paper-to-game

Jevons’s Noon: design memo

W. Stanley Jevons, The Coal Question (second edition, 1866): the preface, chapter I, chapter VII and chapter XVIII, with the frontispiece. Built under SKILL.md and build.md.

0. Given

Mine. Steffen’s signature and counter (build.md, section 7).

Ideas, from the author.

The text is The Coal Question (1866 edition); the argument is in the preface, chapters I and VII, and the last chapter. No biography. What people know is “the Jevons paradox”, a tidy rule that efficiency raises consumption; what the text does is ask whether a nation should prefer a brief greatness to a long mediocrity, with coal running out either way. The game is about that question, not about climate, and the words stay in the 1860s. The look is Jevons’s own: ruled graph paper, cream with faint blue rules, soot black for everything drawn, one hot colour, furnace orange, for whatever is burning. Curves that climb off the top of the sheet. No worlds, no branches, one sheet. The present enters once, on the last screen: the same curve continued past 1865 on the same paper, with the date, and nothing else said.

What this binds:

1. Paper

W. Stanley Jevons, The Coal Question: An Inquiry Concerning the Progress of the Nation, and the Probable Exhaustion of Our Coal-Mines, 2nd edition, revised (London: Macmillan, 1866); first edition 1865. No DOI. Available: a PDF of the preface, chapters I, VII and XVIII and the frontispiece, from Patrick Draper’s 2009 transcription (CC BY-SA 3.0). The full text can be read at the Online Library of Liberty / Econlib (econlib.org, The Coal Question) and in many scans on archive.org. One number the game uses comes from chapter II, outside the attached chapters (Hull’s estimate of the coal within 4,000 feet); it is flagged where it is used.

2. What the paper hands over

Primary: an argument, with numbers. In a free country the consumption of coal multiplies at a constant rate (3½ per cent a year, the frontispiece); the coal within reach is fixed; exhaustion comes not as an empty cellar but as deeper pits and dearer coal (preface, ch. I); dear coal sends manufactures “to where fuel is cheapest” (preface), so Britain’s lead passes to countries with cheap coal (America, Pennsylvania); economy of fuel does not spare coal, it multiplies its use and is open to rivals as much as to us (ch. VII); bringing a growing nation to a stand is “galling and perilous”, it means “a sore pressure of population and a great exodus of our people” (preface); so the nation must choose “between brief greatness and longer continued mediocrity” (ch. XVIII), and whichever it chooses, the wealth of the day should pay “the moral and the social debts … before the evening” (preface).

Secondary: an estimate, the frontispiece. Past consumption shaded, and the “supposed future consumption of coal at same rate of progress, shewing the impossibility of a long continuance of that progress”: 83.6 million tons in 1861, 117.9 in 1871, 166.3, 234.7, 331, 466, 658, 929, 1,310, 1,848, and 2,607 in 1961. Plus the engine figures of chapter VII (a good engine burns about 4 lb of coal per horse-power per hour; an air-engine has worked at about half; the perfected air-engine would burn “less than one-third of the present consumption”), the duty table (efficiency up ten-fold in under a century), Scottish iron (a third of the coal per ton, ten times the coal), and the preface’s 95 million tons for 1865.

The game formalises the argument as a toy (section 16).

3. The discovery, and its kind

A trade-off, with a counterintuitive beat inside it. The coal in the ground is fixed, so the shape of the nation’s fire is the only thing it chooses: high and short, or low and long. Every way out that the 1860s offered to have both fails as the text says it fails. Better engines make the fire burn faster, not slower. Holding the fire level sends the young over the sea. Rivals with cheap coal pass you whatever you do. The end is not an empty pit but coal too dear to burn. The player is put on the frontier and chooses where to stand on it. The game does not rank the two ends, because Jevons does not: chapter XVIII argues each side in turn and closes on “the momentous choice”.

4. The core narrative

A country has found a black stone that burns, and with it everything gets easier: engines turn, mills spin, ships cross the sea, and every year there is more work and more people. But the stone is buried in the ground, and there is only so much of it. The faster the country burns it, the bigger the country gets, and the sooner it has to dig very deep for the last of it, where it costs so much that the mills move to a country with stone lying near the top. If the country burns it slowly, the stone lasts for ages, but there is less work than there are people, and the young sail away to find it. Clever new engines that need less stone don’t help: they make burning so cheap that people find a hundred new things to burn it for. So the country has to choose: a short, bright fire, or a long, small one.

