Most of the art world’s language for technology-driven work has collapsed toward a single catch-all: AI art. Adam Heft Berninger, the artist-curator behind Heft Gallery on New York’s Lower East Side, built his entire program around resisting that flattening — showing work built from “systems,” a category he defines broadly enough to include generative code, machine learning, scanners, and, as he’s put it, maybe even Lego tiles.
Where the gallery came from
Heft opened its doors on April 23, 2025, marking the evolution of Tender, Berninger’s earlier curatorial platform, into a permanent exhibition space. Before launching his own program, Berninger spent years working with institutions including MoMA and the Public Art Fund — a background in mainstream institutional curating that makes his choice to build a gallery around a deliberately non-mainstream category notable rather than incidental.
What “systems” actually means in practice
The gallery’s inaugural exhibition, Truth or, brought together leading and emerging voices in the generative art movement, including Mario Klingemann, Katie Morris, and Michael Mandiberg — artists whose practices are built on rules, logic, structure, or code, but who resist being reduced to a single technique or tool. Berninger’s framing treats that systems-based approach not as a constraint the artist works around, but as an opening: a way of engaging with the actual conditions of the present moment, whether through algorithmic process, procedural mark-making, generative technology, or a purely conceptual framework that never touches a computer at all.
Why the label choice is the actual argument
“AI art” as a category tends to assume a specific, narrow production process — a prompt into a generative model — and it’s increasingly used as shorthand for an entire decades-long lineage of algorithmic and systems-based art that predates today’s generative AI tools by years. Berninger’s insistence on “systems” instead is a curatorial argument as much as a marketing choice: that flattening a scanner-based practice, a machine-learning piece, and a rules-based conceptual work into the same “AI art” bucket erases exactly the specificity that makes each of them worth looking at closely. For a gallery scene that’s mostly reached for the easiest label available, that’s a more demanding — and more useful — standard to hold.