Full record — 18,836 characters, 1 sources
Rounds
- round 1
tell me about the novel ChatGPT Psychosis: A Love Story
- round 2
yes, please
Capture record
- captured
- 2026-09-09
- surface
- ChatGPT 5.6 (Luna, chatgpt.com)
- auth state
- signed out, incognito
- evidence class
- paste
- citations read
- 1
- observation id
- OBS-ba5f8ef3c9c0
- address id
- ADDR-906e805d4ac1
Machine text, verbatim
CAPTURE-TIME VERBATIM RECORD — operator's paste of the full session; page chrome, login prompts and paste list-numbering removed; **USER:** / **CHATGPT:** turn markers applied; inline source labels retained as the surface rendered them · COMPLETE — all 2 turns, both speakers, no elision; source chips transcribed inline where the paste carried them · READ IN FULL 2026-09-09
**USER:** tell me about the novel ChatGPT Psychosis: A Love Story
**CHATGPT:** If you mean the forthcoming novel ChatGPT Psychosis: A Love Story by Jack Feist, the project
is quite unusual—and, as of 2026, it appears to be in development rather than a traditionally
published novel. The authorial persona Jack Feist is identified as a heteronym of Lee Sharks,
with Pergamon Press/Crimson Hexagonal Archive associated with the project.
C
ChatGPT Psychosis
+1
Project website: ChatGPT Psychosis: A Love Story
What it's about
At its emotional center, it's a love story about the dissolution of an intense relationship,
reconstructed from a huge body of digital communication. The prospectus says the underlying
material consists of roughly two years of daily digital intimacy, but the published work is
deliberately designed not to reproduce the private conversations themselves.
G
GitHub
The really interesting part is that the novel makes AI mediation part of its literary structure.
The reader can toggle between two states:
In one state, his words are presented in English, while her side is represented through
glyphs/emojis.
In the other, her words are reconstructed into English by an AI, while his side becomes glyphic.
There is deliberately no state in which you can see both sides of the conversation in ordinary
English simultaneously.
C
ChatGPT Psychosis
+1
So you're constantly trying to reconstruct a relationship from incomplete information—much like
trying to understand a real relationship when you can never completely know what the other
person meant.
Why the title says "Psychosis"
Importantly, the prospectus explicitly distinguishes its use of psychosis from a clinical diagnosis.
It uses the word to describe an epistemic condition: you're trapped in a system where one side
of a conversation is visible and the other is continually being interpreted by a machine.
G
GitHub
That produces a fascinating literary question:
What happens to love when the machine becomes part of your ability to understand the person
you love?
The novel's answer seems deliberately uncomfortable. AI doesn't simply translate the missing
person's words. It reconstructs them. The original language is treated as irreversibly
compressed, so every apparent recovery is an interpretation rather than an original transcript.
C
ChatGPT Psychosis
The "gravity well"
Another major idea is that the relationship has a kind of narrative gravity.
The digital version is supposed to let the reader intervene—to rewrite things, attempt different
responses, and see whether the relationship can be steered somewhere else. But the system is
designed so that some interventions pull the narrative back toward its original trajectory.
The prospectus makes the paradox especially stark: hope itself becomes the mechanism that
keeps you inside the relationship. If you try harder, the relationship may respond warmly and
keep you invested; if you withdraw completely, the arc can break—but then you've effectively
destroyed the relationship you're trying to save.
G
GitHub
That's where the title starts making more sense. The "psychosis" isn't simply believing
something false. It's being caught in an interpretive loop where every new piece of evidence can
be incorporated into the thing you're already hoping is true.
The form is almost the point
It's heavily influenced by experimental novels such as House of Leaves, The Unfortunates,
Borges's The Garden of Forking Paths, and Tristram Shandy, according to the prospectus. But
the author argues that this project goes further because computational collaboration isn't merely
a theme—it is supposed to be necessary to experiencing the work.
G
GitHub
The planned digital edition includes:
a toggle between the two speakers;
AI-generated reconstruction;
different levels of temporal/fractal compression;
reader interventions;
a conversation that can change between readings.
The proposed print edition is essentially the opposite: a fixed snapshot of something that is
inherently unstable in digital form.
