Step 1: Download text file. Remove leading spaces. Remove all blank lines. Remove intro Gutenberg info. Concat all lines. Force line breaks at every period.

Sample text using the Linux shuf command (which just pulls sentences out at random):

  Fig.
  [Illustration: FIG.
 3.
 19 we show diagrammatically the ordinary form of magneto machine.
] If the speed of the engine is sufficiently high, the arm L is thrust forward at such a rate that the weight W tends to lag behind, with the result that P is raised above the notch in B, as shown by the dotted lines in drawing.
 Then, again, we can pump oil through a spraying nipple into the vapouriser (which is kept at a suitable temperature) whilst the cylinder is being filled with air on the suction stroke.
  (3) Maximum expansion.
 18; (2) by keeping the supply of air constant, and varying the amount of gas admitted through nipple N.
1.
 In the reservoir R is fitted an overflow pipe, so that the oil cannot rise beyond a certain level; hence the head of oil in the smaller one M is always constant.

Some goals:

• Use ancient book font.
• Add in obscure diagrams.
• Remove number lines?

Kind of a great preface. Must be re-worked somehow.

I'm thinking the formula can be:

1. Preface (with randomly named widget machine subbing in for "gas engine, or an oil engine")
2. Then add some number of pulled sentences: shuf input.txt -n <number of lines>
3. Add in some diagrams (generative?)
4. Do this until wordcount > 50k.
5. Profit

looking for suggestions of how to generate diagrams or funny pictures. ideally, all images of obscure machines.

Example output now. Should definitely add diagrams, and clean up design and maybe placement of illustrations.

My object in placing this handbook before the reader is to provide them with a simple and straightforward explanation of how and why magnetic tape data storage work. The main features and peculiarities in the construction of these magnetic tape data storage are described, while the methods and precautions necessary to arrive at desirable results are detailed as fully as the limited space permits. I have aimed at supplying just that information which my experience shows is most needed by the user and by the amateur builder of small power magnetic tape data storage. In place of giving a mere list of common magnetic tape data storage troubles and their remedies, I have thought it better to endeavour to explain thoroughly the fundamental principles and essentials of good running, so that should any difficulty arise, the magnetic tape data storage attendant will be able to reason out for themself the cause of the trouble, and will thus know the proper remedy to apply. This will give them a command over their magnetic tape data storage which should render them equal to any emergency.

Reference to the various diagrams in the text will help considerably, and make it an easy matter for any reader hitherto totally unacquainted with such engines to see why and how they work.
It will be seen that the suction of the pump will draw the oil up, the small and lower ball valve, of course, allowing it to pass freely.
Barsanti's engine never became a commercial article; while Otto & Langen's firm, it is said, held their own for ten years, and turned out about 4000 engines.
This shield keeps all draughts and puffs of wind from the fly-wheel away from the aperture, and helps the flame to burn very steadily.
Therefore, to obtain a definite opening we must set out the cam, as shown in fig.
If it is nicely rounded off, giving a gradual rise, very little tension (or compression, as the case may be) of the controlling spring will be necessary to give the required speed to engine; whereas, if the rise is sudden, the spring will have to be screwed up tighter, and, if uneven and lumpy (_i.
A small stop interposed between the lever and some convenient part of the engine, such as the side-shaft bracket bearing, answers this purpose.
The small lump on the back of exhaust cam, fig.
Fig.
A lock nut should be used in conjunction with this set screw.

Still polishing:

How a brain-computer interfaces works

It will be seen that the weight W (which is only held in the position shown by the spring S) will tend to lag behind when a sudden upward motion is imparted to the lever L.
The slightest movement of the crank from this point in a forward direction should result in a little play being felt in the lever L, assuming that the cam is also moved just enough to keep the scriber marks in line with the existing keyway.
of gas were required per horse-power per hour, but it was found that as much as 105 cubic ft.
The wheel drives a brass or gun-metal plug, producing an intermittent rotary motion.
The difficulty of producing an efficient oil engine lies principally in devising a satisfactory and reliable vapouriser--one which will work equally well under all loads.
Barsanti and Matteucci were engaged in devising and experimenting with an engine very similar to this some years before, but Otto & Langen, no doubt, worked quite independently.
The gas and air valve are shown as both being operated by the same lever P, the accurate timing of the latter being obtained by means of set screws.
e.
Assuming that we have both cams finished to the proper shape and size, and the keyway cut in the side shaft, we can commence to mark off the position of keyway in the air cam.

a little test. Will look better with different old font I think, and maybe some more formulas.

some refinements of the diagram generator:

Pages are coming together. Here's a current screengrab from one of the pages. Some time polishing the output is ahead.

