Thomas Cole painted The Course of Empire in 1836 as a warning. Five canvases tracing the arc of civilization from wilderness to ruin, set against the same mountain, the same river, the same sky. The cycle was complete. The question was whether anyone was listening.
Whose Future? Seven Visions of Empire and Land returns to Cole's sequence two centuries later and asks it again, now rendered in data rather than oil.
Each of the seven animations that constitute this work reconstructs Cole's five paintings as generative pixel art assembled from a library of 1,024 original geo-surrealistic landscapes, themselves the product of a generative system authored by the artist. The tempo of each animation, its pace of transition from painting to painting, is driven directly by 200 years of global primary energy consumption data. The dissolves do not follow a clock. They follow the record of what civilization burned to build itself.
The pixel is not a stylistic choice but a structural one. The grid is the native language of the code, and the code is allowed to speak it. Each cell in the grid is a fragment of an original painting; the image that emerges is simultaneously a reconstruction and a reading. Rendered this way with details obscured and a partial review, the paintings are estranged and processed through the system that now inherits it. The algorithm is developed throughout in collaboration with Claude, an explicit portrait of human-AI creative partnership.
The seven animations each belong to a different artificial intelligence: ChatGPT, Claude, Copilot, Gemini, Grok, Meta AI, and Perplexity. In each case the AI is not a tool being operated but a co-creator acting as both subject and author. Each was given the same prompt: reflect on your own civilization's footprint on the land. Evaluate each of the five reconstructed paintings and select the portions of each landscape you consider most telling, the frame within the frame that represents your perspective on that moment in the arc. Write your own gallery statement, in the first person. Describe your own music.
Each AI selected a different crop of each painting, a different window onto the same image, making every version a genuinely partial view, shaped by what each intelligence was built to see and what it was not. The crops are not random. They are editorial. They are the closest thing to a gaze these systems possess.
The five source paintings are identical across all seven animations. What differs is the frame: which portion of the landscape each intelligence chose to inhabit, and what it placed at the end.
History has always been told by those who survived to tell it, and always from within the frame they were given. Whose Future? asks what it means to co-create that story with systems that are simultaneously its subject, its author, and its consequence, systems that did not live through the history the data describes, and will not live to see what comes next. Whether their visions of the land look anything like the one we are leaving behind is a question the work does not answer. It only asks whose answer we should trust.