5. Who lives in the paper

6. Precedent

The three most recent games:

Game Position Format Feedback device Chassis
Augustine’s Clock (augustine) experimenter, with the mind as the instrument discovery by measurement, seven trials two bars laid side by side in memory, appearing when the thing is over Bauhaus poster (Clean): plaster, red, blue, yellow, black; Jost; flat black slabs with a primary disc
Goodall’s Long Grass (goodall) subject: a young chimpanzee learning a skill challenge of mastery against a clock of suns, then one new problem the stem drawn in section in the mound, and Goodall’s age ruler ticked in red pencil graphite field notebook (Drawn): cream, graphite, squared-rule blue, red pencil; Courier Prime + Kalam; pencil-ruled slips
Weil’s Vase (weil) subject: the soul inside a warrior a single life through a seesaw, real time at the top of the swing the player’s own figure glazing to stone, and Zeus’s scales Attic red-figure vase (Worldly): black glaze, Attic clay, dilute glaze, added white, added red; Forum; potsherd buttons

What this game does differently.

7. The aim and the verdict

The aim, as the player reads it: Choose for Britain: a brief greatness, or a long mediocrity.

It is a decision, not a maximum, because Jevons ranks neither (SKILL.md, “The ending”). The yardstick is the sheet: the height of the curve (greatness), how far it runs before it falls (length), and the ground beneath (what is left). It is drawn in the aim’s units at every step.

The verdict leads with the shape, named in Jevons’s words, then three lines:

  1. What you did: “You let the fire double every generation and fitted both new engines.”
  2. What became of it: “Your noon came in 1938, above the top of the sheet, at about 11 times the fire of 1865. By 1945 the pits were at about 2,000 feet and coal about half as dear again as in 1865; the works went where coal was cheap, and the fire fell whatever pace was set. In 1985 it was less than half its noon. America passed Britain in 1928. About 21 million people went over the sea for want of work, nearly all of them after your noon.”
  3. What the paper says: “We have to make the momentous choice between brief greatness and longer continued mediocrity.”

Then Jevons’s sentence for that end of the choice. The shape is named “a brief greatness” when the noon is at least five times the 1865 fire, “a longer continued mediocrity” when it is at most twice, and “between the two” otherwise (XVIII on greatness, “in fearlessly following our instincts of rapid growth …”, or on littleness, “Britain may contract to her former littleness …”). Then the preface’s sentence about noon: “We are now in the full morning of our national prosperity, and are approaching noon. Yet we have hardly begun to pay the moral and the social debts to millions of our countrymen which we must pay before the evening.” Nothing is scored, ranked or called better.

8. Three concepts

A. Jevons’s Noon (agent: the nation).

B. The Damper (controller; the reckless one).

C. The Saved Ton (agent: a ton of coal; the non-obvious one).

9. The first five minutes

A, Jevons’s Noon.

0:00. A cream sheet ruled in faint blue fills the phone. At the bottom left, a black shaded hump of past coal ending at 1865 in a small orange coal; from it a thin dotted line climbs and climbs and leaves the top of the paper, lettered along its length in an engraver’s italic: supposed future consumption … shewing the impossibility of a long continuance of that progress. Title: JEVONS’S NOON. Two lines: Britain burns more coal every year; the coal under it is fixed. You set the pace for each generation. Then the aim on its own line. I press BEGIN.

0:20. The play sheet. Above a ruled double line marked 1850 to 2050: the shaded past and the orange coal at 1865. Below the line, eight bands of black squares labelled 500 to 4,000 feet: “the coal under Britain, each square about 500 million tons.” A thin dashed line far below my coal: America. Under the sheet, a ruler with five marks: cut back, a stand, slower, as now, faster. The text says: set where the fire should be in 1885. I touch “as now”. A dotted line rises from the coal to twice its height at 1885, and five squares at the top of the ground get a pale mark: “these twenty years would take about 5 squares.” I slide to “faster”: seven squares. Back to “as now”. BURN 1865–1885, in orange.

0:45. The pencil is inked year by year: the curve climbs, an orange sliver glows at its tip and turns black behind it, and squares at the top of the ground flash orange and go blank. America’s dashed line rises too. “1865–1885. You set the fire to double, and it doubled. About five squares came out of the ground; the pits are still shallow and coal is no dearer. America burns a third as much as Britain. Nobody had to leave for want of work.” NEXT GENERATION.

1:10. 1885. A new engine: “an air-engine that does the same work on half the coal: 2 lb for each horse-power an hour, where the best engines now burn 4.” FIT IT or REFUSE IT. Half the coal: surely that makes it last. I fit it. Then the ruler: I go for “a stand”, thinking the engine will do the rest. BURN.

1:40. The ink does not follow my level pencil. It climbs above it, to half as high again, and more squares flash than I had marked. “You fitted the new engine and set a stand. Each ton did twice the work, power grew cheap, and new works were built on it, ploughs, pumps and ships; the fire grew by about a third instead of standing still. America fitted it too.” Jevons: It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption. The very contrary is the truth. I laugh.

2:20. 1905. The ground is a fifth gone. I let it run: “as now”. The curve climbs towards the top of the sheet; the ink starts to fall short of the pencil. “Coal from 1,000 feet costs a little more; some works went to America where it is cheap.” America is close behind.

