Science Was Not Driven by Geniuses. It Was Driven by a Few Boring Rules.

On 28 November 1660, after an astronomy lecture at Gresham College, twelve men agreed on the spot to found a society. Until then the default state of natural knowledge was secrecy: alchemists wrote in code names and allegory, and Newton (1642–1727) left behind over a million words of alchemical manuscript, not one of them published in his lifetime. The Society’s first secretary, Henry Oldenburg (c. 1619–1677), was a German-born diplomatic fixer who lived off a correspondence network spanning Europe.

The Society took a line of Horace as its motto: Nullius in verba — take nobody’s word for it. On 6 March 1665 Oldenburg founded Philosophical Transactions at his own expense, printing abstracts of the letters he received, each stamped with the date of receipt. What it solved was not circulation but property: who got there first would now be settled by a dated public record, rather than by racing to stake a claim in an anagram cipher. Its companion rule was that Boyle’s air-pump experiments required named witnesses present. Secrecy became a losing strategy: what you hide, you cannot later claim.

Commit history in a repository, the arXiv timestamp, a patent’s filing date — all descendants of the same move: shifting priority from secrecy to public registration. That frontier model details have gone back to non-disclosure turns this rule down a notch.

Eighteenth-century chemical names carried provenance, not composition: “sugar of lead,” “butter of arsenic,” “water of paradise” — only an initiate could read off how the thing was made; the name itself was the barrier to entry. Lavoisier (1743–1794) was by trade a tax farmer, funding a private laboratory out of revenue collection; his wife Marie-Anne (1758–1836) translated English literature and drew the experimental plates.

In 1787, with Guyton de Morveau, Berthollet and Fourcroy, he published the Méthode de nomenclature chimique, laying down one rule: a name must be determined by composition — lead sulphate is sulphur, oxygen, lead. The Traité élémentaire de chimie of 1789 then listed thirty-three “elements” — among them caloric and light, both wrong. Knowing the system was incomplete, he fixed the language first: once the vocabulary changed, opponents had to argue in his words, and phlogiston theory would gradually lose the terms in which to state itself. On 8 May 1794 he went to the guillotine as a tax farmer; Lagrange remarked that it took a moment to cut off that head, and a century might not produce another like it.

Every paradigm turn in technology begins with a fight over words — “cloud,” “platform,” “agent.” Whoever defines the vocabulary sets the coordinate system the argument runs in; once rivals start asking questions in your terms, the substantive contest is already settled.

In the early nineteenth century research was still a gentleman’s self-funded pursuit, its craft passed privately from master to pupil. Justus von Liebig (1803–1873) was appointed in 1824, aged twenty-one, to a small university in Hesse, and two years later converted a derelict barracks into a teaching laboratory — with neither funding nor reputation.

He did two interlocking things. First, around 1831 he turned combustion analysis of organic compounds into a standard apparatus, the five-bulb potash apparatus (Kaliapparat): one analysis fell from days to hours, and the result no longer depended on the operator’s hands. Second, he organized students into a production line — each assigned a family of compounds, using one method and one recording format, feeding into the single journal he edited. Once the craft is absorbed into the apparatus, people become interchangeable, trainable units. Over the next two decades Giessen produced hundreds of trained chemists; Hofmann was recruited to head London’s Royal College of Chemistry, and his student Perkin stumbled onto mauveine in 1856 — the origin of a German dye industry that held some eighty percent of the world market by the 1870s. In the same period, at the June 1833 meeting of the British Association in Cambridge, Whewell coined the word scientist in answer to Coleridge’s objection; Britain resisted the term for more than half a century.

Modern doctoral training, and the division of labor inside a large-model team, run on the same machine: a unified toolchain, interchangeable execution units, a shared pool of results. Output becomes predictable; the price is a narrowed problem space, as everyone starts asking the kind of question the tools are good at answering.

The Manhattan Project cost roughly two billion dollars and drew on more than a hundred thousand people; science had become, for the first time, an activity requiring industrial-scale organization. Vannevar Bush (1890–1974) directed the wartime Office of Scientific Research and Development and argued for scientific self-governance; Senator Harley Kilgore (1893–1956) had argued since 1942 that postwar research should be under state control.

In July 1945 Bush submitted Science, the Endless Frontier: an agency governed by scientists themselves, money allocated by peer review, patents left with the contractor. Kilgore’s version was its mirror image: a director appointed by the president to carry political accountability, funds distributed across states rather than concentrated in a few elite universities, and public title to patents from publicly funded work. The disagreement was never about whether to spend, but about who is entitled to define what is worth doing. Truman vetoed the Bush-leaning bill in 1947, precisely because the director would not be presidentially appointed; the compromise National Science Foundation was established on 10 May 1950, far smaller than Bush had envisioned, with the gap filled by the military and the National Institutes of Health — the actual shape of postwar American research is a by-product of that compromise.

Money and compute for frontier AI are concentrating in a handful of corporate laboratories, with open publication and peer review giving way to internal review and delayed disclosure. The 1945 argument over who defines what is worth doing is being replayed by a new cast — except this time the Kilgore side, the case for public accountability, has yet to find its place.