Granted! It would be interesting to see what the scaling limits are. If I recall he provides some scaling laws in the paper for the different forces but would likely have some problems creating wings with the necessary flex as they got larger. Maybe some kind of carbon fiber composite?
This is a beautifully written description of what distinguishes a contraption from a device. What stood out to me is the description of contraptions as being from a different universe, and thus fundamentally incompatible with related inventions, both in terms of cross-innovation and in usage (perhaps analogous to language isolates, and other forms of speciation).
It also occurs to me that the insular and monolithic character of contraptions can also be described _procedurally_. Long ago, when I was a student of software engineering, an old professor described the various high-level ways to engineer new software, and said that being able to categorize your task/problems into these ways is essential to avoiding going down too many blind alleys.
The first way of software engineering is _forwards evolving incrementalism_: you have a starting-point (i.e. some software you are paid to work on, like maybe a file-system), and you have a destination (i.e. file-system needs to support files larger than 4GB), and you keep making changes to the file-system until it meets that new constraint. A similar dynamic can be seen when you use libraries, but instead of changing the software, you are changing the details/combinatorics of how it is used. In all cases, you can build a continuous genealogy or gradient that relates much of the software in the world (consider all the various instances of Linux and Unix derivatives).
The second way of software engineering is _randomized evolution_: you have a starting-point, but no destination, and you tinker with the software until it does something interesting. Most games seem to be made in this way (and necessarily so).
The third (and most powerful) way of software engineering is _backwards evolving integralism_: you have a well-understood destination, but no starting point, and you have to work your way backwards to the nearest (yet unknown) starting point. This sounds a little confusing, so let me illustrate, with the helicopter example: you imagine the idealized destination, which is basically a kind of invisible magical force that can make things go up vertically, make it move horizontally to its destination, and then make it go down vertically. In this description, we have not mentioned any mechanical components. The next N steps involve figuring out how to assemble available components into a contraption that can do those things. Because you have a very specific set of constraints, that are not satisfied by existing/related solutions (because otherwise you could just use the first-way to make improvements/optimizations), you end up with a very specialized system, that looks as if it came from nowhere, kind of like a big-bang. These kinds of examples are present in software too. The ZFS file-system can do a set of things that no other filesystem can do reliably _and_ efficiently, and it is no coincidence that it was (a) written from scratch, and (b) attempts to clone it (i.e. btrfs and bcachefs) have failed miserably (mostly because it is nearly impossible to out-ZFS ZFS -- also note that these cloning-attempts are also from-scratch (the problem resists way-one)). It was written from scratch, because evolving ZFS from UFS (or anything else), would no doubt take much longer than just writing a filesystem that was designed around those constraints.
In other words, way-one and way-two are extremely sensitive to prior decisions (especially decisions made by other engineers), while way-three is only sensitive to constraints and available/known components/techniques. Way-three is what you use when you have constraints that are truly novel (i.e. they define a new product category). I suspect that what you call contraptions are almost all the result of "way-three engineering", perhaps combined with certain economic constraints (i.e. only off-the-shelf components wherever possible).
Reduced to a single sentence: Contraptions imply way-three, but way-three does not necessarily imply contraptions.
Anyway, thank you for writing this piece: I have been thinking about this contraption-theory for a few days now.
I'm reminded of the old ribbonfarm article [1] about planning to start and planning to finish.
The first two describe planning to finish and planning to start, respectively.
The last contraption-y angle is planning to finish but backwards? Not sure how to parse it, but it's character makes me feel like working backwards in that way, of having a specific goal in mind, is what allows you to "hover" instead of "flying." You end up needing a chimera of some sort that doesn't obviously descend from nature, but rather remixes natural things in a way that nature itself would never produce. You are going for a *specific* kind of end-experience.
Notably, once it's been achieved "planning to finish" approaches trying to replicate it fail, they have to essentially take that same end-point and work backwards all the same, potentially in an altogether different kind of way.
