It's been shown in tests that most humans can not from memory draw a functional bicycle given pen and paper.

Everyone knows roughly what a bicycle looks like - wheels, frame, seat, peddles, handlebars etc, but the details of the frame and exactly how the other parts connect to it throws people off. It seems people memorize the "concept" of a frame, but not the specifics.

Try it without cheating, then google a picture of a bicycle!

Even if you have memorized what a bicycle (and pelican) look like, writing code to draw one is not the sort of thing humans are good at any more than they are good at mentally calculating cube roots - better to use a computer for stuff like that.

What do you mean by "oneshot"? The term applied to genai makes sense, but it doesn't make a whole lot of sense to apply the term to human art.

But the LLMs are also not one-shotting it, or are they? I assume they have some ways to verify it, e.g. to visualize it (convert to PNG, then feed as vision tokens back to the LLM), or other ways, maybe also pure text LLMs have some ways to verify the result at least somewhat? And with such feedback loop they can iterate.