SNEWPapers

This is genuinely the first time anyone has made 250 years of American newspapers actually usable for research, and the full-text extraction across 6 million+ stories is the kind of infrastructure that should have existed decades ago — the OCR quality is nearly perfect, the semantic search actually works, and The Sleuth AI assistant saves you from manually sifting through microfilm-quality page scans. At ~50% below traditional archives with a 20% Product Hunt discount, this is a no-brainer for genealogists and digital humanities researchers who've been suffering through terrible search interfaces. That a team of one built 115,000+ GPU hours worth of extraction infrastructure is either impressive or concerning, but either way, the data itself — can't be found in Google or any LLM because it literally doesn't exist in this form anywhere else. This is a sleeper hit.