“The funny thing is that creating a universal algorithm to convert PDFs and/or HTML to plaintext is probably comparable in difficulty to building level 5 self-driving cars, ”

At least :)