They are talking about a creative way to manipulate QR codes so they can have a custom design, like an image or logo, without corrupting the data they contain.

QR codes come in different sizes. By choosing a size larger than necessary for their data, there is extra space left over, called "padding." "ECC" stands for Error Correction Code. QR codes have built-in error correction which allows them to be read even if they're damaged or partially obscured. They are suggesting they can manipulate this ECC data, along with the padding, to make the QR code look a certain way, like having a picture in it. Think of a QR code's data as a collection of tiny boxes, each box being a "bit" that can be either on (1) or off (0). They are talking about breaking down this data into separate groups where only one box is turned on in each group. Without diving deep into mathematics, XOR is a simple way to combine two sets of data (exclusive or). It's a fundamental operation in digital electronics and math. In this context, it's used to mix the original QR code data with the custom design data. By using the XOR operation, they can combine the original QR code data with their custom design data without corrupting the original message. So as long as they're only manipulating the extra space (padding) and not the original data, the QR code's message remains intact.

TL;DR: They found a way to add custom designs to QR codes without messing up the information they contain. They do this by manipulating the extra space and error correction data in larger QR codes. It's a bit like doodling in the margins of a notebook page without writing over the original notes.

I originally learned the technique from here