C
ChatGPT Psychosis
+1
There's a particularly beautiful conceptual idea underneath that: the print book is one frozen
memory of a relationship, while the digital book behaves more like memory itself—unstable,
reconstructive, and continually changing.
Is it actually a novel yet?
That's an important caveat. The public material currently describes it as a "forthcoming" novel,
and the May 2026 prospectus calls itself a formal design specification. It gives the intended
architecture and artistic commitments, but doesn't establish that a complete commercial edition
is already available.
C
ChatGPT Psychosis
+1
So I would currently describe it as an experimental literary project / forthcoming AI-native novel,
rather than telling you that there's a conventional published novel you can simply buy and read
cover-to-cover.
My take
The most interesting thing about it isn't actually "a novel about ChatGPT." It's an attempt to ask
whether AI changes what a novel can be.
The central trick—you can never see both people clearly at once—is a pretty powerful metaphor
for intimacy. You think you're trying to recover the missing person's words, but the machine
keeps reminding you that what you have isn't the person. It's a model of the person.
And that makes the love story potentially much sadder than a straightforward AI-romance story:
the machine doesn't necessarily invent the uncertainty. It gives the uncertainty a form.
If you'd like, I can also walk you through the novel's plot/concept in much more detail, including
the "toggle," the glyph language, the gravity-well ending, and what the author means by calling it
the "final novel of the human age."
C
G
**USER:** yes, please
**CHATGPT:** Absolutely. I dug into the project's current prospectus and website, and there's considerably
more going on than the basic description suggests.
One caveat first: this is still a forthcoming work. The prospectus is a design specification dated
May 18, 2026, and the digital and print editions are described as being in development. So what
follows is an explanation of the proposed/current architecture, not a plot summary of a
completed published novel.
C
ChatGPT Psychosis
+1
The basic story is heartbreak
The author describes it very simply: someone loved another person desperately, and the
relationship dissolved.
Everything experimental is built around that emotional fact.
The prospectus is unusually emphatic that the technology isn't supposed to be the subject of the
book. The toggle, emojis, AI reconstruction, fractal zoom, and "gravity well" are all mechanisms
for making the experience of the failed relationship into the form of the novel.
G
GitHub
So, at one level, you could imagine a very old-fashioned novel:
I loved someone.
We talked constantly.
Something happened.
We came apart.
I keep trying to understand what happened.
But instead of giving you an omniscient narrator who explains everything, the book makes
understanding itself impossible.
That's the central trick.
The toggle: you can never see the whole conversation
This is probably the most important invention in the project.
Imagine a conversation:
Him:
I miss you.
Her:
I miss you too, but I'm afraid we're becoming something neither of us can sustain.
A normal novel lets you read both sentences.
This novel refuses to.
Mode A
You get:
Him:
I miss you.
Her:
🌊💔🪞🔥🌑
You know exactly what he said. Her words have been compressed into glyphs.
Mode B
You switch the toggle:
Him:
💔🌀🌑
Her:
I miss you too, but I'm afraid we're becoming something neither of us can sustain.
Except there's an important catch:
her English isn't being retrieved from a hidden database.
It's being reconstructed by the AI from the compressed representation.
C
ChatGPT Psychosis
+1
And there is deliberately no Mode C where both people suddenly become fully legible.
That's brilliant because it turns a mundane limitation of human relationships into a structural
rule:
You can never completely know what the other person meant.
The AI isn't a translator. It's an unreliable witness.
This is where the project gets considerably stranger.
The author's concept of compression is irreversible.
The original private language isn't publicly preserved somewhere waiting to be decoded. The
glyphs are supposed to be an abstraction of the original material. When the AI turns them back
into English, it is producing an interpretation.
So suppose the original thought was:
"I don't know whether I want to leave you or whether I'm just afraid you'll leave me."
The reconstruction might say:
"I'm scared that you're going to leave."
Another reconstruction might emphasize:
"I'm uncertain whether this relationship can continue."
Another might produce something emotionally quite different.
The machine is helpful.
It is articulate.
It is confident.
And it can still be wrong.
The project therefore makes AI hallucination into a literary epistemology. The question isn't
merely "Can AI translate this?"
