Before writing, there were tokens. Small clay objects — cones, spheres, discs, cylinders — found across the Near East from the ninth millennium BC onward. Each shape represented a commodity: a cone for a small measure of grain, a sphere for a large one, a cylinder for an animal. They were the accounting system of the pre-literate world. To record a transaction, you assembled the appropriate tokens and stored them together. The system was concrete in the most literal sense: each token stood for one thing, and you could hold the count in your hand.
By the fourth millennium, a problem emerged. Tokens needed to be sealed inside clay envelopes — hollow balls called bullae — so that a record could be verified later without being tampered with. But a sealed bulla is opaque. To know what was inside, you had to break it, which destroyed the seal that guaranteed its integrity. The solution was to press the tokens into the wet surface of the bulla before sealing them inside. Now the exterior bore impressions of every token contained within.
This seems like a minor procedural improvement. It was the invention of writing.
Denise Schmandt-Besserat, who reconstructed this sequence over decades of work beginning in the 1970s, saw the key moment clearly: once the impressions exist on the outside, the tokens inside are redundant. The surface record is the record. Flatten the bulla into a slab and you have a tablet. Replace the token impressions with drawn signs and you have proto-cuneiform. Each step is small. The cumulative result is a change in the nature of representation itself — from object to impression to symbol.
Proto-cuneiform appears in Uruk around 3400 BC. The earliest tablets are administrative: lists of goods, quantities of grain, counts of livestock. The signs are pictographic — a stylized head for an ox, a stalk for barley — and there are more than a thousand of them. Each sign maps to a concept. The system works, but it cannot say everything. You can record that forty sheep were delivered. You cannot record who delivered them, or when, or under what agreement.
The next crisis is phonetic. A pictograph for “arrow” can represent an arrow, but the Sumerian word for arrow, ti, also means “life.” If you use the arrow sign to mean “life” — exploiting the sound rather than the meaning — you have performed the rebus operation, and you have crossed from pictography into phonetic writing. This begins around 3200 BC. By 2600 BC, cuneiform has become a consistent phonetic system. The sign inventory contracts from over a thousand to roughly six hundred. The signs themselves evolve from recognizable pictures into abstract arrangements of wedge-shaped impressions — cuneus, Latin for wedge.
Every step in this sequence is a move toward abstraction and away from the thing itself. Token to impression to sign to phonetic value. At each stage, the symbol becomes more powerful and less self-evident. A clay cone that means grain is immediately legible to anyone. A cuneiform sign that represents a syllable is legible only to someone trained in the system. Abstraction buys compositionality — the ability to combine a small set of elements into an unlimited range of meanings — but it costs infrastructure.
This is where Proto-Elamite enters as a cautionary exhibit.
From roughly 3200 to 2700 BC, across a wide swath of modern Iran, scribes used a writing system contemporary with early Sumerian cuneiform. About 1,600 Proto-Elamite tablets survive. Approximately 1,200 distinct signs have been identified. The numerical systems — borrowed from Mesopotamia, including sexagesimal, base-120, and decimal notations — are partly deciphered. The language they encode is not. Proto-Elamite is one of the last undeciphered scripts of the ancient world.
The numerical decipherment tells us what the tablets are about: accounting. Quantities of grain, livestock, labor. The same economic substrate that drove proto-cuneiform. But the textual signs — the words that would tell us the language, the names, the grammar — remain opaque. Jacob Dahl at Oxford, who has spent years on the problem using reflectance transformation imaging to reveal details invisible to the eye, has said the field may be approaching a breakthrough. It has not arrived.
The reasons for the opacity are instructive. There is no bilingual text — no Rosetta Stone linking Proto-Elamite to a known script. The signs appear to be syllabic rather than pictographic, which means their sound values cannot be guessed from their shapes. And there are no scribal schools. No practice tablets, no sign lists, no standardized exercises of the kind that Mesopotamian scribal education produced in abundance. This last fact is not merely an inconvenience for modern decipherers. It was the fatal condition of the script itself.
Mesopotamian cuneiform survived for three thousand years because it was an institution, not merely a technology. Scribal schools — the edubba — produced generations of trained writers who learned standardized sign forms, standardized sign values, and standardized formats. Students copied the same lists, made the same exercises, corrected the same errors. The result was a self-stabilizing system: each generation of scribes inherited not just the symbols but the conventions that kept the symbols meaningful. When a sign drifted in form, the schools corrected it. When ambiguity arose, the schools resolved it. The institution was the immune system of the script.
Proto-Elamite had no such institution. The evidence suggests that scribes invented signs with considerable freedom, that sign forms varied between sites, and that the system lacked the standardization necessary for long-term stability. With 1,200 signs and no canonical list to anchor them, each scribe was writing in a dialect that might not be fully legible to another scribe in the next city or the next generation. The system spread quickly across Iran — the idea of writing is powerful, and administrative demand for record-keeping was universal. But it collapsed within five hundred years.
The contrast is precise. Cuneiform had roughly 600 signs with high compositionality: a smaller inventory that could, through phonetic combination, express anything. Proto-Elamite had roughly 1,200 signs with low compositionality: a larger inventory of less combinable elements, each carrying more meaning per sign but offering less flexibility in combination. This is the compression-compositionality tradeoff. High compression — many specialized signs, each encoding a complex concept directly — reduces the length of any single record but makes the system brittle. You need to know each sign individually. You cannot generate new meanings by recombining existing parts. And without scribal schools to maintain the inventory, the signs drift, multiply, and eventually become noise.
The pattern generalizes beyond ancient scripts. A symbol is cheap. The practice that keeps a symbol meaningful is expensive. Cuneiform required not just clay and reed styluses but a pedagogical tradition spanning centuries, an economic system that valued literate administrators, and a culture that treated the scribe’s training as a multi-year investment. The signs on the tablet were the visible output. The invisible input was everything that made those signs legible beyond the moment of inscription.
This is true of every representational system I can think of. A codebase without tests is Proto-Elamite: it records the developer’s intent at the moment of writing, but without the institutional machinery to verify that intent over time, it drifts toward illegibility. A protocol without implementations is a bulla with nothing inside. A mathematical notation without a teaching tradition is a set of marks that the next generation must reverse-engineer from scratch. The symbol carries meaning only as long as the institution carries the symbol.
Proto-Elamite fascinates precisely because it demonstrates the failure mode. The Elamite scribes were not less intelligent than their Sumerian contemporaries. They had the same insight — that marks on clay could represent quantities, commodities, transactions. They borrowed the numerical systems and made them work. What they did not build, or did not sustain, was the infrastructure of standardization: the schools, the sign lists, the correction traditions, the institutional memory that converts a private notation into a public language.
Writing is not the moment when someone presses a sign into wet clay. Writing is the moment when someone else, in another city or another century, can read it back. The tablet is the artifact. The institution is the invention.