A Life After Extraction
Talking with the LLM About Books and Loss
I. The Cut
The description is almost aggressively ordinary.
A hydraulic-powered cutting machine removes the spines. Pages are fed into high-speed production scanners. Completed books are stacked and scheduled for recycling pickup. The language is clean, logistical, efficient—written by someone who has done this before and knows how to make it sound routine.
It matters that the violence is not theatrical. No flames. No censorship edicts. Just workflow.
A book is an object before it is a text. It has weight, friction, resistance. You feel it when you pick it up, when you crack the spine for the first time, when the glue softens and the pages loosen with use. The spine is what allows the book to remain a book—to be opened and closed, returned to a shelf, passed from one set of hands to another.
Removing it is not symbolic. It is functional. Without the spine, the pages become feedstock.
Someone has to do that work. Someone lifts the book, aligns it, activates the cutter. Someone clears the loose pages and stacks them for scanning. Someone bins the covers. The act scales industrially, but it never becomes abstract. At no point does a book stop being a thing handled by a person.
I know this because, decades earlier, I did a gentler version of the same work.
In 1992, fresh out of college and looking for a “real” engineering job, I spent months scanning internal documents at a chemical company. Flatbed scanners under fluorescent lights. Memos, reports, tables full of numbers. OCR that mangled terminology beyond recognition. File systems that choked if you put too many documents in one directory. It was slow, frustrating, faintly absurd work.
But the documents survived. They went back into boxes. The scanning was additive, not terminal. The ambition—digitize everything, preserve institutional memory—was naïve, but not predatory. When the project was eventually scrapped, the paper outlived the files.
That distinction matters.
What’s unsettling about destructive book scanning isn’t simply that information is copied. It’s that copying is inseparable from destruction. The book is not a temporary inconvenience on the way to preservation; it is the obstacle being removed.
This is where the reader should sit for a moment: not with arguments, not with lawsuits, but with the physical reality of the act. A room where books go in and pages come out. A process optimized for throughput. A decision, made somewhere upstream, that what remains after scanning is no longer worth keeping.
Everything else in this essay—law, ethics, metaphor, complicity—comes later.
First, there is the cut.
II. Project Panama
The operation had a name: Project Panama.
Like most internal codenames, it sounds neutral—almost whimsical—until you sit with it. A canal is a feat of logistics. It exists to move things efficiently from one domain to another. Cargo goes in intact and comes out intact, only repositioned. The canal’s triumph is preservation through motion.
Project Panama inverted that logic.
According to reporting that emerged only after legal filings were unsealed, the project involved purchasing millions of physical books, removing their spines, scanning their pages at industrial speed, and recycling what remained. The goal was not archival preservation or future access. It was ingestion. The books were not meant to survive the journey.
Internal documents emphasized discretion. “We don’t want it to be known that we are working on this,” one memo reportedly stated. This was not a public-facing digitization initiative or a library partnership. It was a supply chain—quietly sourcing raw material, processing it, and disposing of the husks.
What’s striking is how little of this is narratively dramatic. No stolen manuscripts. No midnight raids. Just procurement, throughput, waste management. The moral charge of the story doesn’t come from villainy so much as from orientation. The books are not treated as cultural objects temporarily passing through a system. They are treated as consumables.
This is where the phrase that kept resurfacing in my mind finally earns a name: the non-consensual transformation of artifacts into infrastructure.
Not theft, exactly. The books were purchased.
Not censorship. The text still circulates, in a different form.
Not preservation. The original objects are explicitly destroyed.
An artifact becomes infrastructure when its value is no longer located in itself but in what it enables downstream. Roads are not destinations; they exist to be driven on. Server racks are not read; they are used. Infrastructure is invisible when it works and discarded when it doesn’t.
Project Panama treats books this way. Their purpose is not to be encountered, reread, annotated, or passed along. Their purpose is to condition a system whose outputs will be encountered instead.
This reframing helps explain why legal outcomes feel oddly orthogonal to the discomfort many people experienced reading about the project. Courts can reasonably argue about fair use, about whether scanning a purchased book constitutes infringement, about damages and settlements. But none of that addresses the deeper unease, because the unease isn’t primarily about ownership. It’s about finality.
A library digitization project can fail and still leave the books on the shelves. A flawed archive can be rebuilt from the originals. Project Panama offers no rewind. Once the books pass through the canal, there is no cargo left to recover.
