1.1Records and method
The records come from the public catalog of Cosmic Press, which the studio serves as a manifest and a set of numbered detail files. The manifest names 24 topics and 768 entries and gives the catalog's source as "AI-generated examples"; each of the 768 entries also marks itself AI-generated. An entry has a question in the voice of a person, a compressed general form of it, a context, a need, an assumption, a line on what the compression keeps (preserved), a line on what it loses (lost), a counterquestion, a defect and six scale notes. A record has no options and no judgment, and it never answers its question.
The studio files 31 entries under Technology: 9 in the catalog's first 240 entries, 21 in the long middle stretch from q241 to q744, and 1 in the closing stretch. The selection rule is one sentence: every third Technology entry in the catalog's order, counted from the third. It takes 10 entries, q180, q183, q186, q308, q380, q452, q524, q596, q668 and q740, of which 3 lie in the first stretch and 7 in the middle. The starting point keeps the sample clear of the 77 entries that the 6 earlier Cosmic Press studies read. Of the Technology entries, those studies read two, q260 and q764, and counted from the first the same interval would take both. The fetch scrubs any field that names a model; it found none to scrub in the ten, and no identifier or field of the ten names a model or a family of models. Four records speak of an artificial intelligence or an algorithm; these are the subjects of hypothetical questions, not names of systems.
The two stretches differ in manner. The three first-stretch records are terse: their questions run from 39 to 43 Korean characters, and none of their scale notes is a question. In the middle stretch the questions lengthen, to 86 characters at most, and all six scale notes of each of the seven are questions; two of them label the notes Individual, Relationship, Society, Other lives, Planet and Distant future. Of the 60 scale notes in the ten, 42 are questions. The two latest records, q668 and q740, return to small scenes and plain words. The study claims nothing about how these manners were made. It asks whether the way a Technology record places agency holds across them.
Cosmic Press publishes its records in Korean. The English in this paper is a translation made once by the editorial session of this study, sentence by sentence, with nothing added; each case's English seed names the Korean original by its hash, and the appendix gives every field of the ten in that English. Every block quote is that English, word for word. The Korean seldom names the subject of a sentence, and where the English needed one the translation used "one", "I" or "they" as the Korean's grammar allowed, never a gendered pronoun. Every count is made on the Korean by a plain word pattern named beside it in the study's numbers (numbers.md, made by scripts/count.mjs), so that the translation decides none of them. The patterns are crude and serve only to show that a reading is not confined to one record.
The counts set the scene. In 3 of the ten the question or its compressed form puts the choice as progress, and in 6 it uses a verb of replacing, taking away, removing, evaporating, causing or exposing. The defects then turn. In 4 the defect charges the framing with nostalgia, sentiment or retro taste, and in 6 it names a benefit or a helpful use of the technology. In 6 the counterquestion speaks for the technology or its gain. These turns are rare in the catalog and common in its Technology topic: a defect that names a benefit appears in 24 of the 768 entries, and 10 of those 24 are Technology entries; a counterquestion that speaks for the technology appears in 27 entries, 8 of them in Technology. In 6 of the ten the defect says the framing rested on one thing only. In 6 a scale note names law, regulation, policy, an institution or the public, and in 6 the last scale note sets the question in a future or an age.
The script also takes the series' frame measure: for each field, the share of its five-word runs that recur in at least 5% of the records that fill it, and a field whose ratio reaches 30% counts as the form's words rather than the record's. Every field measures 0% in Korean, in the whole catalog and in the Technology topic, and every field of the ten in English measures 0%. No field is set aside as frame. The form itself is fixed, however: every record must criticise its own compression and must pose a counterquestion. That a record finds fault with itself is therefore not a finding. What this paper reports is who acts in the record's sentences, which way the fault turns, and who appears in the place the form holds open.
All 10 contexts call their situation hypothetical. The paper reads every sentence as part of a hypothetical record's framing, draws no fact about any technology from it, and takes no side in any question a record asks.
1.2When the machine acts and the critique reverses the verdict
Some of the middle-stretch records put a technology to a verdict, three of them in the word progress itself, and in each the technology is the one that acts. The record's critique then answers the question's lean with the opposite verdict on the same actor. The clearest is an AI-generated record about a hypothetical smart city running driverless shuttles at night:
If late-night self-driving shuttle buses have been introduced and travel has become safe and cheap, but the warm gaze of the human driver who used to look after drunk passengers and emergency patients has disappeared, is that progress?
The sentence is built in the passive and the past. The buses "have been introduced", travel "has become" safe and cheap, the driver's gaze "has disappeared". No one introduces the buses; no one decided what the driver's work would be. The person asking stands at the end of a process and is offered one decision: whether to call it progress. The compressed form makes the technology the grammatical subject outright:
How do the cost savings and precision brought by making mobility technology driverless and automated replace the human care and safety net within the space of travel?
