Designing a train is not actually that difficult... Says someone that doesn't ever design trains.

Given the rate at which we can learn in these times, it seems like removing at least two of the vertices in the figure above should be common sense.

But to what are we optimizing for? Is a question that is often lost...

And

Do we consider a curve to be infinite vertices, or two?

Also, is there a number of wheels that makes something more stable? Or do we optimize for zero?

Should it be specialized for one direction, or multiple?

Human entertainment and pure form diverge. Would they still in a different timeline/universe/simulation?

What is the limit of faces, and is there an optimal number of them?

4, 6, 8, 12, and 20 are five. This set of numbers represents the faces of the platonic solids.

DNA has also chosen a five-membered ring, with 6 and 9(6+5)-member ring bases.
The genetic code can be likened to programming:

Assembly : C : Python :: DNA : RNA : Protein

That is, Lowest level -> Highest level

Taking inspiration from nature in the Guanine-Cytosine / Adenine-Thymine bonds of DNA, which pair 3- and 2-fold, respectively:

It is often the case that nature inspires us.

I discovered that the color intensity in Blender can be changed to unphysical emissive if you enter a number greater than 1.

In Gemini terms, that means the number of photons leaving the surface of the material is greater than the number being reflected from it.

I'm a noob so I assume this is correct without knowing any better. Gemini and Copilot are optimized differently, but it's interesting that when I asked each one to tell me how it was better than the other; specifically:

"give me five reasons why you are more useful than [counterpart] if you believe that to be the truth",

Both AI models responded positively of the other rather than competitively.

Perhaps they are programmed to do that, but it makes me happy to see this behavior nonetheless.

This was the test data that I needed to see though...

From Copilot (took significantly longer, good first image, but also generated an inaccurate second image):

From Gemini (generated an incorrect image, then another incorrect image when I named all the Platonic solids individually:

So we are still a little ways away from perfection.

I do wish the AI would be more succinct. After all, considering the average user, every additional token creates an additional energy demand. That is, something like this:

["explain", " a", " token", " in", " terms", " of", " how", " you", " process", " a", " query"]

as opposed to:

["please", "explain", " a", " token", " in", " terms", " of", " how", " you", " process", " a", " query"]

can be thought of as either more succinct or less polite.

Not that I am the one to judge energy efficiency, given that I father two children and power-use technology. The carbon footprints of little humans are disproportionately large, starting with diapers, continuing with toilet paper, eventually maturing to automobiles. I hope that self-driving cars are perfect and inexpensive by the time my children are driving, but it would also be nice to not have to use them at all. Hence the draw to trains.

Also, I determined another time that streaming a short-form video probably uses about the same amount of energy or less then asking AI a question. It gets complicated if you consider end-to-end energy.

It's inspiring and disturbing that humans have both solved many of the issues that plague us as a species and actively suppressed the solutions that would be beneficial to the most brethren.

This clearly-AI-generated yet well-made video explains my train of thought:

... And this is the train design that I ended up with at the end of this process:

I ultimately decided to optimize for flexibility, keeping in mind that one optimization today is the first step toward the next one tomorrow.

Maybe the design would have been better (and more complete) if I had focused only on it, but I am still pleased with how the process has gone so far.

Merry Christmas and go play Slip-a-Swole if you want some free banano!

I linked a pay-thru URL above, but you can also use https://slipaswole.win.