Cute. Although rather complex for just basic gates. It's good that he figured out a way to make an amplifier. Although it's rather complicated.
Mechanical logic used to be much more popular. Adding machines, calculators, and tabulating machines used it. So did railroad interlocking systems and race track totalizators.
A Model 14 or 15 Teletype machine's printer is a mechanical logic chain. It's a triggered clock driving an amplifying 5-bit shift register with latches. Those 5 bits are transferred in parallel to a 1 out of 32 decoder, where another level of mechanical amplification drives the typebars. None of the historical documents describe it that way, because there was no general language back then for talking about logic elements.
The simple way to make a mechanical digital amplifier is with an interposer. You have a "clock", a constantly rotating shaft, with a cam generating reciprocating motion. After the reciprocating part, there's a gap, and then something that needs to get pushed as the output. The interposer goes into the gap during the "off" part of the clock cycle, which takes little energy, and gets pushed by the high-powered reciprocating part. This can easily provide a gain of 10-100. This classic mechanism appears in many punching devices, from little ones punching holes in paper tape to big ones stamping out car bodies.
US patent 3659779, "Punch mechanism", assigned to IBM. This is how card punches worked. A little solenoid moved the interposer into position, while a big motor with a flywheel provided the punching power.
The "rotational" logic version was suggested to me as a way to implement a computer by a mechanical engineer in ~2004. He claimed that such a "boolean calculator" had been initially designed as an alternative to the first electromechanical computers during WW2 — trading off speed for reliability. The design he suggested used spur-gear differentials for all parts of the computer, including the one-way gear (using a free-spinning axle and a pall). Using a single "element" (analogous to the transistor) simplifies design & implementation. (The one-way gear shown in the video has failure modes that the one-way spur-gear mitigates to a great extent.)
I've always wanted to build one out of wood and brass.
A rotational mechanical computer will work in a way that a "push-pull" mechanical computer will never do, at scale. The main take-away is to understand that — like a transistor — the rotation is "saturating": you keep turning, but the state doesn't change after a given "settle" time. For "push-pull" the mechanical computer will never truly saturate (friction; compression; ...), and the larger scale components will fail to propagate their "signal".
Once you've got that you need "saturation", then you'll find that there's "saturating" mechanisms all around you! Consider a cup with a hole in its bottom: if the incoming water is less than the outflow, the cup will drain. But, let's say you have two incoming streams such that their sum fills the cup faster than it drains: now the cup will fill up. If the cup is a counter-weight to a funnel, you've got an implementation of an nand-gate: the cup controls whether or not a continuously fed stream of water goes to the next cup, but the water going into the cup only controls its saturating behavior. I've always imagined a Japanese-style bamboo garden very slowly performing calculations [1].
This will sound like a joke, but I would recommend legos for teaching kids about cars, and minecraft for learning to build a computer. As someone else just said, this not very intuitive.
Seems like the message might get lost on someone trying to replicate these and it does get expensive in significant numbers.
Mechanical logic used to be much more popular. Adding machines, calculators, and tabulating machines used it. So did railroad interlocking systems and race track totalizators.
A Model 14 or 15 Teletype machine's printer is a mechanical logic chain. It's a triggered clock driving an amplifying 5-bit shift register with latches. Those 5 bits are transferred in parallel to a 1 out of 32 decoder, where another level of mechanical amplification drives the typebars. None of the historical documents describe it that way, because there was no general language back then for talking about logic elements.
The simple way to make a mechanical digital amplifier is with an interposer. You have a "clock", a constantly rotating shaft, with a cam generating reciprocating motion. After the reciprocating part, there's a gap, and then something that needs to get pushed as the output. The interposer goes into the gap during the "off" part of the clock cycle, which takes little energy, and gets pushed by the high-powered reciprocating part. This can easily provide a gain of 10-100. This classic mechanism appears in many punching devices, from little ones punching holes in paper tape to big ones stamping out car bodies.