Here cost savings and precision do the replacing. The question asks how, not whether, and so the replacement is already granted. This is the stance that MacKenzie and Wajcman, The Social Shaping of Technology (1999), describe and challenge: technological change as something people may welcome or protest but cannot fundamentally alter. The question offers protest, framed as the loss of a warm gaze. The record's own critique offers welcome:
The limitation that, by reducing the benefits of introducing the technology to a sentimental backlash over the loss of the human touch, it may shut off the possibility of collaboration between technology and human care.
The defect glimpses a third possibility, "collaboration between technology and human care", which this paper reads as near to the duality in Orlikowski, The Duality of Technology (1992). But it names that collaboration only as what the framing "may shut off", and its work is to rebalance the verdict: the loss is called sentimental, and the benefits are restored. The reply the record poses to itself completes the turn:
Is a robot bus that carries no risk at all of drowsy or drunk driving, and that threads through every corner of the alleys 24 hours a day without tiring, not far safer for passengers?
Again the machine acts: it carries no risk, it threads through the alleys, it does not tire. The record has moved from the bus as the cause of a loss to the bus as the cause of safety. In neither does anyone design the bus, staff it, or decide where it goes; the hypothetical vehicle is the one who does things.
An AI-generated record about a hypothetical laptop launch makes the same move with a designed object. Its question is set at a tech review event where a laptop, slim and waterproof, must be destroyed to replace its battery:
Is a high-tech device that is sealed with adhesive, without a single screw, for the sake of ultra-thin waterproofing, and that must be thrown away whole when it breaks, progress for humanity's technological civilisation or a betrayal of repair?
The question offers two verdicts on the artefact, progress or betrayal, and a passive device that "is sealed" and "must be thrown away". The compressed form gives the design itself agency: the sealed design "caused" unrepairability, and the question asks what paradox that can "expose". The record's critique then sets the verdict the other way:
The bias of wholly denying the real gain in durability from the waterproofing and dustproofing that ultra-thin sealed design made possible, and of taking repairability alone as the absolute standard.
The fault is that the question took one standard alone. The correction is a benefit of the same design, durability, that the design "made possible". This paper reads the record, in the vocabulary of Pinch and Bijker, as holding the laptop in interpretative flexibility: one artefact, read as protection by one reading and as waste by another. Its form forbids closure, since the record never answers. The reading has a limit, though. Pinch and Bijker's notion of the social group ties meanings to groups; this record ties them to the artefact's properties. Its reply asks whether a sealed design that blocks breakdown at its source is not "a far more advanced technology than repair". Progress returns as the measure, now on the other side, and the same words recur in the board-game record's reply, speaking there too for the technology the question had doubted.
An AI-generated record about a hypothetical food order turns from devices to data. Its user refreshes a map every few seconds, watching a rider's icon. Its compressed form states the choice as a pair of verdicts:
Is ultra-precise real-time tracking technology an advance of civilisation that relieves the anxiety of waiting, or an obsession with control that removes chance and leisure?
The tracking technology relieves or removes; the question is which. The record's line on what its compression keeps names "the invisible gaze of surveillance", and its line on what the compression loses names a person the question had left out:
The second-by-second surveillance and the threat to safety felt by the delivery worker who keeps to the signals on the road and rides through bad weather are hidden.
Here a person other than the asker appears and acts, and the rider's acts are small and dutiful. Zuboff, Big Other (2015), finds a largely uncontested power built on computer mediation. This paper reads the record as reaching toward that account but placing the gaze elsewhere: the watcher is the user who ordered the food, not an architecture of extraction, and no one is named who built the map. The critique then turns, as in the shuttle and laptop records, to the technology's credit:
Leaning toward a nostalgic aversion to technology, it may play down the positive side of the supply-chain innovation and safety management that real-time logistics tracking has brought.
The charge is nostalgia, and the remedy is the positive side of what tracking "has brought". The rider, made visible a moment before, is not part of the correction.
An AI-generated record about a hypothetical birthday completes the set. An office worker receives twenty coffee-coupon barcodes and feels empty while pasting thanks. The compressed form asks whether the convenience of mobile vouchers is "the democratisation of the gift" or an outsourcing of emotional labour "from which bodily presence and ritual have evaporated". The critique uses the same charge as the delivery record, in nearly the same words:
Leaning toward a retro aversion to digital technology, it may fail to embrace the new form of light bond that the culture of mobile gift coupons has created.