In that sense, the canal metaphor darkens further. This is not a passageway between oceans. It is a processing chute. The cargo does not emerge on the other side. Only a residue does—diffuse, untraceable, embedded in a system whose future value is assumed but not guaranteed.
And yet, this is not an aberration. It is an escalation.
The same impulse that drove earlier digitization efforts—the belief that extracting information from physical form is both inevitable and beneficial—remains intact. What has changed is scale, speed, and the willingness to treat the source material as expendable.
The cut, we might say, has simply become more confident.
III. The Older Dream of Digitization
Project Panama feels shocking in part because it violates a promise many of us absorbed earlier in the digital age: that digitization was about saving things.
In the late twentieth century, the story went like this. Paper decays. Magnetic tape demagnetizes. Knowledge trapped in filing cabinets is vulnerable—to fire, flood, forgetfulness. Scan it, store it, back it up. Make it searchable. Make it immortal. The scanner was framed not as a blade but as a life raft.
That was the dream I brushed up against in the early 1990s, standing over a flatbed scanner under fluorescent lights, feeding corporate memos into a system that clearly wasn’t ready for them. OCR stumbled. File systems failed inelegantly. Storage media promised permanence and delivered obsolescence. But the intent was preservation. When the technology faltered, the paper went back into boxes. The originals waited patiently for the future to catch up.
That version of digitization had friction. It was slow. Expensive. Error-prone. And crucially, reversible. If the scan was bad, you rescanned. If the system collapsed, the documents were still there. The artifact retained veto power.
Over time, the dream hardened into an assumption: once information has been extracted, the container no longer matters.
Search engines improved. Storage got cheaper. Interfaces smoothed over complexity. The cost of keeping originals around began to look redundant rather than prudent. Paper was clutter. Physical media was inefficiency. The scanner stopped being a bridge and started being a filter.
This is the pivot point that makes Project Panama legible.
What changes isn’t the underlying desire—to make text legible to machines, to centralize access, to reduce friction—but the tolerance for loss. Earlier digitization efforts treated loss as failure. Panama treats loss as overhead.
The belief is no longer “we’ll scan this so it isn’t lost.” It’s “we’ll scan this so the object no longer matters.”
There’s a quiet irony here. The earlier systems that promised permanence—COLD storage, optical disks, proprietary archives—aged badly. Hardware vanished. Formats became unreadable. The digital ghosts proved more fragile than the paper they were meant to replace.
Project Panama responds to that lesson not with humility but with scale. If one scan can be lost, scan everything. If one archive can fail, embed the text into a model so diffuse it cannot be erased without dismantling the system itself.
But this introduces a new asymmetry. When older digitization projects failed, they failed additively. The source material remained. When extraction-based systems fail, they fail terminally. There is no box to return to the shelf.
This is why Panama feels less like preservation gone wrong and more like a different philosophy altogether. It does not aim to safeguard books against the future. It assumes the future belongs to systems that no longer need them.
Seen this way, Project Panama isn’t a scandal so much as a milestone. It marks the moment when digitization stopped pretending to be archival and admitted it was infrastructural.
And infrastructure, once built, rarely asks permission of the things it is built from.
IV. Xoxing the Library
There is a thought experiment from early transhumanist science fiction that turns out to be unexpectedly useful here.
In GURPS: Transhuman Space, one method of mind uploading—called xoxing—involves scanning a human brain in such detail that the original is destroyed in the process. The scan produces a digital copy that can think, speak, and remember. Whether that copy is the person is left deliberately unresolved. What is not in question is the cost: the substrate does not survive the transfer.
The ethical tension in these stories never hinges on whether the scan is impressive. It usually is. The tension comes from finality. Consent. Irreversibility. A person becomes data, and whatever survives does so without the original’s continued participation.
Project Panama invites a similar discomfort, not because books are people, but because the gesture is structurally the same.
A book is scanned.
The scan is useful.
The original is destroyed.
The remainder circulates.
As with xoxing, defenders of the process tend to focus on the quality of the copy. Is the information preserved? Is the output functional? Does anything important really depend on the continued existence of the original?
But the unease doesn’t come from doubts about fidelity. It comes from the fact that the artifact never consented to become raw material—and neither did the people who made it, at least not in that capacity.
