The short answer is no, and even if we could, the digit index you'd start at would have a larger binary representation than the actual data you were trying to encode.

An example I found: the string of digits 0123456789 occurs at position 17387594880. In this case, it took 11 digits to describe where to find a 10-digit number.

So I think such an algorithm would technically work, but your "start digit" would be so large it would use more data than just sending the raw file data. Not to mention the impossible amount of computing power needed.

Could we already do this by leveraging the Library of Babel?

Genuinely asking, I'm not really sure.

Similarly: if you write a program to randomly run through all the combinations of pixels on a decently large screen (say, 1080p) you will eventually see every important question and answer that can be expressed on a screen.

https://github.com/philipl/pifs

Conceptually this is basically just standard encryption: some math that spits out gibberish unless you have the info to make that gibberish become something useful.

I think if you can ridiculously compress the size down then maybe lol.