Not Just a Camera, Not Just an Engine
Rejecting blind builder and helpless witness narratives in favor of constitutive narratives
It is hard to make sense of events these days because we feel constantly forced into a false choice between blind builder narratives and helpless witness narratives. Stories told by people so enthralled by new agencies they don’t notice their insensibility to current realities, or the poverty of their future visions driving their excited building. Or stories told by people so lacking in agency of any sort that their visions, while richer, are uniformly bleak and framed by their own sense of utter helplessness and doom.
The fundamental inadequacies of these frames, much more than the right/left political leanings usually associated with them, is perhaps the real reason for my refusal to ally with any of the narratives on offer. I don’t want to be either blind or helpless, or move along a tradeoff curve between them.
The Contraptions Book club June pick is Monkey King/Journey to the West (pick either the Lovell or Jenner edition). Chat thread here. We will discuss this week of the 23rd.
An interesting pattern that’s popped for me as a way out of this bind, and a possible stance from which to narrate and inhabit more powerful sorts of stories, is working with media that are simultaneously about seeing and doing.
I first noticed this pattern in Donald Mackenzie’s idea of the economy as an engine, not a camera, which I turned around and applied to AI as a camera, not an engine. But the rhetorical trick in both cases only works because both qualities are present in the thing being described. Both the economy and AI simultaneously exhibit engine-like and camera-like qualities, and require camera-engine stances to make work with. The camera-engine is the duck-rabbit of sense-making and agency.
I just came across a similar idea in Carol Rose’s essay (ht Arvind Narayanan), Crystals and Mud in Property Law, which cites Lawrence Tribe as the source of the idea that every legal decision “tells a story about the kind of society we live in. Decisions … are constitutive,” and the related notion due to James Boyd White that legal language is “constitutive of culture.”
In other words, camera-engine narratives are constitutive narratives that simultaneously observe/narrative and create/steer reality. How can you work with these narratives?
Constitutive narratives are a stronger form of what are sometimes called speech-acts — words that, by virtue of being spoken, create new realities, such as a priest or magistrate pronouncing a couple “married,” creating a new entity. But constitutive narratives do more than create social realities that primarily live in minds. They can create or reshape material realities. For example, courts can require features to be added to buildings to bring them up to code, or declare that money or authority be moved from one locus to another.
And it is not just courts that can speak-act constitutively, narrating new realities into being. Many individuals and institutions have similar power within their domains. The catchphrase from the Ten Commandments, so let it be written, so let it be done, is not just for Pharaohs.
And note that we’re not talking of narratives that merely persuade. We’re not talking of seers and talkers convincing doers and builders to behave in certain ways. Constitutive narratives directly create realities. When a judge declares a couple married, things start happening in certain ways regardless of what the “agents” involved believe or decide to do.
A full engagement with reality involves both seeing and doing as intertwined double-helical rather than serial modes that are to some extent overlapping apprehensions of the same thing. It is not just quantum realities that are altered through the act of observation. Any kind of observation that is full-dimensional must alter the reality it observes. There are things about reality that cannot be grasped through the senses alone, or through the maps and mental models alone. Reality is both a state of being, and a state of becoming. The latter aspects cannot be comprehended through passive witnessing, while the former cannot be acted upon without vision.
Sometimes, to observe a thing is to poke it with a stick to see what it does. Sometimes to act on a thing is to say something about it.
The dispositions of a reality that lend it a tendency to change in some ways rather than others can only be observed by trying to change them, while observing the process of action-entangled change unfold. You can only see what is. You can only do what could be. You can only grapple with both what is, and what could be, by trying to see and do at the same time.
Constitutive narratives subvert falsely separated notions of seeing and doing and offer only a single mode of engagement with reality — a see-do way that gets entangled with both being and becoming processes in a way that erases the sense of individual separateness from reality; of “agent” or “seer” separated from “acted upon” or “seen.” (I called such narratives enactments in my book Tempo, and my account of such phenomena in that book anticipated this updated understanding for me).
From the point of viewing-and-action of see-do people, those who neglect one mode in favor of the other appear ghostly. If you’re not equally part of the being-and-becoming of reality, you’re surreally detached from reality altogether, a ghost harboring illusions of being real.
An ordinary individual who happens to be seeing-doing in an integrated way at modest scales often appears much more real than either an apoplectic billionaire flailing about to convince himself he’s “doing” things or a fretful intellectual endlessly rehearsing favored descriptions of reality, updated periodically with new observations, helplessly gathered.
From the perspective of constitutive narratives, both Silicon Valley doerists and tech-critics endlessly circling their favorite talking points strike me as fundamentally ghostly. Red or blue-tinted clouds of ectoplasm signifying nothing. Their respective political brethren seem even less substantial, substituting theaters of action and observation for the real thing.
This does not mean these ghosts are inconsequential. They have a certain poltergeist-like ability to affect reality without adding commensurate amounts of meaning to it, thereby acting as agents of entropy; closer to non-sentient disaster forces than living forces.
