Different LLMs and analog clock generation

Some co-workers showed me Brian Moore's AI World Clocks project.

The prompt (“Create HTML/CSS of an analog clock showing ${time}. Include numbers (or numerals) if you wish, and have a CSS animated second hand. Make it responsive and use a white background. Return ONLY the HTML/CSS code with no markdown formatting.”) seems simple enough, but it is fascinating each model is able to 'almost' get it on a consistent basis.

My curiosity piqued, I initially asked my local LLM “write me a program to display a clock in javascript” and and it did, but the clock was a digital one: (source)

So in the same session, I simply added "I want an analog clock" and it gave me a pretty decent one, but with no numbers on the face:(source)

Finally, I wrapped up that LLM session with "Add numbers to the face of the clock" and got something that compared pretty favorably to the best clocks found on Brian Moore's site: (source)

But, why are Brian's clocks so wonky?

That's weird, I thought. It was pretty easy to generate a good clock. Why are most of the ones on the AI world clocks project so wonky?

So I tried his prompt, but it was disappointing.