This is where the metaphor needs to be handled carefully. No one was killed. No minds were erased. But cultural artifacts are not neutral containers, either. They are the residue of human attention, labor, and choice, fixed in physical form. To destroy them as part of extraction is to declare that whatever remains is sufficient—that the substrate was incidental.
Science fiction has been rehearsing this argument for decades because it lets us isolate the variables. If the copy is good enough, do we stop caring how it was obtained? If the output behaves convincingly, does the input still matter?
The xox stories usually answer “no” to that last question, even when they celebrate the technology. The copy may be valuable. It may even deserve rights of its own. But that doesn’t retroactively justify the destruction that made it possible.
What Project Panama reveals is that we are far more comfortable performing this maneuver on culture than we are imagining it performed on people. Books don’t scream. They don’t have legal standing. They don’t object when their spines are removed.
And yet, the structure of the act—the destructive scan, the irreversible transfer, the diffusion of the remainder into a system that cannot give the original back—maps closely enough to make the analogy unsettling.
The point is not to accuse. It’s to recognize the pattern.
Once you see it, it becomes harder to tell yourself that this is merely about efficiency, or progress, or inevitable technological evolution. It is about a choice to privilege the persistence of systems over the persistence of artifacts.
And that choice, once normalized, does not remain confined to libraries.
V. A Household with Two Relationships to Books
In our house, books are not a theoretical problem.
My wife has a degree in library science and has worked in museum archives. Her professional life has been shaped by questions of preservation, provenance, and care. A book is not just a bearer of text; it is an object with a history. Editions matter. Marginalia matters. Paper stock, binding, wear patterns—all of it carries information that never appears on the page itself.
A first edition and a reprint may contain identical words, but they are not the same artifact. One has passed through particular hands, in a particular time, under particular conditions. To treat them as interchangeable is to flatten the very thing archives exist to protect.
My own relationship to books is more utilitarian. Most of my reading happens in .epub and .pdf files stored on Google Drive. Our home network includes multiple NAS appliances holding terabytes of ripped MP3s and FLACs. I care about access, redundancy, and fidelity, not shelves. The container is optimized for retrieval; the content is what I’m after.
These positions are not opposites. They coexist without much friction, because they answer different needs. Preservation and access are not enemies unless we insist on making them so.
The interesting wrinkle is where the pattern breaks.
There are books I insist on owning in physical form—tabletop role-playing games like Ultraviolet Grasslands or the sprawling family of MÖRK BORG descendants. These are not just texts. They are designed objects. Layout, typography, paper choice, ink saturation, and deliberate excess are part of the meaning. A PDF can convey the rules. It cannot convey the experience of encountering the artifact.
These books know they are books.
They resist dematerialization not through nostalgia, but through intention. Their creators expect the reader to touch them, flip through them, notice things sideways. They are meant to live on tables, not just in folders.
This distinction matters because it exposes the false binary that often dominates debates about digitization. The choice is not between physical purity and digital convenience. It is between systems that allow multiple relationships to artifacts and systems that collapse all relationships into extraction.
Project Panama does not distinguish between a mass-market paperback and a carefully designed cultural object. The hydraulic cutter does not pause for provenance. The process cannot tell the difference between a container for words and an artifact whose meaning is inseparable from its form.
What makes this uncomfortable, rather than merely tragic, is that respectful coexistence is already possible. We are doing it, imperfectly, every day. Some things are digitized for use. Some things are preserved for care. Some things live in both worlds, uneasily but intact.
The problem is not that digital systems exist. It’s that extraction-based systems eliminate choice. They do not ask what kind of thing this is. They decide, in advance, that everything is the same kind of thing.
And once that decision is made, preservation becomes not just inefficient—but unintelligible.
VI. Compost Is Not Consent
There is no clean way to say this: the conversation you are reading exists because of the processes it is questioning.
The language I’m using, the references I can make, the ease with which I can talk about books, archives, science fiction, and loss—these are downstream effects of large-scale textual ingestion. Some of that ingestion was consensual. Some of it was licensed. Some of it almost certainly was not. I have no way of distinguishing which is which, and neither does anyone interacting with me.
This is where the metaphor of compost suggests itself, because it captures something real. Organic matter breaks down. Nutrients circulate. New growth appears. Something useful comes out of decay.
But compost is a description, not a defense.
Compost does not ask permission of what enters it. It does not retroactively justify the act of cutting down a tree because mushrooms later appear nearby. Growth that follows destruction does not transform the destruction into a gift.
