No, the article says right there it uses search results (presumably by scraping). ECS used to be WSDL - real easy to drive from eg. PHP back in the day. I wouldn't want to be the poor sod parsing the retail website HTML (although perhaps the last redesign improved it).

I'm actually not parsing the HTML at all. Not really. I'm using PhantomJS, which is a headless web browser. Each time I load a page, I inject jQuery and just do a simple JS command like

$('#whatevertheIDisfortheform').submit();

It's the equivalent of opening up your console on each page and interacting with it via jQuery rather than a mouse and keyboard.

So something has parsed the HTML - but at least you didn't have to reinvent that wheel :)

Still, you might want to look at https://affiliate-program.amazon.co.uk/gp/advertising/api/de... - the Product Advertising API looks like the latest evolution of the E-Commerce Service API and might let you do things like stop the bot from making duplicate similar purchases, or accidentally fulfilling your wishlist.

"Every time I run it, I give it a set budget, say $50. It grabs a random word from the Wordnik API, then runs an Amazon search based on that word."

You're right.