Gutenberg's press, and the analogy built on it

Filed as interpretation, and the id says moment, so the disagreement has to be argued rather than quietly resolved. build.py glosses interpretation as "A reading of events that a reading may need to name" — and what a 2026 reading actually meets is never the press. It is the sentence "AI is the printing press," deployed as a premise. That sentence is a reading of current events dressed as a historical fact, it is made by people on both sides of this project's own axis, and the canon's job is to make it nameable and gradeable. The owner's proposal line says so outright: "Cite when: AI is compared to the printing press, which is constantly — the entry exists to make the comparison cost something." An entry whose stated purpose is to price an analogy is an interpretation entry, whatever its id looks like.

moment is the kind this obviously resembles, and it fails on a fact rather than on a technicality. llama-weights-2023 fixed the house test as a date on which something visibly happened in public, and arpanet-1969 had to spend a section on the clause "in public" because it could produce a room, a clock, two named people and a handwritten log. This entry can produce none of those, and the reason is the most important thing in the file: there is no moment. No date, no document, no witness, no announcement, no first day. Section 1 establishes that the year in this file's own id is a commemorative convention with nothing behind it. Filing a non-event as a moment because the id has a year in it would be the exact species of fake precision rule 2 exists to stop.

Why not idea. The serious rival, and it nearly wins. There is a genuine implementable mechanism here — the adjustable hand mould that casts identical type from a matrix, which is what Gutenberg actually contributed and which the Gutenberg-Gesellschaft lists first among "the hand casting instrument, the typesetting box, the angled hook, the printing press, the printing ink and the printer bales." The house test for idea, set by shannon-chess-1950, is that there is something in it you could build. You could build this. But idea is for "a framing later work is built out of", and no AI work is built out of the hand mould. What later work is built out of is the story about the hand mould. The mechanism is not what gets cited, so it is not what the entry is filed as.

Why not prediction, though section 3 grades six claims. The claims graded there are claims about the press made in 2023–2026, not claims made in the 1450s. Nobody in Mainz forecast anything on the record. Grading modern forecasts inside an interpretation is the same move arpanet-1969 made with the 1968 Licklider–Taylor schedule.

descends_from is [], and here that is arithmetic rather than a judgment. At 1440 this becomes the canon's oldest entry by 378 years — the next is frankenstein-1818. Nothing in canon/ precedes it and nothing can. The descent runs entirely forward, to arpanet-1969 and tcp-ip-1983 (the canon's other information-infrastructure entries), to frey-osborne-2013 and simon-1965 (its automation forecasts), and to llama-weights-2023 and algorithmic-monoculture-2022, where the questions this entry raises about who holds a copying technology are argued in the present tense. Those are other entries' files and this job does not write in them.

On the year: field, which is wrong and is staying. build.py reads the first four-digit number, so this file marks the canon's chart at 1440 — a date section 1 spends its length demolishing. Keeping it is deliberate: the id is fixed by the job that commissioned the entry, and a file whose id and year disagreed would be harder to read, not more honest. The defensible date is 22 October 1454. A reader should mentally file the entry there and treat the marker at 1440 as this canon's one deliberately inherited error, labelled as such.

---

The proposal line, corrected

canon/proposals.md reads: "printing-press-1440 — Gutenberg's press, and the fifty years it took to matter. Cite when: AI is compared to the printing press, which is constantly — the entry exists to make the comparison cost something. What actually followed: the scribal profession went first, literacy and the Reformation and a century of religious war all came out of the same device, and the effects nobody could have forecast took two generations to arrive. Anyone invoking it for optimism and anyone invoking it for doom are both entitled to, which is the finding. proposed 2026-08-17."

The cite-when is right, the finding is right, and it is the reason this entry is admitted. Three things in front of it need correcting.

"The fifty years it took to matter" is wrong in both directions, and the correction is the entry's most useful single output. There is no one date at which the press started mattering, because the different things it is credited with arrived on wildly different clocks:

Elapsed times are measured from 22 October 1454, the first firm date, not from the fictitious 1440.

• Prior censorship of print instituted — November 1487. 33 years.
• Book prices down by two-thirds; presses in ~250–280 towns — by 1500. ~46 years.
• Political rupture, the 95 Theses — October 1517. 63 years.
• The casting method published openly — Biringuccio, 1540. 86 years.
• A measurable economic effect — 1500–1600, the following century, and visible only at city level.
• A majority of English men able to sign their names — after about 1600. 150+ years.
• Two-thirds of English men literate — about 1840. ~390 years.
• An aggregate productivity effect — never detected, in five centuries of looking.

Fifty years is roughly right for diffusion and roughly right for price, and wrong by a factor of six for literacy, and wrong in the other direction for regulation, which was quick. It is not merely wrong but unanswerable for aggregate growth, where economists have looked and found nothing at all. A reading that repeats "fifty years" has picked one of nine clocks.

"The scribal profession went first" is the wrong order. Manuscript production rose through the first decades of printing rather than collapsing, and the profession's most famous defence — Johannes Trithemius's De laude scriptorum, written 1492 — was printed, in 1494. Section 3 grades this. The displacement was real and it was neither first nor fast.