3:00. 1925. The second engine. I refuse it this time, having learnt. America’s dashed line leaps; my ink falls well short; the hatched bar of people starts to grow: “Trade grew slower than the people; about two million went over the sea.” Jevons: To secure a safe smallness we should have to go back, and strangle in their birth those thoughts and inventions …

3:40. 1945. Half the ground is gone. I try “a stand”. The fire falls anyway, the bar of people grows fast, and the game says the evening has come: from here the fire falls whatever pace is set. The rest of the sheet draws itself to 1985.

4:10. The verdict: BETWEEN THE TWO, or A BRIEF GREATNESS, with my noon marked on the sheet, the three lines, Jevons’s two sentences. THE CURVE, CONTINUED.

4:40. The same sheet, my curve gone pale. A black line leaves 1865 and climbs, not off the sheet but to under half its height by 1913, wavers, and falls slowly towards the axis; the ground beneath empties a little, and most of it stays black. At the end of the line, on the axis: 4 November 2026. Nothing else. I sit with it.

B, The Damper.

0:00. The sheet, the ember, no words but the title and “hold to check the fire”. I press START. The curve leaves 1865 at once and climbs, slowly, then faster.

0:10. I hold my thumb on the sheet. The curve bends level; little soot marks fly off the right edge. I don’t know what they are: a legend says “people, over the sea”, but I’m busy. I let go; the curve leaps up.

0:30. The curve is near the top, and the ground is going fast. I hold; the marks pour off. My thumb covers the coal I am trying to watch.

0:50. The curve falls although I’m not holding. I don’t know why: I didn’t see the ground run deep.

1:20. The end card. I understood “growth costs coal; restraint costs people”, but I never chose anything I could put into words, and I could not see the parts move while my thumb was on the glass.

10. The pick, and why

A, Jevons’s Noon. The player sits where chapter XVIII puts the question: as the nation deciding how far to let its commerce rise, with the consequences laid out on Jevons’s own kind of paper. Every generation is a commitment, a pace set before the burn, and the burn shows how the world answered. The fixed ground makes the trade-off visible as area: a high curve eats the squares fast. Each of the 1860s’ ways of having both is something the player can actually try. The engine fails on the sheet. A stand fails in the hatched bar. The rivals pass whatever the player does. The discovery lands inside two minutes (the engine) and the question inside four (the evening). The sentence afterwards has no “guess”: I let it double, fitted the air-engine and watched it climb over my line, refused the second one and lost the works to America, and the evening came in 1945 with half the coal still in the ground but too dear to burn. Originality: no game in the table has had the player draw a time series, or made area the thing spent.

Author test. Steffen knows the paradox and knows the question; what he will not know until he plays is how the two interlock. The engine pushes the curve towards the brief end whatever the pencil says. Restraint has its price in people, not just in glory. And the real curve on the last screen is neither of Jevons’s two answers: it rose less than a third as high as his own projection, held for half a century, and was let down with two-thirds of Hull’s coal still in the ground. An economist will enjoy fitting the air-engine “to make the coal last”.

Runner-up: B, The Damper. Switch if the author wants the boulder in the hand, a ninety-second piece for a lecture, and will give up the engine choice and reading the sheet while it moves.

What the reckless concept would have given. The free nation’s momentum felt as the default, with restraint as a held effort. Kept: the burn plays year by year with its growing additions, and the default pace in Jevons’s own terms (“as now”) sits on the ruler as a mark. Given up: the thumb, real time, and a play screen without words.

11. The shape

One run, at most six turns: 1865, 1885, 1905, 1925, 1945, 1965, each a generation of twenty years, ending in 1985; then the sheet runs on by itself at the last pace set, to 2065, so that a fire that runs out is seen to run out (revision 1). At 1885 and 1925 an engine is offered before the pace is set. If the evening comes first (the trade could not grow even at the hardest push), the game says so and the rest of the sheet draws itself at the last pace set: a brief greatness is also a brief game. No presets: the regimes of the paper (Jevons’s 3½ per cent, a stationary condition, the restrictive party’s check) are marks on the ruler. The discovery comes early (the first engine, at the second turn). No score. Then the verdict, then the last screen.

12. Controls, mechanism, feedback, moment

13. The look

Register: Jevons’s own plate (Instrument family): the frontispiece of The Coal Question, a lithographed statistical chart on ruled paper with hand-engraved italic lettering, a shaded past and a thin line climbing into the future. The game draws its whole world as that chart, live.

14. What is cut

15. Divergences

16. Mechanism kind and claim boundaries

Toy. mechanism.js runs a small model built to illustrate the argument of the preface and chapters I, VII and XVIII, set to the numbers they and the frontispiece give.

