Memory Overload

Core idea

The imagined orders of Building Pyramids hit a ceiling early. A single human brain, no matter how clever, cannot remember the contents of ten thousand grain silos, the names of fifty thousand taxpayers, or the terms of a thousand contracts. Once societies grew beyond a few thousand people, biological memory could no longer hold the information needed to coordinate them. The pressure produced an extraordinary invention: writing.

Writing began modestly — clay tokens for counting sheep, then pictographs pressed into wet clay, then the wedge-shaped marks we call cuneiform. The earliest scripts were partial scripts: they could record numbers, names, and standardised transactions, but not poetry or storytelling. That limitation turned out to be a feature. Partial scripts were the perfect tool for accountants, tax collectors, and clerks. Full scripts capable of capturing all of language came later, but bureaucracy had already taken root in the partial form.

Harari's argument: Writing was not invented to record literature or thought; it was invented to keep books, count taxes, and track who owed what to whom. Bureaucracy is older than poetry.

Why it matters

From memory to filing cabinets

A human brain organises information associatively — by smell, place, emotion, narrative. A grain warehouse needs information organised by sack number, owner, date, weight. The two indexes do not overlap. Writing let humans build a separate, artificial memory whose structure matched the kinds of queries an administrator actually needed to run. That second memory is the foundation of every state ever built.

Why partial scripts mattered first

Cuneiform's early users did not need verbs or pronouns. They needed Nabu-shum-iddin owes 32 measures of barley to the temple of Inanna, due in the seventh month. A partial script with numerals, proper names, and standardised commodity signs handles that perfectly. The leap to full script — capturing the full range of spoken language — came only when scribes wanted to record laws, treaties, and prayers. Administration drove the technology; literature was a downstream beneficiary.

Key takeaways

Mental model

Copy sourceZoom

Practical application

Treat any organisation you encounter as a memory system first and a hierarchy second. Ask: what does this organisation remember, and how? The answer reveals what it can do. A startup with everything in a founder's head can move fast on familiar problems and forget anything outside them. A government agency with strict file conventions can outlast its founders by centuries but struggles to act on anything its forms do not have a column for. Bureaucratic forms — and today, database schemas — are the cuneiform of our era.

A second discipline: when you want to scale anything you care about — a team, a movement, a household routine — invest in scripted memory before you invest in headcount. Untracked knowledge does not survive the first turnover; tracked knowledge compounds.

Example

Picture a Sumerian tax clerk in 3000 BCE. His scrolls cannot capture his wife's name in any meaningful way; the script has no symbol for it. But they can perfectly record that field 14, owned by the temple of Inanna, planted with emmer wheat, owes a tithe of 240 measures due at harvest, payable by household 87. Multiply that one entry by ten thousand fields and you have the working memory of an empire. Now jump 5,000 years to a contemporary tax authority. The scripts have changed — SQL columns instead of cuneiform symbols — but the function is identical: capture standardised, queryable information about transactions between strangers, so that an administrator who knows none of them personally can act on all of them at once. The unbroken thread is one of the longest in human history.

Related lessons