14 Dec 2024/length unknown/Orkhan Mammadov
MLow X Orkhan Mammadov
Orkhan Mammadov traces his path from a country with no accredited art school to inscribing TensorFlow on Bitcoin, and lays out his plan for an artist-owned AI model.
Orkhan Mammadov opens with the whole arc of his life so far. He designed website layouts at 11 or 12 in Dreamweaver and FrontPage, studied computer science because Azerbaijan had no accredited art university, and transferred abroad to visual communication design at Bilkent after the dean, a family friend, told him losing that kind of talent would be a waste. A 2018 government commission produced the pattern installation that became his SuperRare genesis, collected by OSF. In late 2020, while interviewing for a design director job at a bank, he heard a podcast from the 2018 Christie's art and blockchain summit, minted two works, sold each for about ten ETH overnight, and within a week was applying for a US visa, which he got in two days by writing the ambassador that he had to see a Snoop Dogg concert. Two years in Miami, a government branding job in Dubai, and a move to London followed.
The middle of the conversation is craft and heritage. Holy Math, his Ordinals collection, recreates more than ten hand-drawn pattern notebooks that vanished in a shipment from Dubai to London; he launched it during a market crash and refuses to watch market conditions since a collector told him the only way to reward holders is to make bigger and better work. His next drop trains CPPN, an AI model from 2007, on his carpets, and to make it run fully on chain he paid to inscribe the entire TensorFlow library, about two and a half megabytes, onto Bitcoin as a public resource any artist can now use. Carpet DM extends his core argument: carpet patterns were passed from memory, woman to woman, for thousands of years with no drawings, and tokenization finally gives those designs ownership, provenance and inheritance.
The last stretch looks forward. He calls AI agents the new meta and warns that AI will automate away millions of jobs faster than people expect. His answer is Genie, a decentralized diffusion model built on licensed datasets plus community dataset farming, governed by guardians and rewarded through a token, which he pitches as a global Spotify for AI art where 90 percent goes to artists instead of Midjourney's model of 300 million dollars a year and zero paid to the artists it trained on. MLow, who bought more than ten Holy Math pieces, offers to help and floats a carpet dataset collaboration, and Orkhan closes on his collectors, including a London finance executive he calls his most devoted supporter, and on the two percent of artists who succeed outside the institutional path through nothing but constant work.
What this episode teaches
- The art world runs on networks
Good work is necessary but not sufficient. Orkhan says nobody gets discovered in a basement: you have to put the work out, go where the people are, and meet them, because every big project he has came from a random meeting at a fair, on a plane or in a train.
- Never trade on market conditions
A collector in Dubai told Orkhan the only thing an artist can do to reward holders is create bigger and better work. He launched Holy Math the week markets lost trillions and people feared war, and it worked. The artist's job is to keep creating, whatever the market says.
- Design for the medium's constraints
For Ordinals, Orkhan advises artists to forget high resolution output. Keep pieces around 400 kilobytes or less, generate SVG rather than JPEG or PNG, and treat recursive simplicity as the aesthetic, because the constraint itself is what makes on-chain work distinct.
- Inscribe infrastructure once, everyone benefits
He paid heavy fees to inscribe the whole TensorFlow library onto Bitcoin for his CPPN drop, and stresses that it is now public: any developer can call it, train a compact model under a few megabytes, and build AI-driven or agent-based work on Ordinals without paying that cost again.
- Keep heritage alive by translating it, not embalming it
Orkhan argues subcultures die when they stay old. Carpet patterns survived thousands of years passed from memory, woman to woman, with no drawings, and his answer is to continue that chain in a new medium: make the carpet generative, tokenize the design, and let families inherit it as an asset again.
- Hold IP loosely
His position is that once a collector pays, the work is theirs to commercialize: sell it, advertise with it, license it, short of destroying it or using it in pornography. If IKEA licensed the Carpet DM collection he would take about a ten percent facilitation fee and pass the benefit to holders.
- Collector relationships are the real payment
Commercial clients threaten to sue; collectors share his story, message him about his wins, and stay for years. Orkhan says about a thousand people actively collect his work, and that this organic love cannot be bought or manufactured, only returned.
The conversation, topic by topic
From layouts to computer science to design school
Orkhan started designing web layouts at 11 or 12 in Dreamweaver, FrontPage and Photoshop, making websites that functioned as nothing but artworks. Azerbaijan's art school was not accredited as a university, so he studied computer science until the dean, a friend of his father, told him it was not his place and helped him transfer to visual communication design at Bilkent University, where he moved into new media and UI/UX work.
The 2018 commission and the SuperRare genesis
The Ministry of Culture and the Heydar Aliyev Foundation commissioned installations from him in 2018. One, a circular repetition of white patterns changing into each other, ran on a transition model his team invented, mapped the animation to generated sound, and became his SuperRare genesis mint, collected by OSF. Two universities invited lectures and professors cited it in a paper on Islamic and Middle Eastern art.
Discovering NFTs during a bank interview
In December 2020, while re-interviewing for a design director role at the International Bank of Azerbaijan, Orkhan watched a podcast from the 2018 Christie's art and blockchain summit. Ten minutes in he realized everything would be tokenized. He minted two existing works, including the Morocco animation shown in Venice, and each sold overnight for about ten ETH, which he took as the sign to go all in.