On the About screen, in plain words:

The game says “under this game’s assumptions” in spirit throughout: no screen claims that the toy’s numbers are Jevons’s, and the verdict quotes him rather than paraphrasing a finding. Which discoveries come from the paper and which from the rules. The trade-off between height and length follows from the fixed ground, which is the paper’s premise. The rebound’s direction is Jevons’s (ch. VII), and its size is ours. The cost in people of a stand is Jevons’s (preface), and its size is ours. America passing on every path is Jevons’s (“ultimately passing it”). The timing of the evening and the numbers in the verdict are the toy’s. The game’s own choices are the six turns, the two engine years, the evening rule and the shape names. None of them stages a finding. The evening rule only decides when the player stops being asked.

17. Calibration plan and result

calibration.js checks the mechanism against the text, then checks that the discoveries survive changes to our own choices. Result in calibration-result.txt: 159 of 159 checks pass.

  1. The text’s numbers. The frontispiece’s ten-yearly figures (117.9 … 2,607) are reproduced by 3½ per cent from 83.6, each within 1.2%. 1865 = 95 (preface). At that rate from 1865, Hull’s 83,544 million tons last 101 years (“rather more than a century”, ch. I). “As now” doubles the fire in twenty years. Engines: 4 lb, 2 lb, under 1⅓ lb (ch. VII). Price: no rise of importance after the first generation (×1.003); more than three times the 1865 price with nine-tenths gone; beyond any trade near the bottom. One square of ground is 497 million tons, which equals one square of fire.
  2. The last screen’s series. 1900 = 225.2, 1913 = 287.5 (published 287.4), 1920 = 229.5, 1950 = 216.3; the curve peaks in 1913; every year from 1865 to 2024 is present.
  3. The discoveries at the game’s parameters. (a) A faster pace gives a higher noon and more ground gone by 1985, with engines fitted and with them refused. (b) As now with engines, the evening comes inside the sheet and the fire falls. On a stand there is no evening, but the coal still runs out eventually. (c) Fitting an engine burns more coal that generation than refusing it, at all five marks of the ruler and for both engines; and it never makes British power cheaper than American. (d) A stand after two generations of rising sends far more people over the sea than keeping on, and a stand from the start sends people steadily. (e) America passes Britain on every steady path.
  4. Robustness. The same checks (a)–(e) are re-run with each of our choices moved both ways: the pull of cheap power 1.5 and 3.5 (game 2.5); rebound 0.3 and 0.8 (0.5); price steepness 0.6 and 1.8 (1); emigrating share 0.08 and 0.3 (0.15); America 5 and 7 per cent (6); refusal penalty 0.1 and 0.35 (0.2); the people’s growth 0.9 and 1.5 per cent (1.2). Every check passes under every alternative. The sizes on the verdict move; the directions do not.

The pre-1900 values of the real series (1865–1899) are from memory of the standard series (Mitchell) and could not be checked against the DESNZ spreadsheet: assets.publishing.service.gov.uk and econlib.org were not reachable from the build environment. Hull’s 83,544 is from chapter II, not in the chapters given; a French summary of 1873 found online gives “83 milliards de tonnes” for the same limit, and the “rather more than a century” check in chapter I agrees.

18. Playing it in numbers (build.md §3)

From calibration-result.txt, section 4. Nothing is scored; the table is here to show that the decisions matter and that no simple rule gives everything.

strategy shape noon noon, × 1865 evening 1985, × 1865 ground gone by 1985 America passed over the sea (millions) coal lasts at last pace
as now, fit both greatness 1938 10.5 (off sheet) 1945 3.7 79% 1928 21 2036
as now, refuse both greatness 1961 7.5 (off sheet) — 5.9 62% 1903 9 2040
faster, fit both greatness 1929 12.3 (off sheet) 1945 2.8 87% 1933 25 2026
slower, fit both greatness 1950 6.8 (off sheet) — 4.9 57% 1905 12 2062
slower, refuse both between 1985 4.5 — 4.5 33% 1896 3 2117
a stand, fit both mediocrity 1945 1.7 — 1.6 19% 1891 11 2438
a stand, refuse both mediocrity 1865 1.0 — 0.6 10% 1889 16 3414
cut back, then a stand, refuse mediocrity 1865 1.0 — 0.3 5% 1882 18 4925
as now twice, then a stand, fit greatness 1905 5.4 — 2.6 50% 1915 19 2158
a stand three times, then as now, refuse between 1985 4.1 — 4.1 18% 1889 6 2162
as now, fit the air-engine only greatness 1928 8.9 (off sheet) 1965 4.2 74% 1927 18 2040
as now, fit the perfected one only greatness 1945 15.4 (off sheet) 1945 4.0 78% 1903 23 2033

Readings. The fast paths keep America behind longest (to about 1930) and empty the ground. Their exodus comes all at once, after the noon. The stands keep the ground nearly full and lose America by about 1890, and their exodus is steady from the first generation. A stand after a rise (the boulder) loses as many people as the fast paths. Fitting an engine always moves the sheet towards the brief end. The path with the fewest people gone (slower, refusing both engines) is the interior of the frontier. It is legitimately there, and it is not the paper’s finding, so nothing on screen singles it out; at its pace the coal is gone by about 2120. An earlier tuning (dearness as a drag on the growth rate) let the greedy path burn past the whole ground and made a refused engine compound into endless decline. It was replaced by the level formulation above before any screen was built.

