Designing the rules, then choosing among the results.
In generative design you do not draw the artwork. You describe a process that draws it, then push the process around until you like what it produces. The output is never one image. It is a space of possible images, and your job is to shape that space and then choose from it.
That sounds like handing over control. In practice it moves control somewhere else. You stop deciding where each dot goes and start deciding what makes a dot appear at all, how it relates to its neighbours, and what is allowed to vary between one run and the next.
Rules place the marks. A few lines of logic will produce things nobody would sit down and draw.
Parameters are the dials on those rules: density, scale, how far things are permitted to wander. Most of the design work happens here, and it is slower and fussier than people expect.
Randomness supplies the variation. It is controlled by a seed, so anything you like is recoverable.
Is this the same as AI art? No. There is no model and no training data anywhere in it. A generative system is ordinary code — geometry, noise, a few rules — and it produces the same result every time you hand it the same inputs.
Do I need to code? Not to use any of this. Every system here is exposed as knobs. Knowing roughly what a system is doing does help you steer it, though.
Is the output mine? Yes. No watermark, no account, nothing held back.