16,777,216: colour as conversation

16,777,216 takes its title from the number of distinct colours available in 24-bit "True Colour" digital space. It is a number so large that it becomes almost abstract: a technical fact that behaves like a metaphor. No artist could meaningfully exhaust it. No eye could hold it all at once. Yet every digital image is drawn from this immense field of possibility, where colour exists before material, before pigment, before touch. This project begins there: with colour as data, selection, potential and pressure. The images appear to be handmade works on paper, suggesting chalk, pastel, ink wash, watercolour, wax crayon, rubbed pigment, scratched marks and deckled cotton rag. Their surfaces seem stained, scuffed, powdered, erased and touched. They have borders, fibres, residues, accidents. They appear to remember a hand. But there is no hand, no paper, no pigment, no studio table beneath them. They are synthetic images of impossible objects: digital hallucinations of colour behaving as if it had weight, dust, drag and consequence.

At first, 16,777,216 seemed as if it might become a project about chromatic range: eighteen specimens drawn from a vast digital spectrum. Cobalt, cyan, indigo, viridian, mint, chartreuse, lemon, apricot, orange, vermilion, oxblood, coral, magenta, lavender, plum, taupe, olive and petrol. But as the work developed, another subject emerged. The project became less about colour alone and more about the way colour was found, tested, revised and held in continuity through conversation. This was not a sequence produced from eighteen isolated prompts. It was made through an evolving exchange with OpenAI's ChatGPT 5.5 + Images 2.0, where language and image occupied the same working space. In my usual process, I write a prompt, send it to an image model, assess the result, and begin again. Each image is a separate negotiation. Here, the prompt became a thread. The model could follow the project as it unfolded: the visual law, the paper surface, the chromatic map, the need for consistency, the need for difference, and the translation of square specimens into a 5:4 hero image. The conversation itself became a form of memory. It became a working surface.

This is what makes the project feel like a discovery. The machine was not simply waiting for commands. It was responsive to context, correction, rhythm and intention. It could continue a visual language without flattening it into repetition, treat each colour as an event rather than a label, and help restore balance when one image weakened the chromatic architecture. That is not the same as pressing a button. It is closer to collaboration, though not collaboration in the sentimental sense. The model does not feel, desire or understand as a person does, but it can participate in structure, extend thought, and make the artist's decisions visible sooner, faster and sometimes more strangely than expected. As artificial intelligence becomes more capable and accessible, the limits of making do not disappear; they change position. The harder question becomes: what is actually worth making? 16,777,216 is therefore both a colour project and a process project. It records a moment when language did not merely describe an image, but shaped a visual system over time. The number in the title promises millions of colours. The project offers only eighteen. That is the point. Art is not abundance; it is selection. From 16,777,216 possible values, a small constellation appears. There are artefacts, yes - but they're not really a problem.