The game is defined in a set of text files which are compiled into Javascript. The text files give the content of the scenes of the game. An example:

VIEW( "Start",
	BOW(
		GRID( {:x=>40, :y=>30, :dx1=>0, :dy1=>100},
			BIG_TEXT_BUTTON( "buttons/about.png", "About"),
			BIG_TEXT_BUTTON( "buttons/help.png", "Help1"),
			BIG_TEXT_BUTTON( "buttons/new_game.png", "New game"),
			IF( TEST( "Started"),
				BIG_TEXT_BUTTON( "buttons/decks.png", "Decks")
			),
			IF( AND( TEST( "Started"), TEST( "Complex")),
				BIG_TEXT_BUTTON( "buttons/calculator.png", "Calculator")
			)
		)
	)
)

The above example defines the first screen of the game. The above is actually Ruby code, VIEW and BOW and the other names are Ruby methods. The Ruby data structures are compiled into Javascript which at runtime will add appropriate HTML elements to the DOM for the game.

The visual elements can be defined as layers in a PSD Photoshop file. The compilation process splits up the PSD file into JPGs and PNGs for the web. This feels like a cheap way of laying out scenes to me, to do it inside Photoshop Elements.