The more contraptiony something is, the more closely it hews to contingencies. Borrowing the idea of ecological niches from biology — which are basically bundles of environmental contingencies — feels like a helpful frame: you find inexhaustibly many contraptions in the biosphere. Evolution is essentially a contraption generator.
In fact, this seems like a primary distinguishing characteristic of evolutionary and intelligent modes of creation. Certainly either mode can produce more or less contraptiony things, but the central tendency of evolution is quite a bit more contraptiony than human technology.
"Are there are other contraptions in history, or around us? Important ones, not marginal curiosities like hovercraft?"
Having grown up in India, the whole genre of "jugaad tech" has always irked me a little bit, even when it solved real problems within all too real constraints. Part of that may just be elitism I'm not quite ready to fess up to lol, but a big part of it has to do with just how lazily most of those ideas are implemented. But the Contraption Factor gives me a very neat way to articulate that! Reading this felt like an annoying guava seed lodged in my tooth for years suddenly come loose!
However, in terms of an example, on a not so recent trip to the Himalayas in the middle of winter, I saw something called an agyaari. It is apparently a fairly standard fixture for space heating in houses in the colder parts of North India. Talk about a contraption with no evolutionary path forward! It is literally a huge metal duct zigging and zagging across the room with the vent smack down in the middle of the space. A typical Western fireplace would be the winged airplane equivalent of that, but the kind of heating the agyaari offered in the middle of that brutal winter is something that a cozy fireplace can only dream of. I'll try and find a good YouTube video just so everyone can get a sense of how janky and contraptiony the thing is.
> [Contraptions are] ... more discovered than invented.
This rubs me as wrong but interesting. I think the ur-discovery is relativity - replacing gravity with warped spacetime, replacing magnetism with relativistic charge-invariance. It's a simplification that will look the same to every alien who finds it.
I think helicopters are an ur-invention, in that they almost 100% environment- and path-dependence. There is the handedness based on region of origin, but there's also buoyancy / reynolds number / power-density. There are no aquatic helicopters, jellyfish are a better plan there. At human size with just enough power density, you're obligated to grab the span-efficiency of a single rotor which means you're stuck with swashplate and tail rotor. But if you've got more juice you can just go quadcopter and avoid all that hassle. The single-rotor + tail-rotor seems like something you'll only bump into in just the right environments with just the right tech-trees.
I think you're gesturing at things like this as deep conserved quantities, and I honestly can't tell if that's usefully more true than relativity or less. DNA replication / repair is contraptiony and path-dependent, and seems like a more important deep conserved kind of thing than even relativity, so as usual I guess you're on to something lol
Is the SR-71 a contraption? I believe it is the single most improbable aircraft ever created. Or contingent, if you will. There had to be a mostly non violent, intense rivalry between superpowers after turbine engines, but before reliable space flight. AND you needed a genius like Kelly Johnson. AND a big enough budget to create all the one off innovations needed to support the SR-71.
Once a society reaches a certain level, a wide body four engine jet like the 747 is probably inevitable. With enough concentration of wealth the Learjet is inevitable. But the SR-71 was special.
As a nit picky quibble, I feel like your parallel universes shtick is unnecessarily floofy. Quirky, fiercely independent or neurodivergent probably all have a higher cosine similarity to the thought vector you’re trying to express.
Record turntables are contraptiony, that’s part of what’s cool about them. So many different designs and price points all trying to more or less do the same thing but within fairly narrow physical parameters
quick poll - how many readers here are or like "contraptioneering"
Here is a prototype of a hovering design that is visibly symmetrical from Leif Ristroph at the Courant Institute: https://www.youtube.com/watch?v=boxeUaFl3R8
It's inspired by the way jellyfish move through water so a clear natural precedent there
Cute but nope.
As in… not practical for large enough payloads to be interesting.
I think flapping is fundamentally limited by the medium. Blue whales can do flapping flight in water. Nothing that size will ever flap in air.
