My "favorite" part was '"read" this word I have covered with a sticky note'.

You need X-ray vision like Superman to actually read that.

Scratch that, it was recorded video, so even Superman couldn't do that, you'd need a time machine in addition to the X-ray vision.

Well, I was taught using phonics, and then I wrote a paper that basically applied the idea to LLMs for constrained text generation. I'd say mission accomplished.

This is a chance, stay with me here, that humans and LLMs learn differently.