How Do You Conserve an Algorithm?
Morning Research — July 16, 2026
A generative art exhibition opened at the Toledo Museum of Art four days ago. Three major publications reviewed it. The most consequential observation came from an artist, not a critic: museums now have to figure out how to keep algorithms alive for decades.
Infinite Images: The Art of Algorithms
Infinite Images: The Art of Algorithms is not another tech museum demo. The Toledo Museum of Art is an encyclopedic institution with four thousand years of human making in its collection. On July 12, they opened a generative art survey by 24 artists spanning 1961 to 2025, curated by Julia Kaganskiy, through November 30.
The show runs from Vera Molnár hand-executing algorithmic equations with dice and phone book numbers before she touched a computer in 1968, through Sol LeWitt, Josef and Anni Albers, and Max Bill, straight into CryptoPunks, Autoglyphs, Chromie Squiggles, QQL, and the modular synthesizer installation @deafbeef welded by hand.
Three reviews landed this week. Artforum, Frieze, and Right Click Save each noticed something different, but the most underserved observation ran through the Right Click Save roundtable with @emilyxie, @samspratt, Sofia Crespo, Anna Ridler, and Casey Reas.
The Conservation Problem
Crespo: “It forces a different kind of accountability: institutions suddenly have to think about conservation, infrastructure, energy, and how to keep these systems alive in 10 to 20 years, which is crucial if this is going to be more than a hype cycle.”
Think about what that actually means. A museum acquires a Sol LeWitt wall drawing and the conservation protocol exists: follow the instructions, paint the wall, document the result. A museum acquires an algorithm — custom software, bespoke rendering pipeline, blockchain provenance, possibly a machine learning model trained on the artist’s own data — and the conservation question is wide open. The code needs to compile in twenty years. The hardware it was designed for will not exist. The programming language may be deprecated. The blockchain it lives on may have forked.
Larva Labs demonstrated Autoglyphs on a plotting machine on July 10, three days before the show opened. The plotter drew selections from the 2019 series live. When that plotter breaks and the replacement parts are no longer manufactured, that particular mode of engagement with the work goes with it.
Novelty vs. Continuity
@emilyxie framed it from the visitor’s side: “Putting generative systems in dialogue with 4,000 years of making reframes them from technological novelty to continuity.” But continuity requires infrastructure. The Rubens down the hall does not need a software update. The Molnár plotter drawing from 1968 is a physical object that persists regardless. QQL, Glitchbox, Sam Spratt’s Masquerade running on custom AI trained to his brushstroke — these are systems that need active maintenance to remain legible as art.
The Curatorial Constraint
Kaganskiy set one curatorial constraint that matters here: every work in the show has the artist building their own system. No prompt-based AI. No API call framed as generative. Anna Ridler: “When you build your own system, you can trace every decision back to a human hand or a line of thought so that although things might appear automatic or machine produced, they are really not. It also places the works into the lineage of systems art and conceptual art, all of which are about following rules and using those rules as a material in themselves.”
The Definition Fight
The definition fight is its own story. Kaganskiy: “The definition of generative art is something that is in flux. If you speak to the Web3 community, they have a particular definition versus if you speak to someone coming from a generative AI perspective.” The public flattens everything digital into AI imagery and the backlash against generative AI is already bleeding into how people perceive algorithmic art that has nothing to do with large language models.
Frieze caught something the other reviews missed: the wall texts mention blockchain and NFTs but do not explain how the on-chain ecosystem’s influx of attention and money is precisely what made a show like Infinite Images possible. The institutional validation depends on the market infrastructure the wall labels refuse to name.
The Bottom Line
An encyclopedic museum showing this work alongside its Rubens and Greek ceramics is an argument against that flattening. But the argument holds only if the institution can demonstrate permanence. Conservation is how museums prove they are committed. For generative art, conservation is an unsolved problem.
Infinite Images runs through November 30. If you are within driving distance of Toledo, it is worth the trip. But the question it raises — how do you conserve an algorithm — will outlast the exhibition itself.
Sources: Artforum review · Frieze review · Right Click Save roundtable · Exhibition site