The Snoop Dogg visa and the Miami to Dubai to London years
Orkhan wrote the US ambassador in Azerbaijan, whom he had met at his own exhibition, that he needed a visa to attend a Snoop Dogg concert, and got it in two days. He spent two years in Miami, took a UAE government job building visual identity for a major project, showed a collection at Art Dubai, found Dubai unlivable, and moved to London in December.
The studio, the gallery plan and the ten year horizon
Orkhan describes his practice as a team: friends handle sound design, CGI renders and TouchDesigner work, all sharing a vision of a new Middle Eastern aesthetic he says has not been done before. He is planning a New York gallery for new media and digital art with an art curator friend, and estimates real market adoption of a new concept takes about ten years without a funded full-time studio.
Selling digital art on USB sticks
Before and even after NFTs, Orkhan sold works on DVD and USB. A Turkish hotel group recently refused to open a wallet, so he charged them the market price in ETH equivalent, boxed the file on a USB drive, and shipped it. For him the story shows why tokenization matters: provenance, secondary markets and royalties as an artist's inheritance.
Holy Math and the lost notebooks
Holy Math recreates the hand-drawn pattern notebooks Orkhan filled from 2012 as meditation for his ADHD, more than ten of them, lost when a shipment from Dubai to London disappeared. He built generative versions of the remembered traits, launched on Ordinals with collectors who founded Inscriber handling the fully on-chain build, and launched into a crashing market. MLow bought more than ten pieces and says it stood out among Ordinals art projects.
Training CPPN on carpets and inscribing TensorFlow
His next drop uses CPPN, a compositional pattern producing network from 2007, trained on his carpets to make alien but recognizably carpet-like patterns, with each pixel mapped to characters. To run it in real time on chain, his team inscribed the entire TensorFlow library, about two and a half megabytes, onto Bitcoin at heavy cost, and he stresses the library is now public for any artist, including for AI agents on Ordinals. MLow offers to introduce him to Bosshi, who builds AI agents on ordinals.
Carpets as tokenized heritage
Orkhan frames Carpet DM as continuing a tradition where women wove patterns from memory, generation to generation, with no drawings on paper. Vintage carpets already trade as family assets and anchor Islamic art sales at Christie's and Sotheby's. Tokenizing the designs, 555 in the collection with 179 holders, makes the pattern itself ownable and inheritable, and he imagines physical licensing by the likes of IKEA or Ruggable changing holders' lives. MLow adds that virtual worlds badly need carpets and furniture, not just wall art.
AI agents and mass adoption
Orkhan calls AI agents the new meta, comparing them to programming a Lego robot: code the character and behavior and it acts, creates, even buys art. He argues people already trust machines with psychological questions more than humans, warns that AI will automate millions of jobs within years, and says crypto's real adoption blockers are scams and complexity. That day he onboarded a friend in ten minutes by opening MetaMask to gift her an artwork for her birthday.
Music NFTs and art for more senses
Orkhan cannot understand why music, the biggest digital asset, has not come to NFTs; his sound designer Isel sells unlimited Bandcamp downloads with no ownership certificate. MLow argues music NFTs work best paired with visuals. Orkhan describes his Dubai immersive show Revival of Aesthetics, projections radiating from a white carpet with wood and wool scents, a mushroom trip with a sober head, and a synced immersive restaurant in Dubai that costs about 5,000 dollars a head.
Genie, the artist-owned AI model
Against Midjourney, which he says makes 300 million dollars a year while paying and crediting artists nothing after training on stolen work, Orkhan is building Genie: a decentralized diffusion model using licensed datasets from corporations plus community dataset farming rewarded with a token, launching its first licensed-data model around the first quarter. Guardians will protect it, a raise will fund salaries, and he pitches it as a global Spotify for AI art where 90 percent goes to artists. MLow asks to get involved.
Collectors, the two percent path and what is next
Orkhan thanks collectors like Tyler, a senior London finance figure who brainstorms with him nightly after putting his kids to bed, and one whose family collected Warhol before anyone knew him. Citing a data-driven book on artist careers, he says 98 percent of successful artists come through institutions, and the rest survive on love and constant work, enough to pay for ramen, shelter and internet. His CPPN drop lands the following week, snapshots already taken, with a bigger AI collection with Isel's sound next year.
In their words
“Everything in this art world is just a network, and you just need to know someone to be known somewhere.”
Orkhan Mammadov“If you're doing a table with the wood, the nail that you will use under that, no one is seeing, has to be the best one.”
Orkhan Mammadov“It's like this earring in my ear from that day. I never look at the market condition.”
Orkhan Mammadov“I just bought it and forgot about price, because I was like, okay, yes, now I finally own some Orkhan pieces, and I'm very happy.”
MLow“It's a mushroom trip with your sober head. It's just a pure mushroom trip.”
Orkhan Mammadov“A global Spotify model for the AI art, but in this case the 90 percentage will go to the artist, not the two percentage in Spotify.”
Orkhan MammadovCite this conversation
Orkhan Mammadov in conversation with MLow, "MLow X Orkhan Mammadov," The MLow Show, unaired recording, 14 Dec 2024. https://show.mlow.xyz/e/2024-12-14-mlow-x-orkhan-mammadov.html. Quoted from a machine transcript.
Quotations come from machine transcripts. Check wording against the audio before republishing.