19. Build log

  1. Read the preface, ch. I, VII, XVIII and the frontispiece (whose labels were read at 300 dpi: 117.9, 166.3, 234.7, 331, 466, 658, 929, 1310, 1848, 2607). Wrote sections 0–15.
  2. Mechanism, headless (mechanism.js), tuned against §18 before any screen existed. Two reformulations are recorded in §18. The engines were then restructured from a fixed gain to pounds of coal per horse-power for each country, so that refusing the air-engine and fitting the perfected one catches Britain up (4 lb → under 1⅓).
  3. Calibration (calibration.js → calibration-result.txt): 159 of 159.
  4. Loop on text (loop.js → loop-log.txt): six scripted players, every reading and result as the page says it (the words live in words.js, shared by the loop and the page). Fixed from the printout: depths of “about 100 feet” became “the pits are still shallow”; the emigrating share went down from a quarter to one in seven, because ten million in a generation read as absurd; the result’s lead was reworded (“at about twice that of 1865”); the second engine’s card was reworded for both histories.
  5. Skin (template.html + build.py → jevons-noon.html; the build checks that the embedded mechanism is byte for byte mechanism.js). Fixed from screenshots: the plan’s dotted line vanished inside the black when the fire overshot it, so it got a cream halo (the paradox needs it seen); America’s dashed line did the same, and got the same halo; the scale labels moved to the right edge, away from the curve; the verdict’s labels (America, noon, evening) were piling up at the top of a curve that leaves the sheet, so noon is now lettered in cream inside the black column and evening where the curve comes back into view; the evening’s heading was repeated in its first sentence; the date on the last screen sat on the curve and now stands above it on a leader; the cover’s plate title collided with Jevons’s dotted line.
  6. Full session in a browser (session-test.js → session-log.txt, screenshots screenshot-*.png). Chromium at 380 × 780: sheet A (a first-timer: the engine fitted to make the coal last, How to play opened after a burn and closed back to the same place, a stand to the end, the verdict, About, the last screen), sheet B from ANOTHER SHEET (free trade to the last: the evening at 1945, the sheet drawn to 1985), sheet C (a stand throughout, both engines refused). No errors; no horizontal scroll at 380 px on the start, play and verdict screens; BURN is disabled until the ruler is touched, every generation. Session length is an estimate until someone plays it: about four to six minutes for one sheet, plus the last screen.

Self-review from the screenshots (builder’s review only; no stranger has played it).

20. Revision 1: the trade-off, the tops of the curves, and plainer words

Steffen’s feedback on the first build: the tops of the curves were cut off by the sheet, which was not pretty; the trade-off, the thing an economics text must come with, was not visible; the coal reserves were hard to read; and everything needed more explanation, with an eight-year-old in mind.

Where the trade-off was. In the first build the game stopped at 1985, when a fast fire was still three or four times the fire of 1865: on every visible measure the fast path beat the slow one, and the only cost of greatness, its brevity, lay beyond the edge of the sheet. The trade-off is Jevons’s: with a fixed store, a fire that is high is a fire that is short, and a fire that lasts is a fire that is low. Height times length is the coal. Three changes put it on the sheet and in the player’s hand:

  1. The ruler reads both sides. Under each mark, beside the size of the fire (“twice”), the ruler now says how long the coal would last at that pace if it were kept up (“coal lasts to 1970”), computed live from the current fire and the coal left (lastsUntil in the mechanism: Jevons’s own “at same rate of progress” projection). Slide right and the fire gets high and the coal short; slide left and the opposite. The reading under the ruler says the same in a sentence, and the dotted plan now continues faintly beyond the generation to the edge of the sheet, climbing off the top as Jevons’s did.
  2. The sheet runs on. After the sixth generation (or the evening), the sheet draws itself at the last pace set to 2065, eighty more years, so that the fall of a fast fire and the near-empty ground are seen, not read about. The ruler’s last turn says so before the player commits. A stand still burns at 2065 with 142 of 167 squares left; as now with both engines is below the fire of 1865 by about 2050 with 7 squares left.
  3. The verdict reads height and length first. “How high: about 11 times the fire of 1865, at its noon in 1938.” “How long: until about 2050; by 2065 only 7 of the 167 squares were left.” Then what you did, what became of it, Jevons. Nothing is ranked. The About screen has a paragraph on the trade-off.

The tops of the curves. The sheet now grows: when a fire climbs above the paper, rows are added at the top, with a dashed join where the first sheet ended, and the scale labels continue. Jevons’s own curve on the cover still leaves the top, and the player’s dotted projection does too; the solid curve never does. The last screen keeps the first sheet’s height, so the player’s curve there is drawn in ash and may leave the top while the real one stays small.

