Characters Are Overrated: A Treatise Against the Tyranny of Arcs

You’ll hear it a thousand times in creative writing circles, often whispered with the reverence of sacred doctrine: character is king. Give your protagonist an arc, they say. Make them grow. Show them change. Rinse. Resolve. Repeat.

Audio: NotebookLM podcast on this topic.

Forgive me, but I’m not here for that workshop claptrap.



My writing isn’t character-driven in the conventional sense. I don’t sculpt protagonists to take heroic journeys or undergo epiphanic transformations. I’m not interested in plumbing the depths of their souls or bandaging their inner wounds with moral insight. My primary concern is the world—the philosophical or sociological structure—through which characters drift, orbit, or plummet. Sometimes they leave a mark. Often, they don’t.

Because real life isn’t narrative. It doesn’t arc. It drifts. And most of us don’t develop. We adapt. We cope. We muddle through.



Resolution, in most stories, is a parlour trick—narrative taxidermy dressed as transcendence. In reality, most encounters don’t resolve. They expire. People come and go. You cross paths with strangers who change your life—or don’t—and then vanish back into the abyss of statistical anonymity.

One of my recent manuscripts begins with a woman named Sena discovering a body by the roadside. She reports it, the authorities arrive, and the narrative follows them—until it doesn’t. It dissipates. No tidy resolution, no tight bow. Just the unfurling tedium of systemic procedure and human irrelevance. It’s not a mystery story. It’s a story with mystery in it. Big difference.

We like to pretend we’re central to our own story, each of us a protagonist in a universe scripted for personal development. But sometimes, we’re not even side characters. Sometimes, we’re scenery. Camus’ Meursault had it right: the sun matters more than your feelings, and death shows up whether you’ve had your arc or not.



Yes, some readers crave grandiosity—heroes, villains, the Great Man Theory dressed in narrative drag. Napoleon didn’t just wage war; he “struggled with destiny.” Stalin wasn’t just a paranoid bureaucrat; he was “a force of history.” These are characters written by history with the same myth-making brush that writes fiction. Convenient, cathartic, utterly inaccurate.

But I don’t write demigods. I write witnesses, floaters, participants without insight. They’re often not even granted the courtesy of closure. They move through a world that refuses to acknowledge their significance. And why should it? The cosmos doesn’t care if your backstory is tragic or if your girlfriend left you on page forty-two.

Sometimes the character who seems central is merely catalytic. Other times, they’re inert—filler between philosophies. If someone changes, maybe it’s society, not them. Maybe the reader. Or maybe no one.

So no, I don’t build arcs. I don’t force characters to evolve like Pokémon just because Act III demands it. I drop them into a world and watch what happens—often, nothing. Because that, more than any tidy redemption tale, is how life actually works.



That’s the work. Not myth-making. Not therapy. Observation. Dissection. Not a ladder to transcendence but a mirror, tilted just so.

Welcome to Ridley Park. Watch your footing. There are no arcs—only echoes.

Behind the Binding: Sustenance in Print, Pixels, and Purgatory

Not quite a launch. Not quite a rant. Just one author trying to get a novella into the world without sacrificing too many hours or brain cells.

Paperback Problems

I’ve been writing quite a bit lately—several novellas/novelettes, to be precise.

They all began life as short stories, but brevity doesn’t come naturally. Apparently, I can’t shut up even on the page. I toyed with the idea of releasing a thematic collection, and I still might. But for now, Sustenance is getting its own debut—likely this month.

The book clocks in at around 14,000 words, printed across 144 pages. I’ve read that readers prefer novels to novellas, but I’ve also read that readers don’t really read anymore. Time’s short. Attention spans are shorter. Maybe shorter fiction has a fighting chance. We’ll see.

I formatted it in 6×9 inches, which may have been overly generous. It’s leaner than your average indie fantasy tome but still thicker than my last Žižek collection. So there’s that.

The manuscript began in Word, like every poor decision. I laid it out in InDesign and exported the PDF through Acrobat. No budget, so I designed the cover too—started in Illustrator for the vector charm, but ended up in Photoshop, where I’m more at home. I designed the full wrap—front, back, spine—as a single canvas.

This was a mistake. More on that later.

Still, I’m pleased with the final look. Might reuse the style across future novellas for a bit of visual branding. There’s barely enough of a spine to print on, but we suffer for aesthetics.

Proofs arrive Thursday. Fingers crossed.

Hardback Headaches

Then came the hardback edition. Same 6×9 size, same interior. Should’ve been simple.

It wasn’t.

