The internet has been read. Every model trained today learns from roughly the same corpus — the fraction of human writing that ever made it online.

Most of what we know is still on paper.

University libraries, institutional archives, private collections. Centuries of scholarship, in dozens of languages, that no crawler has ever touched.

We work with the institutions that hold these collections to bring them into machine-readable form — carefully, with consent, and with provenance documented.

For the institutions, it means preservation and a second life for their collections. For research, it means text that exists nowhere else.

What was printed can be preserved.
What is preserved can be learned from.

We handle the rest quietly.