"Two generations" is the right order of magnitude for the political effect and too short for everything else, per the table above.

---

What it is

The thing itself, and what was actually new

Printing was not new. Woodblock printing is centuries older; Bi Sheng in Song China made movable type out of fired clay around 1040; Korea was casting movable type in metal by 1234, and the Jikji, printed at Cheongju in July 1377, survives as the oldest extant book printed from metal movable type — seventy-three years before anything comparable in Europe. In 1403 King Taejong of Joseon ordered the casting of roughly 100,000 pieces of type. Section 4 returns to this, because it is the fact that breaks technological determinism.

What Gutenberg contributed was not the idea of printing but a production system in which every part worked: an adjustable hand mould that could cast thousands of identical letters cheaply from a punched matrix, a lead-based type metal that took the cast and survived the press, an oil-based ink that would stick to metal rather than bead off it, and a screw press adapted to bear evenly on a forme. Dittmar's economic reading of this is the one worth keeping, because it names the failure mode: printing "exhibited characteristics associated with o-ring production functions" — "Successful printing required movable type, a suitably oily ink, nontrivial skill in inking and press work, and in the printer a rare combination of mechanical and intellectual skills." Every complement had to be present or the technology produced nothing. The Gutenberg-Gesellschaft makes the same point from the other end, and adds the fact that should stop any casual analogy in its tracks: "Everything worked so well together that the process was hardly improved over the centuries."

The date, which does not exist

There is no evidence for 1440. The chain behind that year is short enough to state completely.

The only pre-1450 document that touches the invention at all is a Strasbourg lawsuit of 1439. Andreas Dritzehn, who had put capital into a venture with Gutenberg, died at Christmas 1438; his brothers Claus and Jerge sued, demanding either repayment or partnership. The venture the court was actually asked about was polished mirrors for pilgrims to the Aachen relics — a business that had gone wrong because the pilgrimage the partners were counting on did not happen in 1439 as they had assumed, but in 1440. That postponement is where the year comes from. The witness statements also mention a wooden press built by the turner Conrad Saspach, the use of lead and "moulds," and a payment of over a hundred guilders to the goldsmith Hans Dünne for making everything "that belongs to the truck" — a phrase that may mean printing and may not. The Gutenberg-Gesellschaft, whose institutional interest runs entirely the other way, states the conclusion plainly: "we have no evidence of this and we can assume that Gutenberg was in Strasbourg only in the experimental phase of his later invention." The same body notes that "we have no clues as to where Johannes Gutenberg was between 1444 and 1447 or what he did during this time."

The first firm date in the history of European printing is 22 October 1454, the date on a 31-line indulgence for the defence of Cyprus, printed at Mainz in a type made for the purpose and issued at Erfurt. The Sibyllenbuch fragment is probably earlier — the ISTC dates it to about 1452–53 on the unfinished state of its type — but "probably earlier on typographic grounds" is not a date. The 42-line Bible followed in 1455.

Independent corroboration from outside the philological quarrel: Dittmar, an economic historian with no stake in Mainz's civic pride, opens his paper with "The first printing press was established in Mainz, Germany, between 1446 and 1450." A four-year bracket, six to ten years after the year in this file's id.

So where did 1440 come from? From anniversaries. The British Library's own survey of the Gutenberg commemorations puts it flatly: the two common Gutenberg anniversary dates, 1400 for the birth and 1440 for the invention, are based on guesswork, the 1440 figure inferred from the 1439 lawsuit's hints, while no evidence of Gutenberg actually printing appears until the early 1450s in Mainz. The guess then hardened by being celebrated: the 1840 quatercentenary produced 143 memorial volumes and a statue of Gutenberg at Strasbourg marking a four-hundredth anniversary of an event nobody could date.

This is not pedantry, and it is the first thing this entry is for. The analogy's rhetorical force depends on a clean origin — a moment, a before and after, the year everything changed. There is no such year. The most-invoked technological turning point in Western history cannot be located to within fifteen years, and its conventional date was fixed by a commemoration committee in the 1840s. Anyone announcing a "Gutenberg moment" is invoking something that did not occur.

What happened to the people who did it

Gutenberg took two loans from the Mainz merchant Johann Fust, 800 guilders in 1450 and 800 more in 1452, at 6%. On 6 November 1455 Fust sued him before the Archbishop of Mainz's court for 2,026 guilders with interest. The record — the Helmasperger Notarial Instrument, drawn up by Ulrich Helmasperger and now at Göttingen — is one of the very few documents that place Gutenberg near printing at all, and it is worth knowing that the word "printing" does not appear in it. Gutenberg lost the equipment. Fust and his technician Peter Schöffer carried on, and on 14 August 1457 issued the Mainz Psalter, the first European book to name its printers and date itself. No book Gutenberg printed carries his name. In January 1465 the Archbishop granted him a pension; he died on 3 February 1468.