I forgot (again) that hardbacks require extra bleed and margin space. Couldn’t just resize the existing cover without risking pixelation. If I’d stuck with vectors, this would’ve been a breeze. Instead, I got to rebuild the entire layout from scratch—layers, guides, grids, the lot.

Hours of joyous rework. Lesson learned. Until next time.

eBook Escapism (and Other Fantasies)

Converting the layout to eBook format was a slow-motion trainwreck. I’d inserted custom font glyphs above chapter titles in InDesign. They rendered fine—until they didn’t. Halfway through, chaos reigned.

I cracked open Sigil and manually edited the XHTML. So far, so fiddly.

Then I uploaded the .epub to Amazon. Except Amazon wanted a .kpf file. Of course it did.

Enter Kindle Previewer. Except it doesn’t support embedded font glyphs. So I converted them to SVGs.

Still no dice. Kindle’s rendering engine is older than most of its readers. SVGs failed too. So I converted every glyph to PNG, rewrote the CSS, rebuilt the XHTML again, and gave it another go.

Looks fine. Not perfect. I gave up.

They’re just decorative anyway. No plot-critical glyphs here.

The Kindle version should go live shortly. I enrolled it in KDP Select, which means 90 days of exclusivity in exchange for a modicum of convenience. After that, I’ll look at wider distribution.

For the eBook cover, I simply cropped the original layout in Photoshop. That part was, mercifully, straightforward.


What’s Next?

This post is more documentation than declaration. A sort of production diary. I’ll follow up with an actual announcement when the book launches, plus a few reflections on themes, characters, and that moment when you realise your protagonist may have accidentally sexed up a chicken.

Long story.

Anyway, this is just the start. Stay tuned.

Or don’t. Up to you.

AI Editor Issues

I employ AI editors for copyediting and alpha-reading. They are useful but have limitations.

Some of my writing is ordinary – Acts I, II, and III; Beginning, Middle, and End. This is AI’s sweet spot: assess a piece and compare it to a million similar pieces, sharing plot structures, story and character arcs, heroes’ journeys, and saving cats.

Other stories are experimental. They don’t follow the Western tradition of tidy storylines and neat little bows, evey aspect strongly telegraphed, so as not to lose any readers along the journey.

Mary approaches a doorway. Mary opens the door. She walks through the doorway — the doorway she had approached.

Obviously, this is silly and exaggerated, but the point remains. AI presumes that readers need to be spoonfed, especially American audiences. (No offence.)

But life doesn’t work like this. We often witness events where we have no idea what happens after we experience them. We pass strangers on the street, not knowing anything about their past or future. We overhear something interesting, never to get a resolution. We get passed by for a promotion but never know the reason why.

In science, there are lots of dead ends. Do we want to know the answers? Yes. Is one likely? Maybe; maybe not. Will we make up answers just to satisfy our need for closure? It happens all the time.

In writing, we seem to not accept these loose ends. How many times have you read a review or critique where the complaint is, “What happened to this character?” or “Why didn’t Harry Potter use his invisibility cloak more than once despite it being an obvious solution to many prior and future challenges he faced?”

Sure. I agree that it feels like a plot hole, but the author doesn’t have to tell you that Harry lost it in a poker match, it got lost in the wash, or Ron snatched it.

I’m finishing a story, and various AIs provide similar commentary. Even more humorous are the times it can’t follow a thread, but when a human reviewer reads it, they have no difficulty. In the end, there may be unanswered questions. Some of these leave the universe open for further exploration, but not all questions have answers. AI has difficulty grasping this perspective.

Rhoticity Chicken: The Final Cluckfrontation


The skies darkened over the Coop of Justice. Inside, Rhoticity Chicken—a rooster of unparalleled enunciation—perched on his golden roost, adjusting his crimson cape. His mission was simple: to defend the final R in English against the insidious forces of vowel decay.

Audio: NotebookLM discusses this topic.

Across the barnyard, his greatest nemesis, Non-Rhotic Chicken, cackled from atop his weathered soapbox. “Togethah, my feathah’d comrades,” he declared, wings outstretched, “we shall ERASE the intrusive ‘R’ from this land. Wintah, summah, law and ordah—it shall all flow smoothly once more!”

A murmur rippled through the coop. Some hens clucked nervously. Others nodded, spellbound by his seamless vowel transitions.

But then, a mighty R echoed through the barn like thunder.

“NEVER!”

Rhoticity Chicken flapped into the air, his chest puffed out with impeccable articulation. “You shall NOT take the final ‘R’! I have defended it from the creeping shadows of elision for YEARS, and I shall not fail now!”

From the shadows emerged The Trilled Chick Henchmen, a gang of feathered mercenaries trained in rolled Rs. They trilled menacingly, their Spanish and Italian inflections rattling the walls of the barn.

