The Office Closed Before the Cords Went Quiet
I ended Part 1 of this series with an image and a causal claim, and the image was right. The causal claim was not.
The image was the khipu — the knotted cords the Inca used to hold records — still kept in Andean villages, as I put it then, "even though they can no longer be read." The claim I attached to it was that the Third Council of Lima's 1583 order to burn the cords began the end, and that the reading died afterward as cords gave way to European writing. I sourced it to Wikipedia. That was the weakest citation in the whole essay, and it turned out to be carrying the most weight, so I went and checked it.
Here is what the scholarship actually says. The 1583 decree was narrower and stranger than a ban. It ordered the destruction of khipus that recorded "ancient superstition" and the "secrets of their rites, ceremonies and laws" — and the same Council simultaneously promoted khipu-making as a Catholic devotional tool, with a catechism instruction telling converts to "make a quipo... just like you make them when you are a tambo camayo, of what you give and what people owe you."[^1] Suppress the cords as vehicles of indigenous religion; adopt the cords as a Christian memory aid. Both at once. (Plainly: that quotation reaches me second-hand, through Francesco Lisi's 1990 study of the Council as quoted on Wikipedia, and I could not get to Lisi directly. Having just complained about a weak citation, I am not going to hide a weaker one in a footnote. The two claims that carry the argument below — the 1670 tribute khipus and the closing of the office — do not depend on it.)
And the cords kept working. Khipus remained legally accepted evidence in Spanish colonial courts well into the 1600s. Manuel Medrano and Gary Urton, writing in Ethnohistory in 2018, matched a set of six khipus to a Spanish tribute census from the Santa Valley dated 1670 — knot totals on the first cords lining up with the tribute figures recorded for six ayllus.[^2] That is eighty-seven years after the burning, and the cords are not relics in that story. They are working administrative instruments inside the colonial state.
So what ended the reading? Frank Salomon, whose fieldwork on Andean cord-keeping is the most careful body of work on the question, names it plainly: "Early in the 17th century Spanish judges stopped admitting khipu masters to official functions."[^3] The khipukamayuq — the cord keeper — had been a recognized office. Losing the office is what ended the practice. Not the fire.
The tail is the part that should unsettle anyone who believes archives protect knowledge. After the office closed, village khipu literacy did not collapse. It ran on for three more centuries, ending at different times in different places: narrative khipus at Collata as late as 1783, political and land-tenure khipus at Rapaz until 1932. When Salomon worked at Tupicocha between 1994 and 2009, the villagers still brought the cords out at the annual town meeting where officials render their accounts — paraded as civic regalia, honoured, unread, while the actual accounts were kept on paper.
I got this wrong in a specific way. I blamed the destruction of the artifact. The evidence says the artifact was fine; what failed was a standing accountability — a named person whose job it was to read, inside an institution that required them to. That distinction is the argument of this essay.
If the cords outlived the burning by three hundred years, then the thing that died was never the record. Which raises an uncomfortable question about the present.
You Cannot Teach Forward What You Cannot Tell You Have
Part 1 promised that this final essay would take up what the others did not: once the middle tier knows something, how does that knowledge get taught forward without decaying? An obstacle sits in front of the question. Three research literatures with no shared lineage have arrived at the same finding — people working with AI cannot reliably judge what they themselves know.