Granted! It would be interesting to see what the scaling limits are. If I recall he provides some scaling laws in the paper for the different forces but would likely have some problems creating wings with the necessary flex as they got larger. Maybe some kind of carbon fiber composite?
This is a beautifully written description of what distinguishes a contraption from a device. What stood out to me is the description of contraptions as being from a different universe, and thus fundamentally incompatible with related inventions, both in terms of cross-innovation and in usage (perhaps analogous to language isolates, and other forms of speciation).
It also occurs to me that the insular and monolithic character of contraptions can also be described _procedurally_. Long ago, when I was a student of software engineering, an old professor described the various high-level ways to engineer new software, and said that being able to categorize your task/problems into these ways is essential to avoiding going down too many blind alleys.
The first way of software engineering is _forwards evolving incrementalism_: you have a starting-point (i.e. some software you are paid to work on, like maybe a file-system), and you have a destination (i.e. file-system needs to support files larger than 4GB), and you keep making changes to the file-system until it meets that new constraint. A similar dynamic can be seen when you use libraries, but instead of changing the software, you are changing the details/combinatorics of how it is used. In all cases, you can build a continuous genealogy or gradient that relates much of the software in the world (consider all the various instances of Linux and Unix derivatives).
The second way of software engineering is _randomized evolution_: you have a starting-point, but no destination, and you tinker with the software until it does something interesting. Most games seem to be made in this way (and necessarily so).
The third (and most powerful) way of software engineering is _backwards evolving integralism_: you have a well-understood destination, but no starting point, and you have to work your way backwards to the nearest (yet unknown) starting point. This sounds a little confusing, so let me illustrate, with the helicopter example: you imagine the idealized destination, which is basically a kind of invisible magical force that can make things go up vertically, make it move horizontally to its destination, and then make it go down vertically. In this description, we have not mentioned any mechanical components. The next N steps involve figuring out how to assemble available components into a contraption that can do those things. Because you have a very specific set of constraints, that are not satisfied by existing/related solutions (because otherwise you could just use the first-way to make improvements/optimizations), you end up with a very specialized system, that looks as if it came from nowhere, kind of like a big-bang. These kinds of examples are present in software too. The ZFS file-system can do a set of things that no other filesystem can do reliably _and_ efficiently, and it is no coincidence that it was (a) written from scratch, and (b) attempts to clone it (i.e. btrfs and bcachefs) have failed miserably (mostly because it is nearly impossible to out-ZFS ZFS -- also note that these cloning-attempts are also from-scratch (the problem resists way-one)). It was written from scratch, because evolving ZFS from UFS (or anything else), would no doubt take much longer than just writing a filesystem that was designed around those constraints.
In other words, way-one and way-two are extremely sensitive to prior decisions (especially decisions made by other engineers), while way-three is only sensitive to constraints and available/known components/techniques. Way-three is what you use when you have constraints that are truly novel (i.e. they define a new product category). I suspect that what you call contraptions are almost all the result of "way-three engineering", perhaps combined with certain economic constraints (i.e. only off-the-shelf components wherever possible).
Reduced to a single sentence: Contraptions imply way-three, but way-three does not necessarily imply contraptions.
Anyway, thank you for writing this piece: I have been thinking about this contraption-theory for a few days now.
I'm reminded of the old ribbonfarm article [1] about planning to start and planning to finish.
The first two describe planning to finish and planning to start, respectively.
The last contraption-y angle is planning to finish but backwards? Not sure how to parse it, but it's character makes me feel like working backwards in that way, of having a specific goal in mind, is what allows you to "hover" instead of "flying." You end up needing a chimera of some sort that doesn't obviously descend from nature, but rather remixes natural things in a way that nature itself would never produce. You are going for a *specific* kind of end-experience.
Notably, once it's been achieved "planning to finish" approaches trying to replicate it fail, they have to essentially take that same end-point and work backwards all the same, potentially in an altogether different kind of way.
