Home Research Theory AI & PCT Why AI Lies Audit Kit Consulting Cases Blog FAQ About
1,600
Tablets in corpus
c. 3100–2900 BCE
20
Consecutive tablets
audited in detail
1,500
km span
Susa to highland
7
Findings
2 confirmed · 5 new
Contents CorpusHard-coded constantsThe 20-tablet audit Deficit-reset loopShift of authority CollapseStated limitsCite
// corpus

A relational database inscribed in clay

Proto-Elamite, c. 3100–2900 BCE
Corpusc. 1,600 clay tablets
Sign inventoryc. 1,200 unique signs, c. 400 core ideograms
Find contextsLowland Susa plus highland outposts at Tepe Yahya, Sialk, Godin Tepe and Shahr-e Sokhta
Geographic spanover 1,500 km
Audited in detail20 consecutively archived tablets, P008791–P008810
SourcesCDLI transliterations · Damerow & Englund (1989) · Scheil (1935) · Friberg (1978–79) · Meriggi (1971) · Amiet (1972)
Does not attemptDecipherment. The audit asks what signs do, not what they mean

The structure is positional: headers identify institutions, medial signs serve as personal identifiers or project numbers, terminal operators classify the object being counted, and numerical notations close every entry. Four fields, in fixed order, across a network spanning fifteen hundred kilometres.

// two constants

The economy was hard-coded

// the 20-tablet audit

Twenty consecutive records, one procedure

Structural audit of P008791–P008810
Header M15717 of 20 (85%) · exceptions: P008791, P008798, P008808
Deficit sign M00412 tablets
Reconciliation M3418 of those 12, on the reverse
Debts carried forward4 exceptions: P008793, P008796, P008799, P008802
Arithmetic checkP008791: 128 N 01 S at β = 2.5 predicts 51.2; recorded 52; discrepancy 0.8

The 0.8 discrepancy is reported rather than smoothed. The paper reads it as human approximation of a fractional remainder and states openly that whether the capacity system could represent 1/5 as a standard fraction is uncertain.

// the loop

An accounting cycle that closes itself to zero

Across the eight tablets where both signs co-occur, the pattern is consistent: an outstanding balance is recorded on the obverse, and during summation on the reverse the reconciliation operator subtracts it from incoming grain volume — resetting the ledger to zero for the next administrative cycle.

One tablet demonstrates conditional exception handling. The plow sign acts as a logical trigger that switches the metrological system in use, moving from capacity tracking to sexagesimal area counting and outputting 211 discrete units (3 × 60 + 3 × 10 + 1) under a different scheme. An if-then branch, executed in clay, five thousand years ago.

// institutional reset

The seal moves. The authority survives.

Two tablets document what the audit calls a shift of authority. After an account reaches an unrecoverable deficit, cylinder seal PES0329 migrates from institutional header M157 to header M305 — the auditing node retains its physical seal and its operational authority while its ledger is reassigned.

The failure is contained by moving the account, not the auditor. That is a governance mechanism, and it is legible without reading a single word of the language.

A separate finding: cylinder seal Sb 2425, measuring 1.6 × 2.3 cm, carries Indus-style identification adapted into local sealing technology — an administrative hybrid produced where two systems had to interoperate.

// why it ended

A control system with insufficient degrees of freedom

The constants were not adjustable. The seed-to-land ratio was unyielding and the ration constant was fixed. When yields fell under climatic deterioration, the administration had no elasticity: deficits compounded, the reset loop could no longer close accounts, and the network collapsed around 2900–2800 BCE.

In control terms, the error signal exceeded the operational range of the controller. The system did not fail because it was badly run. It failed because it was specified with constants it could not revise — which is the same fault this research programme audits in modern reporting architectures, four and a half thousand years later.

// stated limits

What this audit does not establish

The script remains undeciphered

The underlying language is unknown, no bilingual text exists, and no lexical lists survive — the "dictionaries" that helped crack proto-cuneiform have no Proto-Elamite equivalent. Nothing here changes that.

The ration-constant sample is small

Two direct proofs and one corroborative case. The paper says the sample is small in its own words, and marks the third tablet as consistent with the constant rather than proving it independently.

Functional readings are inferences

The reconciliation sign as an active deficit-reset operator, the shift-of-authority model, the plow sign as an if-then trigger, and the climatic account of the collapse are all inferences drawn by the author from structural co-occurrence — not previously documented definitions.

// access and citation

Open access, permanent identifier

Diener, Ł. (2026). Excel in Clay III: The Susa Protocol — An Algorithmic Audit of the Proto-Elamite Operating System (version 1.0). Zenodo. https://doi.org/10.5281/zenodo.20396089

Preprint, not peer reviewed by a journal. Licensed CC BY 4.0. First published of the three corpus audits, though numbered third: the numbering is thematic and publication order was III, I, II.

Read on Zenodo Excel in Clay I Excel in Clay II All six audits