The Ledger of Meluhha: Indus Valley Script as Metrological Accounting Code

Abstract

The Indus Valley script (c. 2600-1900 BCE) is a metrological cargo-tag system, not a phonetic language. Five fields: merchant mark, commodity class, weight tier, quantity multiplier, trade route. Two Mohenjo-daro seals (M-52A, M-148A) decoded end-to-end from real CISI sign-sequence data through the Parpola-Mahadevan concordance into a metrological codebook. Both produce jar goods as commodity -- independently the most frequent sign in the Mahadevan corpus (~10% of all occurrences). M-52A produces a Mesopotamia route marker, consistent with its findspot at the primary IVC export hub. Neither result is circular. 20 morphological parallels between Tamil Nadu Iron Age potsherds and IVC seals transcribed from Rajan & Sivanantham (2025). 18 pending -- Harappa/Kalibangan/Rojdi data not yet in the open CISI corpus. Normalised SQLite corpus database: 2,373 signs from 5 sources, 179 CISI inscriptions, 397 concordance entries, 600 Tamil proxy inscriptions. Schema verified by Alloy 6: 16 assertions, all UNSAT at scope 6, zero counterexamples. Five F# scripts (dotnet fsi): codebook seeder, corpus ingestion, LSSC transition entropy analysis, IVC cargo-tag decoder, Tamil Nadu cross-corpus decoder. Codebook frozen as typed F# records generated once by SqlHydra v4 -- decoder runs with zero database dependency. CMake build system with loud dependency checks. All claims queryable with sqlite3. All computations reproducible with dotnet fsi. The Indus decode pipeline produces its result through five independent data paths: corpus frequency, concordance shape matching, findspot geography, cross-corpus overlap, and physical weight calibration. None shares a common error source. Under generous per-path bounds favouring the null hypothesis, the joint probability that all five convergences are coincidental is 3.1 x 10^-5, one chance in thirty-two thousand. The same Bernoulli independence structure that rules out universal stratigraphic disturbance across Tamil Nadu at 10^-7 rules out coincidental decode at 10^-5. The cargo-tag model is not proven. But the hypothesis that these convergences are accidental requires a one-in-thirty-two-thousand coincidence across data sources that do not talk to each other. Noise scatters. Signal clusters. The decode clusters. Indus script is accounting code operating inside a checkpoint-verified commercial network whose primary fraud-prevention mechanism is physical comparison of seal to cargo at every transit node. This paper is about the origin of incentive-compatible distributed fraud detection with accumulated reputational capital, tokenised in tamper-evident physical medium, at civilisational scale, four thousand years before the earliest known comparable system Interactive dashboard: ledger-of-meluhha.html (single file, drop indus_corpus.db to render trade network, decode seals live, filter routes by commodity). Peer review requested. Keywords: Indus Valley, Harappan script, metrological accounting, Bronze Age trade, Meluhha, CISI, Alloy, SQLite, F#, SqlHydra, cargo tag, Tamil Nadu, Rajan-Sivanantham, formal verification License: CC BY 4.0 Upload files:1. ledger_of_meluhha.pdf (21 pages)2. ledger_of_meluhha_overleaf.zip (tex + citations.lua -- set compiler to LuaLaTeX)3. ledger_of_meluhha_dashboard.zip (html + corpus db)