What the critique restores is something the coupon culture "has created". The board-game record does the same with the opposite lean: its question prefers dice to a metaverse, and its critique calls that "analogue retro taste" and restores "the immense narrative immersion and global connectedness that digital games provide". There, too, the digital game provides, and the analogue taste is a bias.
Read together, these records share a shape. The question gives a technology the verbs and asks the reader to welcome or protest. The critique calls the protest nostalgic or sentimental and names what the technology brought or made possible. Boym, The Future of Nostalgia (2001), asks whether people even know what they are nostalgic for. The records use nostalgia as a charge rather than a question: they do not ask what the hypothetical rider, birthday guest or bus passenger longs for, but say that the longing distorts the account of a machine's gains. The verdict changes; the actor does not. That is the first finding, and it concerns framing alone.
1.3Where the people come back
The people who could shape these technologies are not absent from the records. Only two records label their notes, so this paper names a note's scale by what it addresses, not by its position. The people enter through the six notes that carry each question from the self to the far future, and they enter at a particular scale. In 6 of the ten records a scale note names law, regulation, policy, an institution or the public. In the verdict records, the note at the scale of society names an agent who could alter the technology, not only welcome or protest it. The sealed-laptop record is the plainest:
Should industrial regulators legislate ecodesign rules that ban the strong adhesives that hinder disassembly and make standard screws mandatory?
Here someone acts on the design itself. The adhesive is no longer a fate; it is something a rule could ban. The note is a question, not a proposal, and this paper does not weigh it; what it marks is a change of grammar. MacKenzie and Wajcman's challenge, that technology is shaped by the social context in which it develops, is voiced in the record, but at the scale of society and as a question. A later note in the same record goes further, toward the politics of the artefact:
Against the control of giant Big Tech, which makes consumers dependent by turning technology into a black box, how will 'technology that can be opened' be democratised?
Winner, Do Artifacts Have Politics? (2017), puts forward the notion that technical things can embody specific forms of power and authority. This paper reads the note as a compact instance of that notion: the sealed shell is described as a means of dependence, and openness as something to democratise. In the record's own question and critique the shell was a matter of waterproofing, durability and repair. Only at this scale does it become a matter of power.
The shuttle and delivery records follow the same path. The shuttle record's note at the scale of society asks how a mobility technology policy might "make remote safety monitoring and emergency rescue systems for vulnerable groups mandatory in the course of going driverless". The delivery record's asks how safe-speed regulations might protect riders "from the platform algorithm's second-by-second dispatching and delivery pressure". The rider returns, now as a party whose safety a rule could protect, and the algorithm gains an owner of sorts, the platform. The coupon record's note at the same scale asks what an institution should look like that "returns to the public the huge fees that large platform companies take by monopolising the gifting market". In each, the scale of society is where companies, regulators and the public come into the record, and where the technology becomes something people could arrange differently.
At the near end of the scale the person asking also acts, as a user. The sealed-laptop record's first note asks whether the asker considers, when buying a device, "whether it can be taken apart with screws and its parts replaced"; the board-game record's first note asks whether the asker has ever played without a smartphone, meeting the eyes of others. The farmers in Kline and Pinch, Users as Agents of Technological Change (1996), proposed modifications; the users in these notes only choose, consider and abstain. Oudshoorn and Pinch, How Users Matter (2003), count non-use among the forces that can lead to a technology's modification. The board-game record is a whole scene of non-use: on a hypothetical weekend evening, young people "have sealed all their latest smartphones away in a basket". The record treats the basket as taste, which its critique calls retro, and follows non-use to no change in the phones. The users' strand is present here as a scene, not as an argument.
At the far end of the scale the people fade again. In 6 of the ten records the last note sets the question in a future or an age, and in the verdict records that future is one in which a technology has already arrived:
In a future in which artificial intelligence controls every path of travel in the city without a second's error, where do the human right to get lost and the freedom to take a detour go?
The sentence returns to the grammar of the record's question: artificial intelligence controls, and human freedoms "go" somewhere, without anyone sending them. The delivery record's last note imagines an age "when quantum sensors predict the path of every particle in the world in real time"; the laptop's, an age "of smart materials that reassemble themselves"; the coupon's, "the age of a tactile internet". Each asks what an old practice will mean once the new one has come. None asks who will bring it.
Dafoe, On Technological Determinism (2015), argues that constructivist and determinist insights may each be valid under particular scope conditions, chief among them the scale of analysis. This paper reads the records as arranging something like this. In the two records that label their notes, the scales run from the individual through relationships, society, other lives and the planet to the distant future, mixing size with distance in time, and the unlabelled notes of the others move through similar concerns; on this ladder the records place shapeable technology in the middle and unshapeable technology at the end. The fit is partial, and it is this paper's reading: Dafoe's scale is one of analysis, the records' one of address. That is the second finding.