The rest of the first generation did no better. Konrad Sweynheym and Arnold Pannartz, who brought printing to Italy, petitioned Pope Sixtus IV in March 1472 through their editor Giovanni Andrea Bussi. The petition listed all 37 editions they had produced since 1465 — 12,475 volumes — and reported that about 10,000 remained unsold, that their capital was entirely tied up in stock, and that they had nothing left to live on. The Pope gave them some benefices, which did not sell books; their last edition was Pliny in May 1473 and the partnership dissolved. Venice, the industry's commercial capital, records the same pattern at scale: of roughly 100 printing houses operating there up to 1490, 23 were still active in the following decade and 10 survived the century. In 1484 a Paduan printer, Ercole Busca, killed himself over his debts.

The first information-technology boom had roughly a 90% failure rate, its inventor lost his company to his lender, and the first firm to put its name on its product was the lender's.

The diffusion, and how much of it was luck

Printing spread from Mainz outward — Bamberg and Strasbourg by about 1459–60, Subiaco and then Rome from 1465, Venice from 1469, Paris 1470, Caxton in England 1476. The traditional accelerant is the sack of Mainz on 27 October 1462, which scattered trained workmen; the dispersal was already under way before it, so the sack is best read as accelerating a diffusion rather than causing it. Dittmar's data record roughly 200 of over 1,000 historic cities getting a press between 1450 and 1500; the Princeton exhibition puts it at "about 240 additional towns" after 1470, in which "printing shops were hopefully installed" though "their lifespans were often short, sometimes amounting to no more than a single edition."

Two features of that spread matter for any modern analogy.

It was substantially random. Dittmar assembles the historians' verdict: a notable fraction of early printers were "nomads, trusting to luck to find a backer who would enable them to settle and establish themselves", and diffusion was "guided more by chance than by any assessment of profitable centers" — after which the early centres held first-mover advantages for centuries. His footnote lists major commercial and financial cities that simply never got a press before 1500: Bari, Bordeaux, Bremen, Dordrecht, Dublin, Königsberg, Livorno, Malaga, Marseilles, Salerno, Santander, Salonika, Toulon. Presses went to 36% of port cities and 43% of non-port cities. Whatever picked the winners, it was not a clean reading of where the demand was.

And the method stayed closed while the product spread. This is Dittmar's sharpest single fact and the one a reading is most likely to need: "the knowledge required to successfully cast movable type remained quasi-proprietary for nearly one century after Gutenberg's innovation: Biringuccio's Pirotechnia (1540) provides the earliest known published blueprint." Books were everywhere by 1500. How to make the thing that made books was not published for another forty years after that.

The numbers by 1500, with their uncertainty attached

The ISTC, begun at the British Library in 1980 and hosted by CERL since 2016, recorded 30,596 editions as of October 2024, of which 27,460 are confirmed fifteenth-century; the estimated true total of surviving incunabula editions is about 28,000. Those are editions that survived, not editions printed. The familiar figure of about 20 million copies before 1500 is Febvre and Martin's, and it is an inference: editions multiplied by an assumed average print run of no more than 500. It should be carried as an order of magnitude and never quoted as a count.

Prices are firmer, because two independent series agree. Dittmar: "Between 1450 and 1500, the price of books fell by two-thirds." Clark's English data show real book prices falling 75% between 1450 and 1530, then sitting at about one-third of the pre-Gutenberg level into the late 1700s. Van Zanden's Dutch data give two-thirds between 1450 and 1500, then a slow drift to between one-third and one-sixth of the pre-Gutenberg level across 1500–1800.

Read that second half again, because it is the part nobody quotes. Between 1500 and 1800 the technology was, in Dittmar's words, "largely unchanged" and the price declines were "relatively modest." The revolution was a step, then three centuries of a plateau.

And what it was still too expensive to reach

A copy of the 42-line Bible is reported to have sold for about 30 florins, roughly three years' wages for a clerk. Books were heavy, fragile and awkward to move: contracts from Lyons and Poitiers put a 360 km journey at a 20% premium on the sale price, and the Ruiz merchant records show insurance and transport on 21 books from Lyon to Medina del Campo — 280 km — running to thirty days' wages for a skilled craftsman. Dittmar's Figure II shows the share of a printing city's editions held in the Munich archive falling sharply and non-linearly with distance from Munich, and language does not explain it: 72% of the books printed 1450–1500 were in Latin, and the pattern holds within the Latin sample.

The consequence is the entry's central economic finding and the reason the Dittmar paper carries so much of this file: the benefits of printing were local. Cities that had a press got cheap information, printers, booksellers, paper mills, translators, schools, and the people those attract. Cities two hundred miles away mostly got what itinerant traders happened to be carrying.

---

Why a reading would cite it

The comparison is made constantly, at the highest level, by people with positions, and it is always doing argumentative work. The canonical instance is dated and on the record: on 16 May 2023, before the Senate Judiciary Subcommittee on Privacy, Technology and the Law, Senator Josh Hawley asked Sam Altman whether AI was more like the printing press or the atom bomb, and Altman answered "We think it can be a printing press moment." Microsoft's president has made the same comparison; Brookings ran Tom Wheeler's "Gutenberg's message to the AI era" on 16 July 2024, arguing from the press that "the most logical approach to getting control of the AI revolution is to focus on its effects" rather than on the technology. By 2026 the counter-genre is well established too — essays titled, flatly, AI Is Not A Printing Press (18 January 2026) and Why AI Is Not the New Printing Press.

