I'm not sure what search terms you used to search, but you might want to try broader terms. Proposing new mechanics pieces has been popular for a while. Logic and dwarfputing mechanics were a major component of the
Additional Mechanics thread was a major and popular thread back in the Eternal Suggestions Voting days, and managed to get 9th place. As such, at least some of its suggestions were added to the
dev goals under the header "Improved Mechanics".
In fact, if you haven't seen many threads on the topic recently, it's because minecarts radically changed dwarfputing and made it far easier than it had been, previously. Minecarts are less space and CPU-intensive than fluid logic, and also solve some problems of transition between signal and power. (Likewise, the addition of workshop raws allows powered workshops to be added to the game, which quiets some of the concerns in that thread.)
With all that said, Toady has publically stated in the past that he is leery of outright machinery built for the purpose of being giant, physical logic gates, as what we can accomplish right now are basically happy accidents of emergent gameplay, not deliberate game goals. Outright making mechanical logic gates isn't something he's supported in the past. Hence, the Goldberg-esque nature of dwarven mechanics right now is somewhat deliberate, as far as I can tell. Regular dwarves, and even regular Dwarf Fortress players, don't make these things, only some players do, and when they do, they're going against the nature of the world and gameplay. Making fortress mechanics that outright demand the player be functionally programming with fortress pieces is something that changes the tone of the game in a way Toady may not agree with.
That said, "moving fortress pieces" and having to assemble one's own traps mechanically on the map in physical space is something that already delves into this territory, already...
What I would honestly prefer (and have
suggested in the past, actually,) would be "black box" mechanical buildings that allow players to program a set of inputs and outputs. Hypothetically, this could go the extra mile of having a fully simulated internal space that simulates all forces that could be applied to it, and then creates an abstracted set of rules that it operates by when "zoomed out" again to normal fortress space. This allows for hypothetically any arbitrary logic problem whose mechanical implementation fits within the confines of the simulated area of the box to be produced, including all basic boolean logic, although potentially with a greater degree of inputs and outputs. This would also have the advantage of being computationally much simpler than a minecart-based, much less fluid-based "dwarfputer", as the internals of the system could be abstracted to "signals in" and "signals out".
The gearbox would be a single-tile building or possibly a field of buildings that could take inputs from different directions or from a "signal" like from a lever being set to "on".
For example, you might have a gearbox with power from an axle coming from the West, and which transmits that power to the East if there is an "on" signal from the lever. This basically functions like an "AND" gate, as both power from the West and the lever need to be "on" to send the signal forwards.
One might also try to have a "pressure plate" inside the gearbox, such that the same gearbox would send a pressure plate-type signal instead of transmitting power while "on". This way, power to a box might open a door, while losing power will shut the door.
Likewise, one could have a larger and more complex set of inputs or outputs. North, West, and South might all have powered axles as "inputs", as well as the state of a lever or pressure plate might be "inputs" that affect the state of the gearbox, with "outputs" being an axle to the East, Up, and a pressure plate-like signal that might be sent elsewhere. This, obviously, can make it more complex than simple single-stage Boolean commands.