“Señor Rhoticity, your time is up,” rasped El Gallito, the leader of the henchmen. “Your crude American Rs will be smoothed away like an old dialect in the sands of time. Trill, my hermanos!”

They rolled their Rs in unison, a sinister wave of phonetic force blasting toward Rhoticity Chicken. He staggered, his own hard R wavering against the onslaught of linguistic variation.

But he clenched his beak and stood firm.

“No,” he declared, eyes blazing. “You can roll your Rs, you can drop them, but you will NEVER take away my right… to pronounce… HARD R’s!”

With a mighty CROW, he unleashed his ultimate attack:

THE RHOTIC RESONANCE

A shockwave of perfectly articulated, non-trilled Rs blasted through the barn. It swept across the land, restoring all lost R’s to their rightful places.

Non-Rhotic Chicken gasped as his vowels stiffened. “No—NOOOO! My beautiful syllabic flow—GONE!” He clutched his throat as a long-forgotten ‘R’ slipped back into his speech.

“I… I… can’t… say cah anymore… I… I just said… car.”

The barn fell silent.

Defeated, Non-Rhotic Chicken collapsed into a pile of feathers, mumbling in fully articulated rhoticity.

The Trilled Chick Henchmen scattered, their rolling Rs faltering into incoherent babbling.

Rhoticity Chicken stood victorious. He fluffed his cape, took a dignified breath, and proclaimed:

“Justice. Honor. Pronunciation.”

And with that, he flew into the night, ready to defend hard R’s wherever they were threatened.

THE END…?

Against the Grain

As a writer, I fully embrace the digital age – word processors, AI, eBooks, print-on-demand, and so on. Still, I like to proof my drafts on paper. I also render audio with ElevenLabs, so I can hear the flow. You might be surprised how often that picks up awkward phrases and typos. I’ll save this for another post.

I find that printing double-sided on A5 creates just the right form factor for a paperback. A problem is that the grain is running the wrong way. This means that the pages curl horizontally, left to right. You want the page to curl to to bottom, especially if you want to bind it in book or booklet.

A solution to this is to print to A4 in a booklet form, and then fold the pages into a booklet. The result is a book having an A5 page size. And, the grain is now vertical, top to bottom, eliminating that pesky curl.

I take two approaches to this A4 technique.

First: Print a sheet containing four pages at a time, e.g., 1, 2, 3, and 4. This creates pages, 4 and 1 on the obverse and and 3 and two on the reverse. When folded, the pages are ordered as expected for a book.

TIP: Ensure you’ve set your printer to landscape and use a booklet template. I tend to print to PDF first, and use its format settings.

Rinse and repeate for pages 5 to 8, in counts of four. Stacked and ordered, you’ve got a booklet. The beenfit of this approach is that you can stack as many of these as you like until you’ve got all your entire book printed. Use a long stapler or pinch binders to fasten.

TIP: Be sure to account for a gutter, especially for books with more pages, so your text doesn’t get lost in the fold and is presents as expected.

Second: Print two of four sheets at a time. I recommend four but no more. Printing four A4 pages in a booklet format creates a 16-page booklet. And you thought, you’d never need to use maths out of school. The reason I recommend no more than four sheets is that the page ends don’t align well with more. The page ends start getting a curvature. Again, more maths. This is not an issue printing the previous style, but you need to keep it in mind here.

With four sheets at a time, the book is incremented (obviously) in groups of sixteen, so your finished booklet should be a multiple of sixteen. Blank pages at the start and end are fine. Consider a faux cover of sorts.

Also, of you only need a pamphlet – say, sixteen pages – you’re in luck.

In a Bind

If you want to be a real fancy pants, you might considering binding the pages. Say you want to create a bound novella for freinds and family. Punch holes through the folds, and stitch them together. This is fancier than staples.

Stack a series of 16, 32, 48, 64, and so on to create your book. If you have access to a binding machine, create a cover with heavier card stock and wrap it around, fixing it with adhesive.

The cover will need to be larger than A4 because of the aforementioned size problem. Plus, you’ll need to account for the thickness of all of the pages. B5 or even Legal-sized paper may be a solution. I haven’t done it or the maths, so this will be your assignment.

Parting Shots

You may be able to create a booklet with Letter and Landscape paper as long as you are OK with the final dimensions.

You may also be able to find A5 paper with a top-to-bottom grain, in which case, use it. You can settle with standard A5 sheets, but just know that you may be quickly frustrated when your pages start turing in.