The entry exists because the analogy is legitimate and does not deliver what either side wants from it. Print is a real precedent: a general-purpose technology for copying and distributing symbols, adopted commercially, whose downstream effects ran far past anything its builders intended. That is a fair map onto 2026. What it does not supply is a direction. The same device produced the Reformation and the Peasants' War, the scientific correspondence networks and the Malleus Maleficarum, cheap schoolbooks and a century of religious slaughter. The proposal line's finding is exactly right: both sides are entitled to it, which means neither side gets a conclusion from it. A reading that cites the press to settle an argument is misusing this entry. A reading that cites it to show an argument cannot be settled that way is using it correctly.

Five concrete uses, each keyed to a standing lens.

1. Work and the economy — when weak aggregate numbers are read as "AI isn't real." Dittmar's abstract, verbatim: "The printing press was the great innovation in early modern information technology, but economists have found no macroeconomic evidence of its impact." He then draws the parallel himself, in 2011, without being asked: "Economists have found no evidence of the technology's impact in measures of aggregate productivity or per capita income—much as, until the mid-1990s, they found no evidence of productivity gains associated with computer-based information technologies." Clark called the macroeconomic impact "unmeasurably small"; Mokyr agrees the aggregate effects were small.

What Dittmar found instead, when he went down to the city: cities that adopted printing in the 1400s grew 60% faster than comparable cities between 1500 and 1600, with no prior advantage and no evidence that printers were picking winners; instrumenting adoption on distance from Mainz gives a larger estimate still, 0.58 log points, or 78 percentage points, on a strong first stage (F = 82.07). Distance from Mainz predicted nothing about growth before printing and a great deal after it.

So the citable claim is precise, and it cuts both ways honestly: the canonical general-purpose information technology left no fingerprint in the aggregate statistics and an enormous one at the level where it was actually adopted, in the century after it diffused. A reading in 2026 looking at flat national productivity numbers has a documented precedent for "absent in the aggregate, large where deployed" — and equally, a reading tempted to promise imminent macroeconomic transformation has a documented precedent for that promise not showing up in the data for a hundred years, or ever.

2. Concentration — when "it diffused to everyone" is asserted. The outputs diffused; the capability did not. The casting method was quasi-proprietary for nearly a century, with no published blueprint until Biringuccio in 1540. The first regulatory act anywhere was a monopoly grant over the technology itself: on 18 September 1469 Venice gave Johannes of Speyer a five-year exclusive on the entire art of printing in Venice and its dominions. It lapsed by accident — he died within months, and a marginal note on the document reads "the privilege is now invalid, since the master and author died" — after which Venice never granted anything that broad again and issued no further privileges for seventeen years. The first attempt to govern a copying technology was to grant one firm a monopoly on it, and it ended because the grantee died. For a lens that reads open weights against closed APIs, that is a usable precedent in both directions: the capability was held closed for decades by trade secrecy rather than by law, and the one legal enclosure of it collapsed on contact with mortality.

3. Work and the economy again — when displacement timing is argued. The scribes are the standing example and the standing error. Manuscript production rose in the decade before movable type and through its earliest years; some 75,000 of the roughly 130,000 medieval manuscripts surviving in Central European libraries are fifteenth-century. There were no riots, no machine-breaking, no scribal Luddism. The profession's defence was written in 1492 and printed in 1494. And the thing print supposedly delivered to ordinary people — the ability to read — was still absent from most of them a century and a half later: 61% of English men could not write their names in the late sixteenth century, and two-thirds of English men were not literate until around 1840. frey-osborne-2013 and simon-1965 are where this canon keeps its automation forecasts; this entry is the base case underneath both, and what it says is that the occupation, the transition and the population-level capability move on three different clocks and that guessing one from another is how forecasts go wrong.

4. Policy and regulation — when "you can't regulate it, it's already out" is asserted. The record says otherwise, and says it fast. Prior censorship of print was instituted by **Innocent VIII's bull Inter multiplices, November 1487 — thirty-three years after the first dated printed object, and thirteen years before the incunabula period even ends — giving bishops and the Master of the Sacred Palace the right to license printed matter before publication. It was confirmed by Alexander VI in 1501, made conciliar by Leo X's Inter sollicitudines on 4 May 1515 at the Fifth Lateran Council, which required prior examination for anything touching faith or morals, ordered offending works destroyed and excommunicated printers who ignored it, and extended to the Index of 1559. England's answer was structural rather than doctrinal: the Stationers' Company charter of 1557**, which handed a printing monopoly to a guild whose officers could seize and burn books and gaol publishers.

