It is ironic because the article claims images slow the web; before extending HTML with special elements and DSLs it would seem a simpler step to get rid of useless and meaningless stock images.
You are not thinking far enough: some of those stock images could easily benefit from a unified CDN. And why stop there, a number of "office girl on phone" might even be included in the browser distribution. Now _that_ would be an html tag to speed up the web!