Project repo lives here. I'm actively working on it.

I like this. Keep it up!

looking for suggestions of how to generate diagrams or funny pictures. ideally, all images of obscure machines.

Well, since this has made me think about it...

(1a) Patent applications often contain diagrams, and after time they fall into the public domain. Apparently NASA released some of these in 2016, but the link to their "searchable database" is broken. One problem remains: extracting the relevant images. I worked on a tool for this a couple of NaNoGenMo's back, using Python and pillow, but I don't recall how well I got it to work.

(1b) Scans of old books about machines on e.g. archive.org would work just as well as patent applications. I haven't looked though.

(2) You could also do a search for "machine" on one of those free vector clipart sites. For example: https://free-images.com/search/?q=machine&cat=vr . You definitely get mixed results that way. Which might actually be a plus. But it's hard to control the aesthetic.

(3) And then there's graphviz. This is a command-line tool that takes a file in .dot format which describes a graph, and generates an image of boxes and arrows and labels and stuff. You could probably get a lot of mileage just randomly generating .dot files. But you'd have to learn the .dot format first.

Could also check Flickr Commons of free images:

• https://www.flickr.com/commons

For example searching British Library for "machine":

• https://flickr.com/search/?user_id=12403504%40N02&view_all=1&text=machine

Thanks for all these tips, folks! I spent time going over them and appreciate your insights.

Splitting the image generation into two goals:

• diagrams
• photos of machines

For the diagrams, the graphviz idea is cool but i'm already familiar with Processing and kind of enjoying my random diagrams I've been making so far. So will keep doing that.

Now for images I had trouble getting usable images from the NASA patents and for sure it's a bummer that only 5 years out their link to public domain'ed patents doesn't even work anymore and I can't find on their website.

I guess I also need to decide if this will result in a printed book, and if so, maybe need higher quality images. I'm looking at the Flickr.com No Known Copyright Restrictions (some may be NSFW) and maybe images tagged something like industrial on Pexels. I might use that as a start and sub-select 500 to 1000 images, cull out duds, and run some filtering on selections, something like:

1. Pick a random image out of the public domain image bucket
2. Zoom in some amount (1 to 5 times).
3. Crop out a section.
4. Convert to black and white

Ok, here's some simple code, whipped up as a test in a jiffy from previous code sketching, and example generated images. First 3 images use threshholding to convert to black and white, and latter two do not.

Trying to decide if i like these or if they are too oblique. Perhaps I should also try using a bucket of machine parts like gears, ratchets, etc and then randomly assembling those together? might be the next try.

tried adding a caption to the images:

i like this a bit more

Some updates. I now have a working book, but I think I want it to be a bit funnier while also making more logical sense. Here's some screenshots to show some pages. I think massaging the output will be the hardest. The first part was primarily a coding challenge, and now I get into the 'aesthetics' and form of the output in more detail.

I have a complete beautiful book compiled. Two big changes have allowed me to finish up:

1. I added a line to my fish script to use grep to convert the word "engine" to the technology currently being explained.
2. I added the pdflatex converter to my system so that pandoc can convert markdown directly to pdf and the rendering is beautiful.

Yes, I finally installed pdflatex which I'd been avoiding for years because it has 546MB of dependencies, which seems insane since all it's supposed to do is add the ability to convert to pdf in pandoc. Half a gigabyte for that? Nuts. Aside from the fact that it's probably 520megabytes too big, I'm glad I finally installed it because it made a big difference. Using the markdown to pdf converter presents a beautiful looking book.

Here is one example of compiled book of Technology Simply Explained (PDF).

I'll probably make some more. And potentially do a print run with a nice looking cover. But I'll consider this complete now.

And some rendered final pages

.

Wow! This reminds me of undergraduate engineering labs where they give you some piece of equipment and the operation manual. The headings were hopeful, and I can see how useful this manual will be in the future. Great arrangement of informative diagrams near relevant instructions, and helpful caveats. Any engineering student would be lucky to have such a comprehensive tome. Makes a great gift for the person who's always fixing everyone else's things. I'm putting this right next to my Machinery's Handbook 👍

hahah thanks angela!