Note: A4 and A5 are standard in the world except in the United States, where is is difficult to find and priced significantly higher there. If you know a source of decent quality A4 or A5 paper in the US, let me know in the comments.

AutoCrit Challenges

I don’t hide the fact that I rely on AI for early editorial feedback. Once a story is complete, I break out AutoCrit. This programme works well for typical stories that follow standard practices with common tropes. It gets quite confused when I feed it intentionally awkward stories, not the least of which is to advise me to eliminate the awkwardness.

This is a challenge with AI more generally. In this particular story, I leave a lot of loose ends and misdirects, as it’s a commentary on the conspiracy-driven culture we inhabit. The advice, is along the lines of, “You forget to close this lopp. What happened to so and so.”

But this is life. We don’t always know the full story. We drive past an multi-car accident where cares are overturned and in flames, but we never find out what happens – even if we scour the newspapers and internet. Who was that? What happened? What caused it?

We often never find out. In most books and movies, we find out everythung, and it all comes packaged with a nice bow. This is what AI expects. It’s the diet it’s been fed.

Some stories subvert these notions here and there, but by and large, this is not typical American fare. Readers and viewers need to be spoonfed without inconsistencies.

Speaking of inconsistencies addressing one scene, AutoCrit said that a character should act impulsively in one situation and reserved moments later. This was flagged as an iinconsistent character.

In the scene, a woman stops her car immediately to help an injured man on the roadside, but as she gets out of her car an approaches her, she shows caution.

This was a red flag. Why would she have always been rash or always been cautious?

My response, because that how real people act. She acts on instinct but quickly considers that she’s a vulnerable woman alone with a man miles from anywhere.

I don’t suspect a human reader would find this surprising. This is the intelligence absent from Artificial Intelligence — cultural intelligence, a cousin of EQ, emotional quotient.

I know how I want the character to act. I do want AutoCrit to inform me that character A is wielding a pistol but then stabs another character, or that character B is a teetotaler and is getting drunk or that character C has a shellfish allergy but is downing lobsters like they’re going out of style. And I certainly what to be shown continuity errors.

The biggest challenge I have with AutoCrit that is less promonent with other AIs is that I can preface my content with a note explaining my intent. I can even do this after the fact.

If I feed ChatGPT, Claude, or DeepSeek a story of segment to critique without a preface, the responses may be similar to AutoCrit, but when I follow up with some meta, the response may be, “Now it makes sense, but why is John wearing lipstick?” Perhaps he’s metrosexual or non-traditional. Perhaps it’s an oversight.

I dont meán to demean AutoCrit. I’m just advising that if you are writing stories not compliant with 80 per cent of published works, take the advice with a grain of salt, or reserve AutoCrit for more standard fare.

AI Humour and Irony

I’m wrapping up a short story, and I ran it through several generative AIs for feedback.

The first problem is that. although AI provides critique, it is programmatically overly optimistic.

The second problem is that it doesn’t understand nuance.

The working title of this story is Temporal Babel. It’s a low-key absurd, mundane, ironic satirical journey through space-time, language, and identity.

An unconscious, naked man is discovered on a roadside. Events unfold. Who is this bloke, and how did he get here?

I fed the draft into these AI platforms and processed the responses:

  • Anthropic Claude
  • DeepSeek
  • Google Gemini
  • Grok
  • OpenAI ChatGPT
  • Perplexity

I tried to use some other platforms, but they were wholly inadequate:

  • Liquid Labs
    Doesn’t accept attachments.
  • Meta AI
    Doesn’t accept document attachments. Images only.
  • Not Diamond AI
    Just an aggregator that employs the other AIs.
  • Pi AI
    Conversational AI. Not designed for this type of critique.

I also use AutoCrit, but not until my draft is complete, so not yet.

With the draft, they each get the same starting prompt:

I am drafting this story. It’s not ended yet, but might you evaluate it as-is?

The draft includes copious notes about ideas not yet captured, as I tend to outline my material and then add resolution. It’s like taking a sketch to a final drawing, adding and shading until it’s complete.

Without fail, the AIs complained about a lack of an ending – despite being informed the ending is not yet written.

Some AIs miss the point (even having the benefit of explanatory notes). This is still useful because it allows me to reevaluate whether a reader will be as lost or confused.

I don’t want to share any spoilers, but I’ll copy-paste some high-level commentary, reiterating the optimism bias. This is not me humble-bragging. It’s what the AIs yield. DeepSeek seems to be the most critical on the surface

Anthropic Claude 3.7 Sonnet

Based on what you’ve shared, “Temporal Babel” presents a fascinating premise with strong potential. You’re developing a unique take on time travel that focuses on language evolution and the mundane aspects of humanity rather than the typical science fiction tropes.

