NandGame – Build a Computer from Scratch

Silicon Zeroes[0] is also a pretty good next step if you enjoy these games. Doesn't quite get down to the NAND level, but gets into a lot of fun performance optimizations instead.

[0]: https://store.steampowered.com/app/684270/Silicon_Zeroes/

Rethink about what a computer or calculator is? It depends on what you value, TBH. For example, let's go back to what Von Neumann architecture, the ENIAC:

ENIAC: Von Neumann quickly appreciated the crippling limitations of cable plug-board-and-switches “programming,’’ and in fact had already conceived the revolutionary idea of not only stor- ing a problem’s input data inside the machine but also so storing the program itself that would manipulate that data. Von Neumann was smart and apparently he was known to calculate in his head so he could compare what the computational equivalent, but his wife did most of the accounting for him, as did most of the so called human calculators of the day. He held back progress of computer architectures for a long time.

Now, the ENIAC's successor: MANIAC-I 37 commands As a measure of MANIAC-I’s speed, it could add two numbers in ninety microseconds and multiply or divide two numbers in one millisecond.

Briefly, the problem was to study how the vibrational energy of the crystal, if initially located in just one mode of oscillation of the crystal’s atoms, would over time spread itself throughout all the possible modes of oscillation (the meaning of mode of oscillation will be explained soon). This mode spreading of the energy is called the equipartition of energy in statistical physics. These days, "seconds" is already defined (at least in the Si system, but generally you'd want to Metrify stuff because it lets you do paper napkin physics much easier). You can build huge "global wires" of information that the "computer" can precalculate, it may require revisiting the limits of what can be achieved "in silico" or without hybrid "data flow"/"control flow" architectures and distributed virtual power transmission still allow for matter field bulk transport when needed, but dispatch shared resource usage more smartly (this includes stuff we can't "see" like consumption of natural resource flows. these tend to be global issues).

If you like this or want to get into it a bit more, this recent video is lovely https://www.youtube.com/watch?v=QZwneRb-zqA

I find using an FPGA for this an ideal middle point. Building a functional computer from discrete components is fun but the result may not be worth the effort, to be honest. At the other extreme, modeling it in software may leave a weird feeling of not having achieved anything. With an FPGA, on the other hand, you can create something of a usable computer, even one you could run Linux on.