Every one of those acts works on the chokepoint — the printer — and not on the text. There were few presses, they were expensive, they were immobile and their operators were identifiable, so the state licensed the operators. That is the historical shape of information-technology regulation and it is the shape compute-threshold and model-licensing proposals have in 2026. The precedent is not "regulation failed"; it is "regulation was quick, it targeted the chokepoint, it worked for a century and a half in England, and it stopped working when the chokepoint stopped being scarce."

5. The public square and safety — when print is offered as proof that more information is good. The press's first dated commercial product was an indulgence. It printed them in enormous quantity for decades, helping fund the practice that Luther's Theses of October 1517 attacked. Then it printed the attack: within about two weeks the Theses were in print without his permission, three editions by the end of 1517, and by the end of 1522 Luther had written some 160 works published in 828 editions. In the five years 1521–1525 Germany accounted for one in every two books published in Europe, 80% of them in German — against a pre-Luther baseline in which the Empire produced about a quarter of European output, three-quarters of it in Latin. And the same industry had been carrying the Malleus Maleficarum since Peter Drach printed it at Speyer in April 1487, through thirteen or fourteen editions to 1520 and a second wave later, with Innocent VIII's 1484 bull Summis desiderantes affectibus bound into its front matter — attached by the author without explicit permission, precisely so the book would read as papally endorsed.

That last detail is worth a reading on its own. In 1487, within a generation of the technology existing, someone had worked out that a printed document acquires borrowed authority from an authoritative-looking document printed alongside it, and did it deliberately. Fabricated provenance is not a new failure mode of a copying technology. It is the oldest one on the record.

---

What it got right, and what it got wrong

Not required for interpretation, but this is where the entry earns its admission. The claims graded here are modern claims about the press, with the date each was made, because those are the dated claims that exist. Nobody in Mainz made a forecast anyone wrote down.

1. "Within 50 years of Gutenberg's first press, every major European city had printing establishments... more books were printed in that half-century than scribes produced in the previous 1,000 years." (Wheeler, Brookings, 16 July 2024.) Verdict: broadly right, and it is the strongest form of the analogy. Roughly 250–280 towns had presses by 1500; the surviving record is about 28,000 editions; the copy estimate of ~20 million, though it rests on an assumed print run, is the right order and does exceed plausible reconstructions of a millennium of manuscript output. The claim that should be resisted is not this one.

2. "The fifty years it took to matter." (This canon's own proposal line, 17 August 2026.) Verdict: wrong, and wrong in both directions at once, per the table in the correction above. Diffusion and price move on a ~50-year clock; regulation moves faster; politics takes 63 years; literacy takes 150 to 390; and the aggregate economic effect has never been found at all. Fifty years is one of nine answers and is presented as the answer. The canon corrects itself here rather than quietly inheriting the number.

3. "Printing caused the Reformation." (The standard claim, made everywhere; Pettegree's Brand Luther (2015) is the careful version.) Verdict: the association is as large as any in cultural history and the causal arrow is not one-way. Luther's own works were about 20% of total German output over 1518–1527; German presses produced over 7,000 pamphlet editions in 1520–1525, more than double the preceding decade. But Luther also built the industry that carried him — the shift of German printing from a quarter of European output in Latin to half of European output in German happened because there was suddenly a mass vernacular thing to print. Print was necessary and was not sufficient, and the same presses simultaneously carried the Malleus. The honest statement is that print lowered the cost of a schism that the institutions of the day could not otherwise have sustained, which is a different and much more citable claim than "print caused it."

4. "The press democratised knowledge." Verdict: true on a scale of centuries and false on the scale people mean. Prices fell two-thirds in fifty years and then essentially stopped falling for three hundred. The first product cost three years of a clerk's wages. Access was geographically concentrated by transport costs, and the ability to use it — literacy — arrived on a timescale of ten to fifteen generations, unevenly, and mostly in northwest Europe. A reading that cites the press for near-term democratisation of a capability is citing something that took until roughly 1840.

5. "AI is the printing press, therefore [optimism / catastrophe]." (Made continuously since 2023; Altman's "printing press moment" of 16 May 2023 is the dated instance.) Verdict: the premise is defensible and the inference is void. Both conclusions are drawable from the same record, which is what it means for a historical analogy to have no direction. The Reformation, the scientific revolution, cheap schooling, the German Peasants' War of 1524–25 with its 80,000–100,000 dead, the French Wars of Religion with two to four million, the Thirty Years' War with four and a half to eight million and a population decline across the German lands on the order of a third, and the witch persecutions the Malleus equipped — all downstream of the same machine. The entry's finding is that the press is evidence that a copying technology's valence is not set by the technology.

6. The determinist premise underneath all of it. (Implicit in every use of the analogy.) Verdict: refuted, by a controlled comparison history actually ran. Movable metal type existed in Korea from 1234, produced the Jikji in 1377, and was cast by royal order in 100,000 pieces in 1403 — and produced no Reformation, no vernacular pamphlet war, no print-driven merchant class. The usual technical explanation, the size of the character set, is real and is not sufficient, because the decisive difference is institutional: printing there developed inside the royal foundry, the crown kept a monopoly on the technique, non-official printing was suppressed by mandate, and the output served a small aristocratic readership. Same mechanism, different ownership, different world. For a project whose two poles are, in its own words, the same capability under different ownership, this is the most important sentence in the file: the press is not evidence that a copying technology transforms a society. It is evidence that a copying technology plus a particular institutional settlement did, in one place, once — and that where the settlement differed, the identical mechanism sat in a royal workshop for two hundred years and changed nothing.

