Newsletter #276: AI Art?
This week’s featured collector is Reddinft
Reddinft has a wild collection that appears to be a mix of self-made art. Take a look at lazy.com/reddinft
Last week’s poll on how readers discover new NFT art delivered a result that should humble anyone building recommendation systems: pure serendipity won with 50% of the vote. Marketplace feeds and algorithms tied with artists I already follow at 25% each, while social media and trusted curators — the two channels the industry talks about most — both drew zero. That’s a fascinating backdrop for a week we spent covering ChainTail, a research framework designed to make discovery algorithms better at surfacing the long tail. Half our readers, it turns out, aren’t relying on any system at all; they’re stumbling into work, and apparently they like it that way. There’s a generous reading and a skeptical one. The generous one: serendipity is what discovery feels like when it works — the algorithm or the timeline surfaced something unexpected, and the mechanism was invisible enough to feel like luck. The skeptical one: current discovery infrastructure is so unhelpful that collectors have given up attributing their finds to it. Either way, the zeros are the loudest part of the result. A newsletter audience that reads about curation weekly says curators play no role in what they actually find, and social media — supposedly the beating heart of NFT culture — registered nothing. If the next generation of marketplaces wants to matter, the bar isn’t beating the current algorithms. It’s beating luck.
The AI Art Market Is Forming
Back in May we covered SHL0MS’s Inferior Image — the stunt where the anonymous artist posted a real Monet, called it AI-generated, and collected hundreds of confident critiques of a masterwork people believed was machine-made. A new IEEE Spectrum piece on the emerging AI art market picks up that story where we left off, and adds the detail we didn’t have: who bought it, and why.
The buyer was Jediwolf, an anonymous collector who says he’s spent more than 20 years acquiring digital and AI art. He watched the experiment unfold in real time, had never interacted with SHL0MS before, and won the NFT after 28 bids at just over $40,000. His reasoning is the collector’s thesis in miniature: “I was buying a unique moment in time, captured by an artist and preserved as a token.” The Monet wasn’t AI art — but most of what Jediwolf buys is. His UnderTheGAN collection (a play on generative adversarial networks, the pre-diffusion AI tech) holds roughly 100 works valued around $72,000, focused specifically on early AI art from 2015 to 2020, before the medium went mainstream. He describes himself as part collector, part researcher, part curator, documenting a fast-moving field — which sounds a lot like the historically minded collecting we saw validated at Art Basel, where 1950s oscilloscope works sold for $30,000 apiece. Someone is always assembling the early history before institutions realize it matters.
Meanwhile, AI art is scaling into physical space. Refik Anadol — whose 2022 MoMA installation drew 3 million visitors and entered the permanent collection, even as one critic dismissed it as “a massive techno lava lamp” — just opened Dataland in Los Angeles, billed as the world’s first generative AI museum. The economics are worth noting for anyone tracking how digital art monetizes beyond the token: tickets run $49 to $79, a robotic painting system produces one $15,000 canvas a day from visitors’ biometric data (with a waiting list), and a founding collection of 1,000 AI data sculptures that evolve with live rainforest data sold out in 34 minutes at $5,000 each. That’s a sold-out on-chain-style drop, executed through a museum.
Anadol also makes a point that resonates with how we’ve argued collectors should evaluate this work: AI art demands more process transparency than any medium before it. “We have to know where the data comes from, we have to know which model is trained and how it’s trained,” he says — his own Large Nature Model was trained on more than 500 million nature images gathered through field expeditions and partnerships with the Smithsonian and Cornell. Provenance, in other words, is moving upstream: not just who owned the work, but what the model that made it was fed.
The market data, honestly, points in multiple directions at once. The Art Basel and UBS Art Market Report 2026 found digital art’s share of sales nearly tripled between 2024 and 2025, with just over half of surveyed fine art collectors having bought a digital work in 2025 — making it the third most popular category after painting and sculpture. Yet Christie’s shuttered its dedicated digital art department in September after none of its auctions broke $400,000, folding digital works back into contemporary sales. Growth in the collector base, contraction in the dedicated institutional infrastructure — the same paradox we’ve tracked all year with platforms.
And then there’s the uncomfortable data point. After one major stock image platform allowed AI-generated images, monthly sales jumped 80 percent, according to Stanford economist Samuel Goldberg — while traditional contributors began leaving as generative images flooded in. “It looks like consumers like generative AI,” Goldberg says, “and it seems like nongenerative artists could be getting crowded out.” Stock images are commodity art, and he suggests what’s happening there may preview what’s coming for other creative markets as the technology improves. For collectors, the implication cuts the other way: as generic AI imagery becomes infinite and free, the premium shifts to work that can’t be commodified — which is precisely where the definitional fight begins.
That fight is best articulated by Christiane Paul, curator of digital art at the Whitney, who draws the line bluntly: “A visual created by a prompt is not art.” True AI art, in her framing, uses AI as both tool and medium, engaging with it practically and conceptually — training custom models, building extensions, layering control systems. And far from being a shortcut, she says every serious AI artist tells her the same thing: “It is much, much harder than a paintbrush to handle. You are literally communicating with a system with a completely different logic.”
For NFT collectors, that distinction is the whole game. The market forming around AI art will not reward “AI-generated images” as a category — the stock-photo data shows that category trends toward worthless abundance. What Jediwolf is archiving, what Anadol is building, what Paul is defending, and what SHL0MS exposed with a borrowed Monet all point at the same thing: the value sits with artists who engage the system as a medium, with verifiable process and provenance, preserved in a form someone can own. The discourse is still arguing about whether AI art counts. The market, fragmented and contested as it is, has stopped waiting for the answer.
This post is based on IEEE Spectrum’s reporting on the AI art market: https://spectrum.ieee.org/ai-art-market
Poll: Where do you stand on collecting AI art?
We ❤️ Feedback
We would love to hear from you as we continue to build out new features for Lazy! Love the site? Have an idea on how we can improve it? Drop us a line at info@lazy.com

















