The Civilization Didn't Vanish — Its Readers Did

While building the Lahore–Multan railway in 1856, the Brunton brothers found a mound of ancient brick and quarried it for ballast, laying a hundred kilometres of track bed on it — that was Harappa. In 1921–22 Sahni and Banerji opened Harappa and Mohenjo-daro respectively; on 20 September 1924, John Marshall, Director-General of the Archaeological Survey of India, announced that the Indus valley held an unknown civilization contemporary with Mesopotamia.

What it left behind is deeply anomalous. Roughly four thousand inscribed objects, averaging about five signs, the longest seventeen; no bilingual object, no royal tombs, no king lists. What it did leave is a different category of thing: Mohenjo-daro's city-wide drainage, standardized bricks in a strict 1:2:4 ratio, and chert cube weights distributed from the Punjab to the coast of Oman — base unit about 0.86 g, doubling 1:2:4:8, barely drifting across a thousand kilometres. A state that spent its resources on metrology and sanitation rather than on monuments happened to route around every decipherment tool the nineteenth century possessed, all of which depended on bilinguals and royal names.

Observability design: the metrics you chose to emit determine how deep a postmortem can go. Whatever is not in the logs did not, after the fact, happen.

The German geologist Karl Mauch reached Great Zimbabwe in 1871 and pronounced it the palace of the Queen of Sheba. In the 1890s Cecil Rhodes's company took over the site, and Richard Nicklin Hall, between 1902 and 1904, stripped several feet of deposit wholesale from inside the enclosures under the heading of "clearing later occupation layers."

In 1905 the Egyptologist David Randall-MacIver, sent by the British Association, excavated and published Mediaeval Rhodesia the following year: the building technique and the finds belonged to a local African tradition. In 1929 Gertrude Caton-Thompson re-examined the site with rigorous stratigraphic method and reached the same conclusion — locally built, flourishing roughly in the twelfth to fifteenth centuries. The evidence was already sufficient in 1905, and in the hundred and twenty years since, no contrary empirical finding has appeared. Yet after the unilateral declaration of independence in 1965, the Rhodesian government made local origin a taboo: Peter Garlake, Senior Inspector of Monuments, resigned and left the country in 1970 for holding to that conclusion, and official guides were required to preserve the possibility of "foreign builders."

Log scrubbing in compliance audits: a foregone conclusion can still be corrected, but deleted primary records cannot — no amount of subsequent methodological rigour restores them.

Andrew Sharp's Ancient Voyagers in the Pacific (1956) laid down a hard claim: long-distance settlement of the Pacific could only have been accidental drift, and deliberate navigation was impossible. The anthropologist Ben Finney had been running voyaging-canoe reconstruction experiments since the 1960s and founded the Polynesian Voyaging Society in 1973. The obstacle was that Hawaiʻi's navigational lineage had long since been broken: no living person knew how to do it.

They eventually found Mau Piailug (1932–2010) of Satawal in Micronesia, one of the very few still actively using the etak star-path system. On 1 May 1976 the reconstructed double-hulled canoe Hōkūleʻa left Maui; carrying no instruments, Mau navigated by star positions, the phase relations of swells, and the direction of seabirds returning at dusk, reaching Papeete in Tahiti on 4 June. This did not uncover new evidence; it re-ran a procedure. The cost was not romantic: crew conflict broke out before the return leg and Mau left the ship; in 1978 Hōkūleʻa capsized and crew member Eddie Aikau was lost paddling a surfboard for help. Mau nevertheless went on to teach the whole system to the Hawaiian Nainoa Thompson, breaking Satawal's rule against passing it to outsiders — not passing it on meant losing it entirely.

Knowledge that exists only in execution: however complete the documentation, it is worth less than having someone who can still do it run the legacy migration once more.

On 16 November 1532, Pizarro and some hundred and sixty men seized Atahualpa at Cajamarca. The Inca state had no writing in our sense; its administration ran on the knotted-cord quipu and the hereditary cord-keepers, the khipukamayuq: labour levies, population, storehouses and genealogies were all encoded in the position, direction and colour of knots.

In the early conquest the Spanish used the system: in 1542 Vaca de Castro convened cord-keepers to dictate a chronicle, and in colonial courts of the 1560s Andean lords itemized requisitioned goods from quipus and were believed. The turn came in 1583, when the Third Council of Lima declared quipus idolatrous objects and ordered them destroyed; in the same years Viceroy Toledo's reducción policy forcibly consolidated scattered villages into new towns, severing the hereditary training chain of the cord-keeping families. What burned was cord; what was lost was the reading. Around a thousand quipus survive today, and Gary Urton's database confirms that many encode decimal values; but whether narrative quipus carry non-numerical information, nobody can read — this is the only recording system for which we hold a large body of originals and not a single reader.

Proprietary formats and decommissioned runtimes: the tape is intact and the bits are fine, but the environment and the people who could decode it are gone — which amounts to nothing.