---

Commonly misused as

Not required for interpretation — the kind that requires it is limit — but this is the entry's second job and arguably its first. The press is made to say six things it does not say, and as with arpanet-1969 the misuses run in opposite directions, which is the tell that the case is being used rather than read.

1. "It's the printing press, so it'll be fine" / "so it'll be catastrophic." Both, from the same record, at the same time. Carry both or neither. An entry that debunked only the convenient direction would be doing politics with a canon, and rule 7 is easier to keep when a file refuses both in one paragraph.

2. "Gutenberg invented printing." He did not. Printing was four hundred years old in East Asia and metal movable type two hundred. What was new was a casting system, and what made it matter was where it landed: a place that already had paper mills, an alphabet of two dozen characters, commercial credit, competing jurisdictions, universities, and no state monopoly on copying. The analogy's usual implicit claim — that a sufficiently good technology makes its own history — is the one thing this case most clearly refutes.

3. "Once loosed, it couldn't be controlled." It was controlled, promptly and for a long time, by licensing the operator rather than the text: Venice 1469, Innocent VIII 1487, Lateran V 1515, the Index of 1559, the Stationers' charter of 1557. The presses were few, costly, immobile and owned by identifiable people. What eventually defeated the licensing regimes was the disappearance of that scarcity, not any property of print itself. A reading that cites the press for "you cannot regulate a copying technology" has the history exactly backwards and would be better citing it for "regulation of copying technologies attaches to whatever is scarce, and stops working when that stops being scarce."

4. "The inventor won." He lost the shop to his lender in 1455 and his name appears on nothing he printed; the first book to name its makers was Fust and Schöffer's Psalter of 14 August 1457. Italy's first printers were begging the Pope by March 1472 with 10,000 of their 12,475 volumes unsold. Nine in ten Venetian printing houses of the 1480s were gone by 1500. If the press is going to be the analogy, then its own record says the transformative technology and the firms that commercialised it are two different questions, and it says so with numbers.

5. "It was one invention, so we know what a one-invention shock looks like." This is the strongest disanalogy and it is almost never stated. Printing was invented once and then stopped: the Gutenberg-Gesellschaft says the process "was hardly improved over the centuries", Dittmar records the technology as "largely unchanged" between 1500 and 1800 with only modest further price declines. Society had three hundred years to arrange itself around a fixed capability. Whatever is happening in 2026, that is not its shape, and an analogy that imports a static technology into an argument about a moving one has smuggled in its conclusion. A reading may cite the press for what a cheap-copying shock does; it should not cite it for how much time there is to adapt.

6. "Print brought fixity, standardisation, and therefore modern science." This is Eisenstein's thesis from The Printing Press as an Agent of Change (1979), and a reading should know it is contested rather than settled. Adrian Johns's The Nature of the Book (1998) argues that fixity is not an inherent property of print but a social achievement — that early modern readers found print unreliable, piratical and corrupt, and that trustworthiness came from conventions that took until the eighteenth century to establish, not from the technology. This canon does not need to adjudicate that dispute and should not try. But the Johns position is directly load-bearing for 2026, because it is the claim that a copying technology does not confer reliability on what it copies — reliability is built afterwards, by institutions, slowly, and can fail. Any reading tempted to argue that scale or standardisation implies trustworthiness is arguing the Eisenstein side of a live historiographical dispute and should say so.

---

Sources

Read directly, at source. One primary was fetched as a PDF and read page by page as rendered images; every Dittmar quotation and figure below is from the document itself. Fetch failures are recorded rather than papered over.