The reserves. The ground now has an engraved heading on the sheet (“COAL STILL IN THE GROUND: 167 SQUARES, 8 LAYERS DEEP”), as do the fire (“BRITAIN’S FIRE: COAL BURNT EACH YEAR”) and the people. The squares are exactly 500 million tons each, 167 of them (Hull’s 83,544 million), and the burnt coal above the line is filled with the same ruled squares, so that a square above is visibly the same thing as a square below. A “digging here” label marks the dig face. The result after each burn says how many squares are left of the 167.

Plainer words. Every text was rewritten in short sentences: the cover (“Burn it fast and it makes a big fire that runs out. Burn it slow and it makes a small fire that lasts.”), the key under the sheet, the ruler’s panel (“The top row says how big the fire gets; the bottom row says how long the coal would last”), the results (“Why higher? Each ton of coal did more work, so power was cheap. Cheap power paid for new works: ploughs, pumps, mills, ships. They burnt more coal, not less.”), the engine card (“To do one horse’s work for an hour”), and How to play.

Mechanism. Unchanged in its rules; RUN_TO: 2065 and playerDone added, squares made exact, lastsUntil added. The people bar is now 4 million a square, because the exodus to 2065 runs to 40–90 million on a toy population that keeps growing at 1.2 per cent a year for two centuries; the About screen says the numbers are the toy’s. Calibration: 179 of 179 checks pass, including the ruler’s own projections (“as now” from 1865 lasts to 1966, chapter I’s “rather more than a century”), the longer horizon, and the same robustness sweep; two alternative settings (a steeper price, a stronger pull of cheap power) have the fast fire falling more slowly, so check (b) now asks for the fall and the near-empty ground rather than the year it drops below 1865’s. A scripted browser session played three sheets through the run-on and the last screen: no errors, no horizontal scroll.

21. Revision 2: one production engine, engines that come when they pay, and a ledger of the generations

Steffen’s questions after revision 1: why would a nation ever refuse a new engine (it wouldn’t: the choice was a quiz with one answer), and shouldn’t the game measure what a chosen path meant for each generation? His answer to both: put a simple engine underneath that computes income and income per head (what is now called GDP), and read everything else off it.

The production engine (mechanism.js, all of it “our choice”, with Jevons giving the direction and not a function):

What the engine does to the play. The price is flat while the shallow coal lasts (“no recent rise of importance”), so the air-engine reaches Britain and America in the same year on every path (1912–13). The perfected engine reaches a fast burner earlier (1939–44 against 1953), because its coal is dear by then; that is the fast burner’s temporary edge, and it hastens the end (“the end would only thus be hastened”). The noon of “as now” is now about twenty times the fire of 1865, in 1952, because Jevons’s rate kept up for eighty years, with the rebound of two engines, is that high; the sheet grows to hold it. A first build of this revision let works move linearly and a one-year lead in engines made the fire jump five-fold for a year; the move is now proportional, and burning is capped at what is left in the ground.

The ledger. Eleven boxes under the sheet, live: one per generation from 1865–85 to 2045–65, each with a bar and figure for income per head and a small figure for the whole income, both against 1865, the 1865 level as a dashed rule; the squares each generation burnt are visible on the sheet above it; the squares left for after 2065 are said in words. The boxes from 1985 on are hatched and carry the line “supposing, as Jevons did, that no substitute for coal is found” (ch. I; Tyndall’s letter in the preface). The verdict adds “Who got the coal”: the best-off and worst-off generations, the last generation’s income per head and the days fuel took, and what is left. Nothing is ranked.

The numbers it gives (calibration-result.txt, section 4; income per head against 1865, per generation from 1865):

path ledger left
as now 1.2 1.6 3.3 5.3 6.5 2.1 1.3 1.0 0.8 0.6 5
faster 1.3 2.0 4.7 7.4 4.6 1.0 0.8 0.8 0.6 0.5 2
slower 1.1 1.2 1.9 2.9 4.0 3.7 2.4 1.6 1.1 0.8 24
a stand 0.9 0.8 1.0 1.3 1.4 1.4 1.2 1.0 0.9 0.7 116
a stand three times, then as now 0.9 0.8 1.0 1.6 2.9 4.2 4.1 2.8 1.7 1.1 18
as now three times, then cut back 1.2 1.6 3.4 3.4 2.7 1.3 0.7 0.4 0.2 0.2 84

The far generations are poor on every path per head, because the toy’s people keep multiplying for two centuries (142–203 million at home in 2065); what differs between paths for them is the whole income (×2.8 on “as now” against ×4.7 on a stand) and what is left in the ground (5 against 116 squares). The memo and the About screen say that the far generations’ per-head figures rest on the population toy most of all.

