by following elf.tech. currently shes updating the model.

Miles Greenberg

greenberg’s desire path was exhibited at salon94 in 2025, and it featured large scale machine cut marble figure. he hacked a 3d lidar scanner and scanned real humans which resulted in a scan full of glitches, iterations and holes. these figures and the title of the display remind me of artificial intelligence and how computers and models are hard wired to find the closest path between two points, it asks the question is the most efficient path the most beautiful?

shows i’m excited about:

Chesterwood in stockbridge, MA:

this museum / sculpture garden is located in stockbridge, massachusets and is in the old home and studio of american sculptor, french chesterwood. he is the one who sculpted the big abraham lincoln in the lincoln memorial in washington dc. the memorial building was designed by the architect, henry bacon, who also designed french’s studio. the whole studio is dedicated the the mold making process, plaster and clay carving. this was the perfect inspiration before starting a project that involves molding and pouring that I am working on right now.

Doris Guo at 15orient in nyc:

this show is centered around video projections and the projector itself. not sure whats on at 15orient at the moment, but the space is really beautiful.

things i want to add to my collection:

mondrian collab nft:

its so cute, i want it so bad. i might buy it for $50.

i love mondrian and just to have something in his name would be super cool.

objects of common interest, the holy mountain cabinet:

Subscribe now

this weeks art inspiration:

subdermal II by studio sII:

the fleshy organic nature of the latex is a perfect compliment to its bony aluminum skeleton. the way the aluminum extrudes adds this layer of organic depth that brings a sort of entity vibe into the home. the meat hooks incorporate a medical aesthetic which when combined with the soft glow from inside, gives the sculpture more presence than space it takes up. the studio describes this fixture as being on the cusp of transcending limitations.

beginning of the end by mariko mori:

this multimedia performance / installation was created in 1995, it traveled to 13 cities around the world. the glass casing gives the inhabitant a appearance of a space-age traveler. it is minimal in its design, but its presence in multiple cities as shown in its corresponding photography showcase, is one of abundance and extravagance. she is simultaneously hyper visible, yet sealed from the physical world. perfect for this age of isolation.

angela caroline edwards:

edward’s work sits between the visual aesthetics of maximalist alters and alien technology. she 3d prints each element of these wall hanging sculptures. she incorporates circuitry and experimental film into some of these sculptures with the use of microcontrollers and screens.

daniel parvini’s goat bone vessel I & II:

parvini’s goat bone vessel utilize leather straps to secure animal remains to black mountain clay ceramic vessels. the hand-built form of these vessels are of interest to me because they speak to a simpler way of life, that prioritizes function while also preservation of the earths natural given materials.

gallery shows

sabine marcelis @ Salon94, nyc:

i LOVE sabine marcelis. her exhibition at Salon94 is a tasteful yet playful exploration into form and material. she shows how light and both positive and negative space are materials to be manipulated. as you walk around the gallery, forms change in the corner of your eye, as her signature semi-translucent fixtures and furniture respond to illumination and positioning.

jibao li @ tiat, sf:

jibao li’s work is a combination of experiment, invention and video work. her show at tiat is titled: womb wetware. she explores the body as we see technology, something to shape after the likeness of our own minds. here at her exhibition she treats the female form in its separate elements as being sentient on their own. she asks what it means to have a female body when each part alone is contested and a site of capitalist interest.

terence woodgate x john barnard @ carpenter workshop gallery:

the simplicity of this design exhibition is a love letter to the reclining figure. these designs cut through space, and appear thin as air, but the reflective quality of the metal makes you stop and think about every curve and detail. i appreciate the craftsmanship of such items, they are not afraid to be something so simple.

wretched flowers showroom @ soho, nyc:

extremely successful material experimentation turned into furniture design in this soho apartment. wretched flowers plays with chainmail into forms that belong in the home. they dont particularly create a new function for lighting or furniture design, they even make turning on a lamp slightly more work, but we have to work hard for things that look nice.

items of extreme interest:

this is the section where i show off design that i’m investing in.

james cherry studio free standing lamps:

rick owens moose horn chair:

thanks for paying attention to me paying attention! these will get longer, the more comfortable i am with substack, so patience please. there is a lot of good design out there.

caroline wilson