USP - where a lot of this started

The pattern is visible across this timeline without being stated anywhere on it. A national computing industry needs three things at once - engineers, a reason to employ them, and somewhere they overlap before they have jobs - and a large public university supplies the first and third at no cost to the companies that benefit.

It is also why the market-reserve argument had a technical constituency and not only a political one. The people making the case that Brazil could build its own machines were, in many cases, the people teaching the subject.

The pattern above has a founding instance, and it is the machine on display in the corridor of the Escola Politécnica's directorate. In July 1972 the first computer entirely conceived, designed and built in Brazil was finished there, as the final project of a graduate course in computer architecture, on the school's own budget. It was a metal box a metre square and eighty centimetres deep, weighing a hundred kilograms, and the students had built the hardware, the software and every peripheral interface themselves, because none of it was made in the country. They had to learn to make each item.

The name is the thesis of this catalogue in miniature. Unicamp had announced, with the Navy, that it would build the first Brazilian computer, the White Swan. A master's student at Poli said they were building the Ugly Duckling, but that it would be the first to work. The White Swan was never finished. The catalogue's recurring finding - that the announced first is not the one the record keeps - was true in Brazil before it was true of most of the companies here.

The public demonstration is worth recording exactly as the university's own magazine does. The bishop, the state governor and the rector stood around the machine; as the governor reached for the switch, a photographer tripped over a cable and pulled the plug. The memory emptied and the demonstration did not happen. The first computer built in the country was defeated at its unveiling by the oldest failure in infrastructure, which is somebody standing on the cord.

What followed is the lineage the Cobra entry picks up. The Navy wanted a machine for its programme to nationalise shipboard electronics, and the Patinho Feio became the basis for the G-10, built by Poli with PUC-Rio. The university's administration could not move at the Navy's pace, so the rector created a foundation in December 1972 to run the project outside it - an institutional workaround that became the model for how universities, government and industry would work together here. The G-10 was the prototype for the MC 500, the first Brazilian commercial computer, and Cobra was created in 1974 to build it.

The engineers, the reason, and the overlap: the machine supplied all three at once. The computer engineering course at Poli was consolidated because of it, the mechatronics department grew out of the same group, and the generation that built it went on to found the companies elsewhere in this section. The market reserve that followed had a technical constituency because a room in São Paulo had already shown the argument could be built, if not, on the day, switched on.