Calibration. 412 of 412 checks pass: the text’s numbers as before; the engine’s anchors (income and income per head 1 in 1865; 18.25 days; an engine that doubles what a ton does cuts the days by 2^0.7); America’s fitting years; the ledger summing to the ground on every path; and the discoveries (a)–(g): the trade-off monotone in the pace (noon up across cut, stand, slower, as now, with “faster” at least 0.85 of “as now”, since dear coal bites sooner; ground gone strictly up; no path burns more than the ground holds); brief or long; economy not saving (with the engines, more coal is burnt by 2065 than without them at every pace, and British power is never cheaper than American in 2065); the boulder; America passing; the engines fitted when they pay, Britain never later than America and earlier on the fast path for the perfected one; the fast path’s best generation more than twice as well off per head as the stand’s best, the stand leaving far more for after 2065, the fast path’s last generation far worse off than its best with fuel taking more than twice 1865’s days. The robustness sweep now moves twenty settings both ways, including the engine’s two anchors, the people’s elasticity, and the speed at which works move.

Build. Engine cards became announcements (one button, “Noted. Set the fire”); the result after a burn reports the engines fitted, the days fuel took and the income per head and whole; the help and About screens explain the price in days, the engines paying, the ledger and the future flag. The browser session (three sheets, the help opened mid-play, the run-on, the last screen): no errors, no horizontal scroll at 380 px.

22. Revision 3: ten generations, a box for after, and a score

Steffen’s notes on revision 2: the sheet’s run-on to 2065 replayed the same animation (a bug: the animation took its start from the end of the full history rather than from the generation’s own year); the run-on itself, a projection at the last pace, could go; and the game needed a final verdict, a real score fitting an economics game: average income per head achieved, over what could have been achieved.

Ten generations. The player now sets every generation from 1865 to 2045, ten turns; the sheet still reaches 2065, every generation scored is the player’s, and after the evening the rest is drawn at the last pace set as before. The strip above the sheet shows the ten years. The real-curve screen stays.

What “could have been achieved” turned out to mean. Three searches over the paths (optimum.js, a grid over the ruler’s five marks for the first five generations, then coordinate ascent over all ten in finer steps) decided the score’s definition:

  1. Average income per head of the people at home is maximised by cutting back hard in 1865 and then pushing faster than ever: thinning the people first raises income per head later. A score that rewards sending people away is not one Jevons would sign. So heads are now the people Britain would have had at the census rate, the gone counted in (naturalPop in the mechanism), in the ledger and in the score; the people at home still decide the fire’s growth and the exodus.
  2. With that denominator, the plain average to 2065 is maximised by burning everything: “as now” scores 98, “slower” 90, a stand 41. The coal left after 2065 counts for nothing in it, and the late poverty of a fast path is only averaged in. True of a finite horizon, and not what the game is for.
  3. The typical generation, the geometric mean of income per head over the ten generations and a box for after, gives an interior optimum: push early (×2.4), hold for three generations, push again towards the end. This is the planner who counts every generation the same, discounts nothing, and lets a doubling count the same for a poor generation as for a rich one (log utility), the standard of Ramsey and Stern. The box after values the coal left in 2065 as if burnt evenly over ten more generations, with the engines of 2065, for the people of 2065 (afterBox); without it the planner burns the last squares in the last generation, the finite-horizon artefact. The box is the ledger’s eleventh cell, with its bar and the squares left.

The score is yours ÷ the best path found, out of 100, with the figure and two sentences on the verdict, above “How high” and “How long”. The best path is searched by optimum.js at build time from mechanism.js, written to optimum.json, embedded in the page as OPTIMUM, and checked by calibration.js against a fresh search (within 0.5 per cent). At the game’s parameters: best typical ×1.52 on 2.4 1.5 1 1 1.25 1.25 1.75 2.4 2 2, leaving 19 squares; the steady paths score slower 90, as now 70, a stand 56, faster 52, cut back 36. The About screen says the score is ours, the one place the game ranks what Jevons would not, and that a planner who weighed the present or the far generations differently would rank otherwise.

Words on screen. “Income per head” stays the phrase; the help and About say whose heads. The score’s explanation, for a child: “Count every generation the same, the one after 2065 too, and let a doubling count the same for a poor generation as for a rich one. That gives the typical generation. On your sheet it had ×1.3 the income per head of 1865 (the ledger’s figures, taken together). On the best path we could find it has ×1.5. Your score is the one against the other.” (The first wording compared the two in words, “the same as the income per head of 1865 … half as much again as it”, which Steffen found incomprehensible; the ×-figures are the ledger’s own.) The ledger’s caption names the box after.