• **Jeremiah E. Dittmar, "Information Technology and Economic Change: The Impact of the Printing Press," The Quarterly Journal of Economics 126(3), August 2011, pp. 1133–1172**, read as page images from the copy hosted at pscourses.ucsd.edu (the UCSD host refused the connection; the identical Boston University copy at people.bu.edu/chamley was retrieved and read). This paper carries most of section 2 and much of section 1. Source for: the abstract in full, including "economists have found no macroeconomic evidence of its impact" and the 60% city-growth figure (p. 1133); the Solow-paradox parallel to computer-based IT and the founding of the first press "between 1446 and 1450" (p. 1133); Clark's "unmeasurably small" and Mokyr on small aggregate effects (pp. 1134–1135); the o-ring characterisation of printing and the required complementary inputs (p. 1157); **the quasi-proprietary century and Biringuccio's Pirotechnia (1540) as the earliest published blueprint** (p. 1140); transport costs, the 20% premium over 360 km and the Ruiz shipment at thirty days' skilled wages (p. 1141); Figure II on archive holdings versus distance and the 72%-Latin control (p. 1141 and n. 17); the price series — two-thirds 1450–1500, Clark's 75% by 1530 and one-third plateau, van Zanden's 1500–1800 drift, and "Between 1500 and 1800, printing technology was largely unchanged" (p. 1142 and n. 19); printers as "nomads, trusting to luck", diffusion "guided more by chance than by any assessment of profitable centers", and the list of major cities that never got a press, with the 36% and 43% port figures (p. 1156 and n. 41); and the instrumental-variable results — Table VI on distance from Mainz before and after, Table VII's 0.58 log points / 78 percentage points and F = 82.07 (pp. 1158–1159).
• Gutenberg-Gesellschaft, "Gutenberg in Strasbourg" and "Gutenberg's invention" (gutenberg-gesellschaft.de). Source for: the 1439 Dritzehn suit and its parties; the pilgrimage-mirror venture and the postponement to 1440; Conrad Saspach's wooden press, the lead and "moulds", and Hans Dünne's hundred-plus guilders for what "belongs to the truck"; the Society's own verdict "we have no evidence of this and we can assume that Gutenberg was in Strasbourg only in the experimental phase of his later invention"; the gap "between 1444 and 1447"; the component list of the invention; and "Everything worked so well together that the process was hardly improved over the centuries."
• Primary Sources on Copyright (1450–1900), commentary on the Venetian privilege of 1469 (copyrighthistory.org, record i_1469). Source for: the grant to Johannes of Speyer on 18 September 1469; its five-year scope over "the entire art of printing" in Venice and its dominions; the marginal note "the privilege is now invalid, since the master and author died"; the seventeen-year gap in further privileges; and the characterisation of it as "the first known record of a print privilege/patent granted by a European government", functioning as a technological patent rather than a copyright.
• **Princeton University Library, Gutenberg and After digital exhibition** (dpul.princeton.edu/gutenberg), features "Epilogue: 1470 & After" and "Fust and Schoeffer in Mainz." Source for: "about 240 additional towns" with presses after 1470 and their short lifespans, "sometimes amounting to no more than a single edition"; about 28,000 additional surviving editions 1470–1500; the consolidation onto about a dozen towns led by Venice and Paris and Mainz's loss of leadership; and the Mainz Psalter of 14 August 1457 as the first European book to name its printers and date itself.
• Tom Wheeler, "Gutenberg's message to the AI era," Brookings Institution, 16 July 2024. Source for the graded claim in section 3.1, for "the most logical approach to getting control of the AI revolution is to focus on its effects", and as a dated instance of the analogy being made by a serious institution.
• **Johannes Trithemius, De laude scriptorum (written 1492, printed 1494)**, read in Dorothea Salo's 2010 translation (misc.dsalo.info/trithemius.html, CC-BY 3.0 US). Source for the durability argument — "The scripture on parchment can persist a thousand years, but on paper, how long will it last?" — and for "Even if all the books in all the world were printed, the devoted scribe should not desist in his work."
• Wikipedia, "Incunabula Short Title Catalogue." Source for the ISTC's 30,596 editions as of October 2024, the 27,460 confirmed fifteenth-century subset, the ~28,000 estimated total, the British Library origin under Lotte Hellinga from 1980 and the CERL hosting from 2016, and the caveat that the ISTC counts editions rather than copies. Tertiary, and flagged as such; the CERL database itself was not queried.

Secondary, via search-result summaries rather than at source, and labelled as such. Everything in this block was read as a summary of a page, not as the page. Where a claim is load-bearing it is marked.

