What's the messiest document you've had to extract data from?

We built Parser because real documents are rarely clean: crooked scans, stamps over totals, handwritten notes, 300-page PDFs with the one number you need on page 212. I'm curious what the worst one you've dealt with was, and what you ended up doing about it (manual entry, OCR plus scripts, an LLM?). If you have a file like that, you can try it on the homepage playground, no account needed. I'd love to hear where it breaks.