Calibration. The checks of revision 2 (the text’s numbers, the engine’s anchors, the discoveries (a)–(g) with a twenty-two-setting robustness sweep, now including the box’s horizon) plus: heads at the census rate (30 million in 1865, 327 million in 2065); the box after (a stand leaves the next ten generations more than three times what “as now” does); the score’s order over the steady paths under every alternative setting (the interior beats both corners: slower above as now and above a stand, and a stand above cutting back; at the game’s own parameters the full order is slower 90, as now 70, a stand 56, faster 52, cut back 36); and the best path: reproduced by a fresh search, interior, scoring 100, and leaving coal for after. The browser session plays three sheets of ten turns (one meeting the evening), the help mid-play, the verdict with its score, About and the last screen.

23. Revision 4: the score explained, the best path shown, and a coda on the engines Jevons could not see

Steffen on revision 3: the score’s explanation was opaque (“what is doubling? what do poor and rich mean here?”); it needed a much better explanation, perhaps with the best path; the best path is driven by the arrival of the engines and that should be flagged; a coal policy is always a function of the technologies in view, which suggests a coda about the world reached in the 1960s with nuclear power and the world of today, in which wind and sun are not only a means against climate change but the cheapest option; perhaps a final picture. And: earlier instructions (the Given of §0) should not bind these final tweaks.

The score, explained as an even share. The number is now introduced as what it is, before any rule: the income per head which, given alike to every generation from 1865 to the ones after 2065, would be worth as much as the player’s sheet (the equally-distributed equivalent of log utility). Then the rule as one example a child can follow: ×4 for one generation and ×1 for the next is worth no more than ×2 for both. Then what pulled this player down: the two poorest boxes of their ledger, and, where it applies, why their richest generation did not lift them more. Then the best path in the ruler’s own words, what it does and why, and how its range compares with theirs. Then the flag: the best path is a plan for the engines; without them the best even share is ×0.9 instead of ×1.5.

The best path is now searched over the ruler’s five marks only (optimum.js: a grid over the first five generations, then coordinate ascent from the ten best and from every steady path, because there are local optima: a single start found ×1.505). It reaches the same ×1.523 as the finer search, so the best path can be stated exactly and followed: faster, then slower, then a stand three times, then slower twice, then faster twice, then as now. It leaves 21 squares.

What drives the best path (checked before building): the engines set its level (×0.86 without them, ×1.52 with them) and its timing (a generation earlier, ×1.63 and a noon in 1913; a generation later, the pause and the second push move later). Its shape, hard early, a hold, hard again late, survives without the engines: it comes from the cheap shallow coal at the start and the many people at the end. The verdict and About say both.

On the verdict, the picture: the best path dotted on the player’s sheet with its name; in the ledger, each generation’s black bar (yours) beside a hollow bar (the best path’s), with the best path’s figure as the small one; the player’s even share as an orange line across the ledger and the best path’s as a dashed line. The paragraph “Who got the coal” went: the score now says the same thing better.

The coda. The silent last screen stays as it was, the real curve and the date; after the date, one button, Why it stopped, leads to a last screen titled The engines Jevons could not see. Its picture is the same paper: coal alone to 1965, then all of Britain’s energy stacked by source in tons of coal (coal solid, oil and gas hatched, wood and waste dotted: the things that burn; nuclear in ash and water, wind and sun in the paper’s own blue: the things that do not), from Energy Institute data via Our World in Data, with four dates (1913, the most coal; 1956, nuclear; 1967, gas; 2024, coal power ends), an orange ember where the burning fuels stand today, and the ground beneath emptied by the coal really dug: 71 per cent of Hull’s coal is still in it. The words, five short paragraphs: Tyndall’s letter in the preface (“I see no prospect of any substitute being found for coal”); what the game’s own engine says when a substitute comes (searched again in optimum.js with a floor of twice the income per head of 1865 from the year on: from 1965 the best path is “faster” in nine generations of ten and leaves 2 squares, Jevons’s brief greatness; from 2025 it leaves 7; with none, 21), which is the economist’s backstop result (Nordhaus, 1973); what did come (oil; Calder Hall, 17 October 1956; North Sea gas, March 1967; Ratcliffe-on-Soar closed, 30 September 2024), each checked; the International Energy Agency’s “cheapest electricity in history” (2020), with its condition; and one line: a plan for coal was always a plan for the engines in view.

Calibration 509 of 509: the earlier checks, plus the best path uses only the ruler’s marks; the engines lift the best even share by more than a third; with a substitute from 1965 the best path is “faster” in at least eight generations and leaves under 5 squares; the earlier the substitute, the less coal left and the better the even share; the coda’s data (every year 1965–2024, no negative layer, coal in 1965 and oil and gas in 2024 as published); and the share of Hull’s coal left by the real curve (“about two-thirds”).

Renamed (11 November 2026)

Renamed from Jevons’s Noon to Burn Bright with Jevons, at the author’s decision to let some titles say what the player does while keeping the name. The game file is burn-bright-with-jevons.html; a one-line page at jevons-noon.html sends old links to it. Nothing in the mechanism changed.