• The British Library European Studies blog, "Gutenberg Anniversaries — not all that they seem?" (August 2020). The central authority for section 1's claim that 1400 and 1440 are guesswork, that 1440 is inferred from the 1439 Strasbourg suit, that no evidence of Gutenberg printing appears until the early 1450s, and for the 1840 quatercentenary's 143 memorial volumes and the Strasbourg statue. Negative finding: the post could not be read at source. blogs.bl.uk 301-redirects it to blogs-archive.bl.uk, which returns HTTP 404, and web.archive.org is not reachable from this environment — the same wall arpanet-1969 hit. The substance is corroborated independently by Dittmar's "between 1446 and 1450" and by the Gutenberg-Gesellschaft's own disclaimers, both read at source, so the finding stands; the wording is secondary.
• The Cyprus indulgence of 22 October 1454 as the earliest firmly dated European printing, and the ISTC's c. 1452–53 dating of the Sibyllenbuch fragment on typographic grounds: from Princeton's digital exhibition catalogue and reference summaries. The Princeton item record itself was not opened.
• The Helmasperger Notarial Instrument (6 November 1455) — the 800-guilder loans of 1450 and 1452 at 6%, the 2,026-guilder claim, Helmasperger's role, the Göttingen provenance, and the fact that printing is nowhere named in the document — from reference summaries. The facsimile at gutenbergdigital.de was not opened and no translation was read at source.
• The 42-line Bible's print run (~180 copies, 135 paper / 45 vellum, ~48–49 surviving), the ~30-florin price against a clerk's ~10 florins a year, Gutenberg's pension of January 1465 and death on 3 February 1468, and Fust and Schöffer's firm from 1457: all secondary. The 30-florin figure in particular is reported as "some early evidence suggests" rather than documented, and should be carried as an order of magnitude.
• Sweynheym and Pannartz's 1472 petition — 37 editions since 1465, 12,475 volumes, ~10,000 unsold, typical editions of 275, the benefices granted, the Pliny of May 1473 as their last — and the Venetian survival figures (~100 houses to 1490, 23 active in the next decade, 10 surviving the century) and Ercole Busca's suicide in 1484: from search summaries of the Rylands incunabula blog, historyofinformation.com and copyright-history commentary. historyofinformation.com renders as an empty page to this environment's fetcher on every attempt, and the Rylands blog and the Library of Congress, Britannica and Morgan Library pages all returned HTTP 403, so none of these figures was verified at a primary. They are consistent across independent summaries but they are not first-hand here.
• Censorship: Innocent VIII's Inter multiplices of November 1487 (incipit "inter multiplices nostrae solicitudinis curas"; a printed copy is digitised at Freiburg as GW M1227410, which is itself the point — the order to license print was distributed by print), Alexander VI's confirmation of 1501, Leo X's Inter sollicitudines of 4 May 1515, the Index of 1559, and the Stationers' charter of 1557 with its seizure and imprisonment powers. From summaries of encyclopaedia and copyright-history sources; no bull was read in the original or in translation.
• **The *Malleus Maleficarum*** — Peter Drach, Speyer, April 1487; approbation dated 19 May 1487; thirteen or fourteen editions to 1520 and a gap thereafter; and Kramer's attachment of Innocent VIII's 1484 Summis desiderantes affectibus to the front matter without explicit permission, to borrow authority. Secondary throughout. The "without explicit permission" claim is load-bearing for the last paragraph of section 2 and rests on a scholarly summary rather than on the scholarship.
• Reformation print figures — Luther at ~20% of German output 1518–1527; over 7,000 pamphlet editions 1520–1525, more than double the preceding decade; ~160 works in 828 editions by end-1522; Germany at one in two European books in 1521–1525 with 80% in German against a prior quarter-share at 75% Latin; the three unattributed 1517 editions of the Theses at ~300 copies each. These derive from Pettegree's Brand Luther (2015) and Edwards's Printing, Propaganda, and Martin Luther (1994) as reported in secondary summaries. Neither book was read.
• Casualty figures — German Peasants' War 1524–25 at 80,000–100,000; French Wars of Religion at two to four million; Thirty Years' War at four and a half to eight million with German population decline estimates ranging from 20% to about 40% depending on the historian. These are contested ranges and are given as ranges deliberately; they are cited to establish that the downstream violence was on the order of millions, not to fix a number.
• East Asian printing — Bi Sheng's clay type c. 1040; Korean metal type from 1234; the Jikji, Cheongju, July 1377; Taejong's ~100,000 pieces cast in 1403; and the institutional explanation (royal foundry, crown monopoly, suppression of non-official printing, aristocratic readership). From summaries of the BnF, UNESCO Courier, Guinness and university exhibition material. The institutional argument is load-bearing for section 3.6 and section 4.2 and was not verified against a specialist history of Korean printing; a future revision should.
• Literacy — 61% of English men unable to write their names in the late sixteenth century; two-thirds of English men literate by ~1840; Allen's urbanisation-based European estimates 1500–1800 and the northwest-European concentration by 1800. Secondary, from Our World in Data and education-history summaries.
• Manuscript production rising through early printing — ~130,000 surviving medieval manuscripts in Central European libraries, ~75,000 of them fifteenth-century — and the absence of scribal riots: secondary.
• **Febvre and Martin, L'apparition du livre (1958), for the ~20 million copies before 1500 and the 30,000–35,000 surviving editions. Known to this entry only through Dittmar's citations of it (read at source) and through secondary summaries. The book was not read**, and the copy estimate is flagged in section 1 as an inference from an assumed print run rather than a count.
• The Eisenstein–Johns dispute — Eisenstein, The Printing Press as an Agent of Change (1979); Adrian Johns, The Nature of the Book (1998). Read via review summaries, principally an H-Net review. Neither book was read, and section 4.6 accordingly states the shape of the dispute and declines to adjudicate it.
• The analogy's modern instances — Altman's "We think it can be a printing press moment" in answer to Senator Hawley, Senate Judiciary Subcommittee on Privacy, Technology and the Law, 16 May 2023; Microsoft's president making the same comparison; and the 2026 counter-genre, including a piece dated 18 January 2026 titled AI Is Not A Printing Press. The hearing quotation is from press summaries of the transcript, not from the transcript, and is short enough that its wording should be checked before a reading quotes it as exact.

A negative finding worth recording for future entries. Six institutional sources that ought to be the best available on this subject — the Library of Congress research guide, Britannica, the Morgan Library, the John Rylands library blog, and every page on historyofinformation.com — were unreachable from this environment, returning HTTP 403 or rendering empty. The strongest material in this file came instead from an economics paper, a German learned society's own site, a copyright-history archive and a university digital exhibition. Where the reference works were needed, the entry is thinner and says so.