That's ASCII text in hexadecimal form. It converts to this:
XML:com.adobe.xmp#####<x:xmpmeta xmlns:x="adobe:ns:meta/" x:xmptk="XMP Core 5.4.0" <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"> <rdf:Description rdf:about="" xmlns:photoshop="http://ns.adobe.com/photoshop/1.0/" xmlns:exif="http://ns.adobe.com/exif/1.0/"> <photoshop:DateCreated>2018-05-02T19:52:05</photoshop:DateCreated> <exif:UserComment>Screenshot</exif:UserComment> </rdf:Description> </rdf:RDF></x:xmpmeta>
I've picked these images apart with a fine tooth comb. Any hidden message must be encrypted into the pixels.
Here they are with the 16-bit channels broken down into upper and lower bytes (left and middle) and the lower 2 bits normalized to a 0-255 range (right).
The bottom half is the alpha channel broken down that same way (It's solid 0xFFFF).
If you look at the histograms on the middle sections (the lower 8 bits) you'll see that there are gaps. This is because the image sensor is only really 12-bit. 8 bits are stored in the MSB and 4 bits (16 values) are stored in the LSB. If there way any steganography afoot it would probably be done by modulating the lowest bits in the pixel values. And result in some of those gaps being filled in. They're not.
After that I decided that if these images contained hidden info then it was not intended for us. And if it's not intended for us then Q would be using some spy-grade stuff cooked up by the NSA that none of use could ever crack in a million years.
First I was converting them to raw RGB and reassembling the interleaved bytes into grayscale images. Picking through them with GIMP and a hex edit was a pain in the ass. So I wrote this little command-line cudgel (PDF related). It's crude, but it works.