Their narratives are inconsequential in other words, even if their presence in a situation is not.
A hurricane is a consequential force, but if some clever AI model builder found a way to reconstruct and narrate a speculative pseudo-sentient internal monologue of a hurricane as it storms around, it would… not matter. You just have to model (a kind of seeing) the laws of the atmospheric physics and act according to the behavioral predictions. Silicon Valley these days must be taken seriously as far as its behaviors go, but the narratives it operates by are largely irrelevant derps.
Similarly, the constitutive aspect of the visions of leftist activists — the lawsuits they initiate, the protests they organize — must be taken seriously and accounted for in your own actions, but their governing narratives can also be ignored.
These are hurricanes haunted by ghosts that imagine themselves to be living and real.
If you’re acting strongly in either camera-over-engine or engine-over-camera modes, your internal narrative can be ignored. Even riding the tradeoff curve is not enough to stop being a ghost. You have to transcend the engine-camera dichotomy in your behaviors.
Otherwise, you can be reduced to your constitutive behaviors (which are likely going to amount to meaningless entropic flailing and may have no relationship to the story you’re telling yourself about what you’re doing), and modeled like a non-sentient entity. You have reduced yourself from a you to an it.
Assuming you find this theory of constitutive behaviors and narratives to be persuasive, how should you see-do? How can you transcend?
I think a big part of the answer lies in the medium through which you choose to act.
What sort of camera-engine medium should you adopt for your own behaviors?
A simple example of a constitutive behavior today, for a writer, is to simply use AI tools in your writing. Regardless of what you co-write with AI about, or how well it turns out, by the mere fact of trying at all, you’re reshaping reality for everyone. AI-generated words today have a constitutive power that human-generated words generally don’t (outside of contexts like legal language). They may be consequential though, without being constitutive.
There are camera-over-engine media available to you: Ordinary words, ordinary artworks, giving talks, sharing insights, “owning” political opponents with rhetorical coups, peddling stimulating insights that constitute no new realities.
There are engine-over-camera media available to you: Woodworking, building hardware or software that sees/says nothing, founding a by-the-book Silicon Valley startup, building ships, highways, and bridges, and so on.
And then there are the strongly entangled engine-camera media: working with AI, legal codes, lawsuits, legislative efforts, code-as-law, and my emerging personal favorite, protocols.
When you act through a camera-engine medium, you cannot help but see and do at the same time. You cannot help constitute realities while observing them.
There are also what I call theatrical media: Those that are really near-pure engines that offer illusions of being cameras too (camera theaters), and near-pure cameras that offer illusions of being engines too (engine theaters).
A quick test — if the actions seem full of energetic noise and fury but curiously lacking in efficacy or lasting effects, you have an engine theater. Many protest movements and commentariat subcultures, at least in America, are engine theaters.
If your words seem full of gloriously recycled cliches, leaving you with a lingering sense of having learned absolutely nothing from listening to them, you’re dealing with a camera theater. Much of the Trump regime’s actions are camera theaters. There is an illusion of being presented with a real story describing the way the world was, is, and could be, but poke a little and there’s nothing there. It’s just random acts of acting for the sake of acting, adding up to nothing but entropic destruction.
I don’t yet have a strong handle on camera-engine media and their properties, and how to wield them well, but I think I’ve already developed a strong sensitivity to degenerate media where one mode overpowers or excludes the other. I can tell when there are ghosts rattling about, pretending to be humans, and am slowly getting better at not taking them at their own narrative estimations.



This kinda reminds me of my gnostic psychobabble phase where I stumbled into camera mode while operating in engine mode.
There’s another duality worth considering here :
brain churn x field connection.
They’re mutually exclusive yet braided in interesting ways.
Setting each dial to max creates distortions like the brightness / contrast setting or bass / treble.
All Brain churn no field connection is actuarial living where everything is underwritten. These are people who believe that love is just a bunch of undiscovered chemistry lessons.
They’re also people who endlessly need to fact check rhetoric that’s generating strong goosebumps in others.
All field connection no brain churn are heroin trips. Goosebumps all the way down. All gas no brakes.
LLMs represent a braiding where synthetic brain churning clashing with human prompts could plausibly catalyze a field connectivity similar to someone deep in Hare Krishna chanting or evangelical trances.
It feels like we are entering into an era dominated by an ability to connect with fields rather than make brains churn.
Where Goosebumps are the new oil.
This puts me in mind of how our understanding and of the human genome is progressing. At first it seemed like the DNA was the camera and the RNA was the engine but it turns out there is a ton of mixing via regulatory systems, the shapes things take and factors in the environment. It would be much easier to understand if DNA was actually a blueprint and RNA just decoded it but alas it's much more complicated. I wonder how the approaches scientists are currently using to understand regulatory networks might be relevant to these weird feedback loops we are constantly faced with now.