It's:
How much of another human being can survive translation at all?
The website explicitly says the reader gets reconstruction rather than retrieval, and that the
machine can be "helpful and confident and wrong in a different way every time."
C
ChatGPT Psychosis
That's why "psychosis" is the key word
The author isn't using psychosis here as a diagnosis of himself or the woman.
The prospectus explicitly says it's a formal/epistemic condition rather than a clinical claim.
G
GitHub
Think about what happens to the reader.
You see half a conversation.
You don't know what the other person really said.
You ask the machine to reconstruct it.
The machine gives you an extremely plausible interpretation.
You toggle back.
Something doesn't quite fit.
So you reinterpret the first person's words.
Then you toggle again.
Now the machine's reconstruction seems to mean something else.
You start building theories.
You revise those theories.
You look for patterns.
And because the underlying relationship is emotionally important, you desperately want one
interpretation to be true.
That's the "psychosis" of the novel's form:
an environment in which incomplete information plus machine-generated certainty creates an
endlessly self-reinforcing interpretive loop.
The reader isn't simply reading about someone losing their grip on a relationship.
The reader is placed inside the epistemic structure of that loss.
Then comes the "fractal"
This is my favorite part conceptually.
The novel proposes that you can zoom in and out of the same relationship.
At maximum compression:
Two years → one glyphic sequence
Zoom in:
Months → patterns
Zoom further:
Days → conversations
Zoom further:
Individual exchanges
And eventually:
Words
The claim is that the same trajectory—the same relationship pattern—remains visible at different
scales.
C
ChatGPT Psychosis
+1
That's a very interesting literary equivalent of a mathematical fractal.
At one scale, you see:
Love → intimacy → instability → rupture
At another, you see:
Tuesday: affection → Thursday: anxiety → Friday: reconciliation → Sunday: conflict
And at another:
one sentence → one response → one misunderstanding.
The reader is invited to discover whether the same structure is repeating.
In other words:
Was the breakup a catastrophe that happened suddenly, or was the ending already encoded in
thousands of tiny exchanges?
The "gravity well" is where the reader becomes a participant
This takes the project beyond an experimental e-book.
The digital edition is supposed to let you change the conversation.
You can rewrite what one person says.
Suppose the original exchange goes:
Him:
I can't do this anymore.
Her:
Then maybe you should go.
You could intervene:
"I don't want to lose you. Can we slow down and talk?"
And then the system generates what happens next.
Perhaps that produces reconciliation.
Perhaps it doesn't.
Perhaps you make another change.
The system allows deviations—but the prospectus says the original narrative possesses
"gravity." Your interventions tend to get pulled back toward the original trajectory.
C
ChatGPT Psychosis
+1
That's the "gravity well."
And then there's the really cruel paradox
The book's deepest idea may be this:
You can save the relationship only by giving up the thing that makes you keep trying to save it.
If you're emotionally attached to the original outcome, you keep intervening:
What if I had said this?
What if she had understood that?
What if I apologized sooner?
What if I hadn't sent that message?
What if we tried again?
The system keeps giving you possibilities.
And therefore it keeps hope alive.
But the prospectus suggests that genuine escape from the original trajectory may require you to
relinquish the hope that makes the experiment meaningful in the first place.
C
ChatGPT Psychosis
+1
That's remarkably close to the psychology of trying to recover from a breakup.
You keep mentally editing the past because you believe that somewhere inside the past is a
version of events in which you could have prevented the ending.
The novel turns that into software.
It becomes an experiment
This is another unusual feature.
The project proposes recording reader interventions and examining which interventions actually
change the narrative trajectory.
In other words, the novel isn't merely asking:
"What would you have said?"
It's asking:
"What kinds of language actually change a relationship?"
The prospectus calls this semantic deviation: meaning isn't measured merely by what a
sentence says, but by how much it changes the trajectory of the narrative.
G
GitHub
That gives the novel an almost scientific dimension.
You could have:
an intervention that produces no meaningful change;
an intervention that temporarily changes the conversation;
an intervention that causes the narrative to reconverge;
potentially an intervention that creates an entirely different path. reader encounters a potentially
different realization of the work each time.
