The AI drew the garden. Now what?
On the strange new world where clients arrive with a vision already rendered — and what it means to be the person who makes it real, alive, and true to the ground beneath it.
She slid her phone across the kitchen table without saying anything. On the screen was her back garden — or rather, a version of it that no longer looked like her back garden. The tired lawn had become a gravel terrace. The fence that let in no light was gone, replaced by pleached hornbeams in full leaf. There were grasses — feathery, luminous, catching a late-afternoon sun that London produces perhaps four times a year. It was beautiful. It had taken, she told me, about forty seconds to generate.
She looked at me the way people look at a builder when they hand over an architect's drawing. This is what I want. How soon can you start?
I have been thinking about that moment — and the dozen like it I've had this year — for a long time. And I've found, unexpectedly, that a few concepts have helped me understand what is actually happening in that kitchen, and where I stand inside it.
Einstein's theory of relativity dismantled one of the most intuitive assumptions in human thought: that there exists, somewhere, a fixed and neutral point of observation from which reality can be correctly measured. There is no such point. What you perceive — time, distance, even the order in which events occur — depends entirely on where you are standing and how you are moving.
An AI-generated garden render is, in Einstein's terms, an observation from a single fixed point in spacetime. One angle. One light condition. One idealised season. It is a photograph of a garden that does not yet exist — and that, once built, will never look quite like that image again.
The render cannot show you what happens when those pleached hornbeams, planted in London clay with a west-facing aspect, take three years longer to knit than the algorithm assumed. It cannot show you the Knautia that self-seeds into the gravel in year two, in a colour combination nobody planned, and turns out to be the best thing in the garden. It cannot show you the quality of the light in November, when everything the render depended on for its beauty — the full leaf, the long evening sun — is gone, and what remains is either structure or disappointment.
A garden is a four-dimensional object. It exists in space, yes — but also through time. And no algorithm yet built can render time. What my clients are holding on their phones is not a garden. It is a single frame extracted from a film that will run for decades, presented as though it were the whole story.
Lottie Delamain's recent book makes a quieter argument, but perhaps the most important one. A garden's deepest value, she writes, lies not in how it looks but in what it actively does — for the people inside it, for the ecology surrounding it, for the city it is part of.
"Gardens are the vanguard of positive change — modern-day crucibles for ideas and innovation that provide solutions to some of our most persistent problems." — Lottie Delamain, Gardens That Can Save the World
London's domestic gardens, taken together, cover more green space than all the city's public parks combined. Every decision made in one of them is, in aggregate, an environmental act. The AI render optimises for the visual. It cannot optimise for the mycorrhizal network forming silently in the soil through year two. It cannot optimise for the marsh fritillary finding the Succisa pratensis you planted in the back corner.
And it cannot optimise for this: the child who kneels beside a plant in this garden, learns its name, and carries that name — and the memory of that moment — for the rest of their life.
That image, I think, deserves a moment to sit. Because it is the thing that no render, however photorealistic, however instantly generated, can ever promise. A garden that teaches someone something. A garden that is remembered not for how it looked in a photograph, but for what it felt like to be inside it, once, on an ordinary afternoon in an ordinary summer, when something small and alive caught their attention and held it.
So where do I fit?
The AI does something I cannot do: it gives a client a clear, immediate, emotionally engaging image of possibility. The renders my clients arrive with tell me a great deal about what they value — the quality of light they're drawn to, the register of planting that feels right, the relationship between open space and structure they've been reaching for without knowing how to name it. They are, in the best sense, an efficient brief.
But a brief is not a garden. The distance between the image on the screen and the living system in the ground is exactly the space I occupy. It is the space of garden knowledge — what that specific soil will support, what this aspect means across twelve months, which plants will perform here and which will look like the render for one season before quietly declining. It is the space of time: the understanding that the garden worth having is not the one that looks right on day one, but the one that surprises you in year seven.
My conclusion? A garden is a set of relationships. And relationships cannot be generated. They can only be cultivated — with knowledge, with patience, with honest attention to a specific piece of ground.
The AI drew the garden. I'll build you the one that lives.