[1] https://contraptions.venkateshrao.com/p/planning-to-start-planning-to-finish
The more contraptiony something is, the more closely it hews to contingencies. Borrowing the idea of ecological niches from biology — which are basically bundles of environmental contingencies — feels like a helpful frame: you find inexhaustibly many contraptions in the biosphere. Evolution is essentially a contraption generator.
In fact, this seems like a primary distinguishing characteristic of evolutionary and intelligent modes of creation. Certainly either mode can produce more or less contraptiony things, but the central tendency of evolution is quite a bit more contraptiony than human technology.
This was an awesome read!
"Are there are other contraptions in history, or around us? Important ones, not marginal curiosities like hovercraft?"
Having grown up in India, the whole genre of "jugaad tech" has always irked me a little bit, even when it solved real problems within all too real constraints. Part of that may just be elitism I'm not quite ready to fess up to lol, but a big part of it has to do with just how lazily most of those ideas are implemented. But the Contraption Factor gives me a very neat way to articulate that! Reading this felt like an annoying guava seed lodged in my tooth for years suddenly come loose!
However, in terms of an example, on a not so recent trip to the Himalayas in the middle of winter, I saw something called an agyaari. It is apparently a fairly standard fixture for space heating in houses in the colder parts of North India. Talk about a contraption with no evolutionary path forward! It is literally a huge metal duct zigging and zagging across the room with the vent smack down in the middle of the space. A typical Western fireplace would be the winged airplane equivalent of that, but the kind of heating the agyaari offered in the middle of that brutal winter is something that a cozy fireplace can only dream of. I'll try and find a good YouTube video just so everyone can get a sense of how janky and contraptiony the thing is.
> [Contraptions are] ... more discovered than invented.
This rubs me as wrong but interesting. I think the ur-discovery is relativity - replacing gravity with warped spacetime, replacing magnetism with relativistic charge-invariance. It's a simplification that will look the same to every alien who finds it.
I think helicopters are an ur-invention, in that they almost 100% environment- and path-dependence. There is the handedness based on region of origin, but there's also buoyancy / reynolds number / power-density. There are no aquatic helicopters, jellyfish are a better plan there. At human size with just enough power density, you're obligated to grab the span-efficiency of a single rotor which means you're stuck with swashplate and tail rotor. But if you've got more juice you can just go quadcopter and avoid all that hassle. The single-rotor + tail-rotor seems like something you'll only bump into in just the right environments with just the right tech-trees.
I think you're gesturing at things like this as deep conserved quantities, and I honestly can't tell if that's usefully more true than relativity or less. DNA replication / repair is contraptiony and path-dependent, and seems like a more important deep conserved kind of thing than even relativity, so as usual I guess you're on to something lol
Is the SR-71 a contraption? I believe it is the single most improbable aircraft ever created. Or contingent, if you will. There had to be a mostly non violent, intense rivalry between superpowers after turbine engines, but before reliable space flight. AND you needed a genius like Kelly Johnson. AND a big enough budget to create all the one off innovations needed to support the SR-71.
Once a society reaches a certain level, a wide body four engine jet like the 747 is probably inevitable. With enough concentration of wealth the Learjet is inevitable. But the SR-71 was special.
Sorry. Pet peeve. The Fermi paradox is deeply disturbing if you take it seriously.
As a nit picky quibble, I feel like your parallel universes shtick is unnecessarily floofy. Quirky, fiercely independent or neurodivergent probably all have a higher cosine similarity to the thought vector you’re trying to express.
https://chatgpt.com/share/b00760fd-e589-4c37-827a-e12670b9887d
Link doesn’t work
try https://archive.ph/GY7pz
silly prompt pretending a paper exists
Record turntables are contraptiony, that’s part of what’s cool about them. So many different designs and price points all trying to more or less do the same thing but within fairly narrow physical parameters