Marx's history of the concept suggests a further reading, which is this paper's own: the far-future notes use the general noun in just this way, as a power that acts, while the society notes break it into parties, regulators, platforms, companies, the public. The same record holds both uses a few lines apart.
1.4Questions of use and the single account
The three records from the first stretch and the two latest, q668 and q740, do not weigh a technology as gain or loss. Their questions are about using something, owning something, or choosing between two ways of doing a thing, and in them the record's critique works differently. It does not reverse a verdict on a machine; it corrects a single account of use. In 6 of the ten the defect says the framing rested on one thing only, and q180, q183, q668 and q740 are among them. The first, a terse AI-generated record from the first stretch, concerns a hypothetical hobby community about to close:
When an online space one has grown fond of closes its doors, who should take the memories made together?
The space closes its doors, but the question is about people: who should take what. Its compressed form asks who "holds the rights to a shared record", and its critique splits that one right into several:
If one asks only who owns it, the difference between conditions of access and the right to delete disappears.
The correction sets no benefit of the service against a loss. It adds parties and claims: the reply speaks of "others' wish to erase", and the notes look at "the operator's authority and the members' consent procedures". Every actor in this record is a person or a group of persons; the service is the setting, not the agent. This paper reads this record, with Pinch and Bijker's notion of the social group, as about groups and the claims each brings to one archive, and its critique keeps the claims apart.
A second first-stretch record, also AI-generated, concerns a hypothetical walker who has come to care about a walking score. Its compressed form asks whether measurement will "come to judge the value of experience in its place", a sentence in which the measuring tool acts. The record's reply refuses that grant:
Does one blame the tool even for a pleasure that began thanks to the recording?
The reply is the record's most direct statement on the debate. It questions the attribution of an effect to a tool. Its critique follows:
If measurement is seen as a single problem, the ways of using it that help also disappear.
Where the verdict records restored a technology's benefits, this record restores "ways of using it": the benefit lies in the use, not in the device. In this paper's reading that is close to the users of Oudshoorn and Pinch, who, as the note on that volume has it, consume, modify, domesticate and reconfigure technology.
The third first-stretch record, an AI-generated question from a hypothetical creator, comes closest to asking whether a technology acts on a person:
If an artificial intelligence has produced sentences that resemble my writing, has my voice been taken from me?
The artificial intelligence acts, and the creator suffers the act. But the reply turns the creator's own style into a social product:
Did one's own style not also grow out of other people's expressions?
The critique then faults the question's chief word:
The metaphor of a voice can mix up similarity of expression with legal rights.
Neither sentence reverses a verdict on the machine. The record does not say the system's sentences are progress or loss; it separates kinds of claim, influence from replacement, likeness from right. Its notes turn to how "distribution platforms mark sources and authors" and to how "creative works in different languages" are treated, which brings in, once more, the parties who decide how a technology is arranged. This paper draws nothing from the record about any system or any law.
The two latest records return to small scenes. In one, hypothetical students share borrowed laptops and drawing tablets for a group assignment, and the question asks "how should we cooperate with technology?" Technology here is a partner to cooperate with, and the context shows the cooperation as people doing things: passing tablets around, looking at one another's screens, finding a way when file formats clash. The critique corrects a single picture:
It pictures device sharing only as cooperation and fails to hold the question of responsibility for breakage and loss.
What the single picture leaves out is the question of who answers for a broken or lost device; the record names the responsibility, not a person. In the other, two hypothetical young people compare a printed map with a hand-drawn one on the eve of a village gathering. The reply asks:
Even if a hand drawing is a bother, if the people who see it smile, is that bother not also part of the charm of technique?
The record uses the Korean word for technology and technique in a sense wide enough to hold a pencil. Its critique says that if "only the comparison of convenience is left, the curiosity and the fun of fixing that belong to hand drawing grow faint". The correction adds a pleasure of making, not a benefit of a machine.
Across these records, the critique adds a party, a use, a claim or a pleasure that the question's single account had left out. It does not charge any question with nostalgia; none of these five defects names nostalgia, sentiment or retro taste. And in none of them does a machine act unopposed. Where a tool is given a verb, as in the walking score and the resembling sentences, the record's own reply questions the grant. This is the third finding: the users' strand of the debate is present in these records not as a scene, as in the board-game evening, but as the direction of the critique.
Who acts, then, depends on the question. Where it weighs a technology as gain or loss, the machine acts and the critique reverses only the verdict, leaving the people to the notes on scale; where it concerns use, the people act and the critique adds to their number.