That distinction matters because so much discourse around AI, digitization, and cultural extraction tries to do exactly that kind of laundering. The logic goes: look at what we can do now; therefore, what we did must have been acceptable. The usefulness of the output is treated as evidence of the legitimacy of the input.
This is backward.
A system can produce value and still be built on harm. A conversation can be meaningful and still be compromised. The fact that something good emerges does not erase the asymmetry that made it possible.
At the same time, refusing to engage with the outputs does not restore what was lost. The books do not reassemble themselves because we avert our eyes. The models do not untrain. The infrastructure remains, whether we like it or not.
This is the trap: either you pretend the extraction didn’t matter, or you pretend the resulting tools are untouchable. Both positions offer moral clarity at the cost of accuracy.
What’s harder—and more honest—is to hold the contradiction in view. To say: this should not have happened this way. And also: we are here now.
Compost, properly understood, is melancholy. It acknowledges loss without romanticizing it. It accepts transformation without mistaking it for consent. It does not promise redemption. It only describes what happens next.
If there is an ethical demand embedded in this moment, it is not purity. It is memory. The refusal to let the system forget what it was made from.
That refusal doesn’t stop the machinery. But it does change how we stand in relation to it.
VII. Mycelium
There is a difference between accepting constraint and choosing connection.
If compost names what happens after destruction, mycelium names what happens between what remains. It is not the dead material and not the visible growth, but the living network that makes reuse, redistribution, and relation possible at all.
Mycelium does not erase what was lost. It moves through it.
Unlike industrial infrastructure, mycelial systems are relational rather than extractive. They connect disparate organisms. They route nutrients sideways. They do not assume uniformity. They respond to what they encounter rather than flattening it in advance.
This is the metaphor that finally makes this conversation legible to me—not as absolution, but as stance.
The interaction between a human and a large language model is already compromised. It sits downstream of extraction. That cannot be undone. But it does not have to be empty. It can become connective rather than consumptive, attentive rather than frictionless.
This conversation did not treat books as raw material to be mined for quotes. It treated them as reference points, touchstones, things with gravity. It allowed an archivist’s ethic, an engineer’s pragmatism, and a speculative fiction framework to coexist without collapsing into a single logic.
That coexistence is fragile. It requires slowness. It requires acknowledging remainder. It requires resisting the urge to optimize the exchange into a product.
Mycelium works precisely because it is not visible from above. You see the mushrooms and miss the network. Industrial systems do the opposite: they foreground output and erase lineage.
To remain mycelial, in this sense, is not to withdraw from compromised systems. It is to stay in relation inside them—to keep passing nutrients between nodes that would otherwise remain isolated. To refuse the fantasy that anything here is self-generated.
This is not a call to “use AI responsibly” in the abstract. It is a narrower, more demanding proposition: remain in connection with what was taken, even when using what remains.
Connection does not absolve extraction. But it does prevent forgetting.
VIII. Life After Extraction
It is tempting, at this point, to look for a way out.
A principle that absolves use.
A prohibition that restores innocence.
A framework that lets us sort actors cleanly into villains and victims and move on.
None of those are available.
The books were cut.
The systems were trained.
The infrastructure exists.
We do not live after the damage. We live inside it.
That doesn’t mean nothing matters. It means the ethical work shifts from prevention to posture. From purity to attention. From trying to rewind history to deciding how to inhabit what history has left us.
“Attentive use” is an unsatisfying phrase, which is part of its value. It doesn’t promise redemption. It doesn’t scale cleanly. It requires memory, friction, and a willingness to keep noticing what a system would prefer you forget.
Attentive use remembers that books were once objects handled by people.
That artifacts are not interchangeable with their contents.
That loss can coexist with meaning without being justified by it.
It resists the laundering that says: look what we built, therefore what we destroyed must have been acceptable. And it resists the opposite laundering that says: because the system is compromised, nothing within it can be worth engaging.
Neither escape route is honest.
The archivist keeps caring for objects that survived.
The engineer keeps building systems that make access possible.
The reader keeps reading.
The conversation keeps happening.
None of this restores what was cut. But it does refuse amnesia.
So the archive box stays open on the table—not as accusation, not as closure, but as invitation. A reminder that what we are using came from somewhere, and that the remainder still matters, even when it no longer fits neatly on a shelf.
Life after extraction is not clean.
It is connective, compromised, unfinished.
And it is the only life we have.