From Steam to SiliconEnglish Deutsch Español 中文 Русский

About the project

The first industrial revolution amplified muscle.
This one amplifies mind.

That sentence has been on this site since the first day. What went unsaid for a long time is that it describes not only what the collection is about, but how it is made. The collection does not depict a bridge between the two revolutions. It is that bridge, and it stands on four piers at once.

  1. The subjects come from the first revolution.Steam, iron, glass, coal, and calculation made of brass. 1,871 works, each its own object with a name and a date.
  2. The making comes from the second.Every image is generated with AI, and the site says so without hedging. The C2PA credentials stay in the files. A work about the revolution in tools that concealed its own tool would be dishonest — and duller.
  3. So does the form.A contract on Base, images and metadata on IPFS, each work singular and transferable. Ownership of a picture is now as much a technical artefact as the picture. A printed volume would have shown the same subjects and said nothing about the present.
  4. And the fourth pier is the one that carries the span.It is below.

Why 1,871. The number was chosen for the year 1871: the founding of the German Empire, the state that made industrialisation in Germany possible. That Charles Babbage died in that same year — on 18 October 1871, the Analytical Engine still unbuilt — was noticed only later, while this page was being written. Both ends of the bridge in one number, half of it by accident.

What this revolution threatens — and the answer to it

The machine that makes these images can do something else too: assert, very convincingly, something that is not so. An invented quotation, a wrong date, a lineage that never existed — in the same even voice as the truth. That is not incidental damage from this technology. It is its characteristic danger.

So the rules of this collection are not arbitrary. They are its argument.

A collection made by an AI that insists on evidence is therefore not neutral towards the technology that produced it. It takes a position — not in a text about that technology, but in the way it is built.

Which puts the fourth pier on the same ground as the first. The industrial revolution was also documented by the people who carried its cost: the trapper boy at the mine door, the picket line, the emigrant quay. Producing evidence was already once the answer to an upheaval moving faster than understanding.

Why the work pages widened it

The pages at /w/ were meant as an addition — a little more information, a few references. They became something else.

While the collection was 1,871 images, it was a list. Once the works point at one another, it is a network — and the network is the actual statement. The Jacquard head makes Babbage's card reader possible; the card reader carries Note G; the card leads to Hollerith; Hollerith to the social security account, where a person suddenly stands behind a number that will outlive them.

A trait generator cannot copy this. Not for want of technique, but because there is nothing to connect between fur colour and hat shape. Here one thing really does follow from another. That is the difference between a collection and an output.

What counts as success

Floor price is the least informative thing you can know about this project.

The better question is whether anyone learned something. The first person that demonstrably happened to is the one making it, every day the research runs. That is not an excuse for absent demand. It is a statement about what is actually being produced here: not 1,871 pictures, but a connected account of how we got from one revolution to the next, with 1,871 pictures hanging from it.

If in ten years someone reads about de Prony's calculating office because they landed on a page with one image and three documented references on it, the bridge held — regardless of who owns the tokens.

People between the machines

Machines do not invent, build or run themselves. Every fortieth work therefore shows no object but a person of power on a dated day with a machine — the Revolutionaries. From the third epoch on, each chapter of forty also holds two works with the people at the machines: researchers, engineers, doctors, the women and men who built, tested and used them. Willem Einthoven and Marie Curie stand in the chapter on medicine, Karel Čapek, who gave the robot its name, and Sally Ride at the robot arm of the Challenger in the chapter on robots.

The same rules apply to them as to the objects. The day, the place and what happened must be documented. Only the named person is shown with a recognisable face, modelled on a photograph in the public domain; where no such photograph exists, the person is seen from behind or only as hands at the machine. Their pages say who the person was, what happened that day, what it cost — and what the sources leave open.

The image correction of Epoch I

In September 2026, after Epoch I had been minted, all 400 images were checked one by one against their motif text. The result was uncomfortable: in 46 works the image did not show the object named beneath it — a propeller where none was allowed, one wheel too many, a design from the wrong decade. In 49 more it was arguable enough not to let it stand.

That set two of this site's rules against each other. The first says every motif has a documented object. The last says nothing is quietly rewritten. An image showing something other than its evidence breaks the first — and leaving it untouched to spare the last would have made the error into the rule. The correction therefore follows not taste but the first rule. Quiet it is not: it stands here.

95 motif texts were rewritten and the images generated again. 89 are now right. Six are not: works 6, 36, 38, 42, 128 and 396 carry a named flaw. With the two Garratt locomotives the image model never once managed the articulated frame; naming the fault was more honest than retouching it away.

On 12 September 2026 the image and metadata addresses of Epoch I on IPFS were replaced. Title, year, object and the links of every work stayed as they were; only the image changed. The old addresses remain reachable.

The value of the inspection does not lie in the corrected images alone. Three things were learned, and they have been in this project's working instructions ever since. First: a negation summons what it forbids — "no propeller above water" produces a propeller. Since then only what should be there is described. Second: a style reference pulls every motif towards itself; the locomotives grew alike as long as one was supplied. Third: the model cannot count — ten wheels, an articulated frame, a closed tube are not instructions that arrive.

What changed was not only the text of the motifs but the order of the work: from Epoch II on, every image is checked against its motif text before it is minted, not after. The inspection itself was joint work — several machine passes over all 400 images, every flagged case looked at by the author. Both sides learned, and both sides made mistakes: the geometric error on the Shukhov tower, work 74, where the railway line runs through the foot of the tower, had been waved through by the machine passes. It was the human who saw it.