Been on this as well: https://igorkomolov.com/projects/screenshot-and-ai-resistant...

There was a pretty popular video showing this exact concept (hiding shapes/text in high frequency noise ) about a year ago, for those interested:

https://www.youtube.com/watch?v=Bg3RAI8uyVw

Earlier: https://news.ycombinator.com/item?id=48870381

A different link: https://www.mixfont.com/ghost-font

Can AI vision models process "Magic Eye" images? the demo video reminds me of text in SIRDS images.

Please note - I’m the submitter, not the creator