Do HTML demos tell us more than benchmark scores?

Looking through posts about Hy3 after its launch last week, I noticed something interesting.

The posts getting the most attention weren't benchmark charts. They were one-shot HTML demos, ranging from rotating Earth visualizations and Canvas-based physics simulations to complete browser games.

Most of them followed the same constraint: everything had to run as a single HTML file using vanilla JavaScript, with no frameworks or external assets. That meant the model had to handle everything itself.

I've attached a few examples and links I came across. One of the posts even compared the generated HTML and API pricing side by side, and it got a lot of engagement. TBH, I ended up clicking through every one of them.

These aren't the kinds of projects I work on every day, but they got me wondering whether one-shot HTML demos are becoming a useful complement to benchmark scores.

Do these actually tell us something benchmark scores don't, or are they just cool demos?

For anyone interested, here are the original posts: