AI is a garden

AI doesn’t behave like any other thing I’ve thought of as “technology” before. It behaves like something alive.

If you expect predictable results, you’ll be disappointed. (”Of course we expect predictable results!” scream enterprise IT departments.)

AI pushes back, surprises us, and, if we’re open to it, teaches us things we didn’t expect to learn.

Gardens are like this. They need tending—watering, pruning, observing. They live or die by their inputs.

You Get Out What You Put In

I tried to grow strawberries in sandy soil once. The leaves looked fine. The fruit was deep red. But they tasted like nothing.

I started collecting kitchen scraps in a bowl in my kitchen. Every day, there’s a new pile of spare atoms—carrot tops, a couple of egg shells, coffee grounds wrapped in paper. Instead of discarding them in the green bin, I compost them. And when I add them to my strawberry patch, the flavor is unbelievable.

I’ve got a folder on my desktop called CONTEXT. In it is a growing list of text files, full of relevant information about the projects I’m working on in simple, unformatted text. Any time I need to improve my AI workflow, I have lots of spare bits to throw into the conversation, and the outputs are unbelievable.

Each Season is Different. Each Session is Different.

Gardens humble their creators. Plants sometimes don’t grow where you want them, and do grow where you don’t.

Every gardener knows that the strongest tomatoes are the ones you didn’t plant. They’re the ones that spring up from your compost. You don’t even know what type they’ll be until they fruit.

AI does the same, showing you new approaches to problems you thought you understood. If you walk into an AI workflow expecting the same results every time, you’ll be as disappointed as if you expected identical outcomes each growing season.

AI’s best outputs often emerge from prompts that might drift from your original intent. Stop trying to make AI into predictable software. Start tending it like the living system it is.