Name the questions, keep the word
This guide gives the lessons of the game, each tested against 400 games played by the built-in bot with and without the advice (seeds 9000–9399). The numbers are there so you can judge the advice, not so you can trust it. The accounting puts on one scale what does not belong on one: the elements of a machine, the questions it leaves open, and the warnings of the stories. The guide says so where it matters.
1. Name the questions; do not dismiss them
A workshop that names every question the counter-voices ask, from Avicenna's flying man to Peirce's two impotencies, scores 95 in the median. One that dismisses them all builds a little more, because each dismissal brings Insight (63 against 59 for what became calculable), and scores 49: the questions it waved away come back as hubris. That is the game's thesis in two numbers. What could be written as a rule became calculable; what the counter-voices asked did not go away because no one answered it.
2. Build all five elements
The machine is complete on paper only if every element, Symbol, Combination, Number, Mechanism and Program, stands at 5 or more at the threshold. A workshop that builds only the logical elements never gets there: median 74, the machine on paper in no game at all. One that leans on the machine elements still makes it (92), because the logical stations raise Symbol and Combination by themselves. Mechanism needs Number 1; Program needs Mechanism 2 and Symbol 2: build in that order.
3. Answer Poe
Poe's question, whether a machine can judge where the steps are not fixed, is the only one the game lets you answer, and only with Program 2 and Combination 3 by 1836. An answer is worth twice a name: 95 against 93 for naming it. Poe was right about the Turk, which hid a man, and wrong about machines in general; the game lets you be right about both.
4. Keep the stop-word before the broom
The sorcerer's apprentice floods the hall without Control 2; Rossum's robots rise without Control 3. A workshop that never keeps the stop-word, dreams the self-working tool, forms the golem by letters and then commands the broom, scores 79; leaving the broom in its corner saves a little (82). Two or three points of Control by the 1790s cost little and pay at both stories. Asking whose work the tool does, and keeping the word on the golem's forehead within reach, gives you the first point for free.
5. Every temptation costs about three
Several stations offer a choice that builds a little and adds hubris. Each costs about three points in the median: trusting the zāʾirja's answers (90) instead of following Ibn Khaldūn's critique (95); building Hobbes's artificial animal (92) instead of reckoning with names (95); embracing La Mettrie (92) instead of holding him as a hypothesis (95); leaving the brazen head to the servant (91) instead of keeping the watch yourself (95); forming the golem by letters (91). The stories are consistent: what is made without a check is lost or turns.
6. Butler: read him as satire
'Darwin among the Machines' offers three answers. War on the machines destroys two steps of Mechanism and buys Control (93); watching them evolve adds hubris (93); reading it as the satire it half is, and taking the Insight, scores best (95). Butler himself let the Erewhonians smash their machines and left open whether they were wise.
7. Choices that make no difference
Kircher's explosion can be met by counting everything or by pruning by rules; both come to 95. Frege's concept-script or his sense and reference, al-Khwārizmī's number or his procedure: by the threshold the workshop has caught up either way. Choose what you want to read.
What the accounting cannot decide
The accounting rewards a workshop that names its open questions. A historian would add that naming a question is not answering it, and that most of the questions on your list are still open after 1936, after 1950, and today. The game ends at the threshold because the apparatus does: the papers that come next are not yet in the public domain. What it can show, station by station and with the passage open, is that the questions were on the table before the machine was, and who put them there.
Benchmarks: node mc.js 400 in the repository; the bot and its switches are in game.js. Every number is provisional; the game is a prototype.