Every inscribed object in the corpus — 2,543 documents. Click one and the grammar that parses 80% of unseen inscriptions prints its functional reading: roles, not words.
WHAT AM I LOOKING AT?
These are real archaeological objects — seals and tags from Mohenjo-daro, Harappa, Lothal and beyond, c. 2600–1900 BC. Nobody can pronounce their signs. But research published in 2026 (MDP Paper V) showed the inscriptions obey a four-rule document grammar: header → content → quantity → closer, like a receipt. The decoder runs those exact rules live. ACCEPTED means the document fits the grammar; FLAGGED means the simple four-rule version can't parse it — mostly long compound documents, the paper's documented ~21% residue. The labels describe each sign's role in the system, not its translation — no one knows what the words sounded like, and this page doesn't pretend to.