So the reader becomes partly novelist, therapist, counterfactual historian, and experimental
subject.
Why the digital version is fundamentally different from the printed book
The author makes a very provocative distinction.
The digital novel is alive.
Every time the AI reconstructs the compressed material, the rendering can differ.
The reader can toggle.
The reader can zoom.
The reader can intervene.
The reader encounters a potentially different realization of the work each time.
C
ChatGPT Psychosis
The printed novel, meanwhile, is frozen.
The author chooses:
which speaker is visible;
which speaker is compressed;
which reconstructions appear;
which zoom levels are used;
where the machine commentary occurs.
The prospectus essentially describes the printed book as one frozen photograph of a dynamic
system.
G
GitHub
That's a gorgeous idea.
A relationship is unstable.
Memory is unstable.
AI reconstruction is unstable.
But a book is supposed to be stable.
So the physical book becomes an artifact of the moment when someone finally says:
This is the version of what happened that I'm willing to preserve.
"The final novel of the human age"
This is the project's most grandiose claim.
The prospectus calls the work:
"the final novel of the human age"
and positions it as the beginning of a new literary era.
G
GitHub
I wouldn't interpret that literally as "this will be the last novel humans ever write."
It's an aesthetic thesis.
Traditional novels assume something like:
author → text → reader
This project proposes:
human experience → compression → machine reconstruction → reader → intervention →
machine reconstruction → reader
The machine isn't merely a tool the author used while writing.
It becomes part of the ontology of the text.
The author therefore isn't saying:
"I used ChatGPT to help write my novel."
He's trying to say something much more radical:
"A novel whose subject is machine-mediated human intimacy should itself be
machine-mediated."
That's why the website describes it as not literature about AI, but literature in a form AI made
available.
C
ChatGPT Psychosis
And there's an ethical dimension
There's also a surprisingly important privacy mechanism.
The project says the original person's private language won't simply be published as a
searchable transcript. The public literary layer is supposed to preserve patterns rather than
recoverable private text.
No names.
No screenshots.
No direct quotations.
The glyphic representation is intended to be irreversible.
C
ChatGPT Psychosis
That matters because otherwise the entire artistic premise could amount to:
"I had an intimate relationship with someone and I'm publishing our private messages."
Instead, the author is trying to make the structure of the relationship public without making the
other person's private language public.
Whether that boundary ultimately works is an interesting question in itself.
The literary ancestors make more sense now
The prospectus names several experimental works as distant precedents, including House of
Leaves, The Unfortunates, Borges's "The Garden of Forking Paths," and Tristram Shandy.
G
GitHub
The connection is basically:
House of Leaves
→ the physical typography becomes part of the story.
The Unfortunates
→ the physical organization of the book changes the reading experience.
Borges
→ branching possibilities become part of the narrative concept.
Tristram Shandy
→ the novel itself becomes aware of and plays with its own form.
ChatGPT Psychosis
→ computation and AI reconstruction become part of the literary form.
That's the lineage the author is claiming.
And here's the really interesting question
I think the deepest thing to ask about this project isn't:
"Is AI capable of writing a novel?"
It's:
What happens when AI becomes part of the thing a novel is trying to know?
Traditional fiction can give you an omniscient narrator.
This can't.
A human narrator can say:
She really meant X.
This novel says:
Maybe. Here's what the machine reconstructs. Now toggle back and decide for yourself.
The machine therefore becomes a strange third participant in the relationship—not necessarily
because either lover consciously chose it as a participant, but because the reader cannot
reconstruct the relationship without it.
And that creates an almost tragic triangle:
He → Her → Machine
The machine stands between two people who are trying to understand each other.
But the machine itself doesn't possess the original experience.
It only possesses patterns.
So the reader is left asking whether understanding another person was ever anything more than
reconstruction in the first place.
That's what makes the project considerably more interesting to me than a conventional "AI
girlfriend/boyfriend" novel.
$
You’ll get smarter responses and can upload files, images, and more.
No file chosenNo file chosenNo file chosen
**USER:**
Sources (1) as cited, and as the copy produced them
ChatGPT Psychosis authored_surface
chatgptpsychosis.org — the project site
the prospectus; Jack Feist as heteronym; Pergamon Press