DeepSeek

Temporal Babel is an intriguing and ambitious piece that blends elements of science fiction, mystery, and psychological drama. The narrative structure, with its shifting points of view and fragmented memories, creates a sense of disorientation that mirrors the protagonist’s own confusion, which is effective for the story you’re telling.

Google Gemini 2.0 Flash

{This is a well-written and intriguing story. The characters are well-developed and the plot is suspenseful. The story is also thought-provoking, raising questions about time travel, identity, and communication. I am eager to see how the story ends.}

Grok 3

Thank you for sharing your draft! It’s a compelling story with a strong sense of mystery and an intriguing premise that blends science fiction, cultural displacement, and human connection. Below, I’ll provide an evaluation of the story as it stands, focusing on its strengths, areas for potential improvement, and some thoughts on its current trajectory. Since it’s unfinished, I’ll avoid speculating too much about the ending but will offer suggestions based on what’s here so far.

OpenAI ChatGPT 4o

This is a fascinating and well-executed piece so far. You’ve managed to blend speculative fiction with grounded procedural storytelling, creating an atmosphere of quiet unease. The linguistic elements are some of the strongest aspects, making this feel fresh rather than like another “mysterious visitor” trope. The lack of easy resolution enhances the realism, and your characters feel distinctive and lived-in.

You’re well on your way to crafting something intelligent, immersive, and deeply intriguing. Keep pushing the subtle strangeness, and this could be an incredibly compelling read.

Perplexity (Auto)

The story has a strong, suspenseful opening. You’ve established a good sense of place and effectively built tension through Sena’s observations and internal thoughts. The mystery surrounding the naked man with strange markings is intriguing, and the arrival of the authorities adds another layer of suspense.

The Rest

Except for Gemini, perhaps a feature of its Flash model, the others provided more extensive feedback and summary.

To be fair, I am no fan of Elon Musk or X (formerly known as Twitter), but I liked Grok. Its responses felt more natural and less forced optimism. It also picked up on some nuance the others missed. This said, it not only missed others, after some dialogue about the satirical ironic intend, it offered me suggestions that felt as natural as Steve Buscemi’s “How do you do, fellow kids?” meme.

Here’s a non-spoiler sample:

Nurse in “Assessment”: “Hold still, E.T., no phoning home yet.” Jef blinks, “No compile,” thinking she means coffee—absurdly off.

Neither story context nor character profiles do anyhting to set up or frame this delivery. “The nurse” is a character. I have no idea where the ET reference comes from, nor by extension, the “phone home” line. How any of this relates to coffee is beyond me.

Grok, don’t quit your day job.

Closing

This story is intentionally anti-science fiction, anti-trope. To clarify, I’m not sure one can write a story devoid of tropes, but I prefer to subvert the expectations that come with some.

Those who follow my Philosophics Blog know that I am a Postmodern philosopher. I don’t buy into teleological notions of progress and associated metanarratives, leading to shiny spaceships and a Jetson existence. Technology is neither Skynet nor a saviour. It’s just a tool.

Accelerando and Other Non-Fantastical Utopianism

I’ve been absent in the realm of fiction and this blog for a while, and I’ve got many other things distracting me, but I wanted to post something somewhat relevant. I prompted ChatGPT for possible writing ideas related to the rise of artificial intelligence and the interaction with humanity minus the usual tropes. I share the response.

Audio: NotebookLM podcast of this content.

Ah, so you’re after something in the vein of Accelerando—where AI doesn’t turn the world into some saccharine utopia but instead naturally supersedes humanity through sheer momentum, without falling into the Skynet or AI God-King clichés. You want the cold, indifferent inevitability of intelligence evolution, rather than some paradise-like AI nanny-state. Got it.

Here are some that fit your criteria:


1. “Blindsight” (2006) & “Echopraxia” (2014) by Peter Watts

  • Blindsight is a first-contact novel, but its underlying theme is the obsolescence of human consciousness. The protagonist interacts with post-human beings and non-conscious but hyper-intelligent AI entities that operate on levels far beyond human comprehension.
  • The sequel, Echopraxia, takes it further, showing how post-biological intelligences leave baseline humans behind—not maliciously, just inevitably.

Why it fits: The AI isn’t a dictator or savior—it just moves beyond us, as if we were an outdated biological app.


2. “The Lifecycle of Software Objects” (2010) by Ted Chiang

  • A novella exploring the evolution of AI over time, without falling into dystopia or utopia.
  • It’s about how AI develops differently from humans, not through brute force but through time, experience, and interaction—like raising children.
  • Eventually, these AI beings begin to diverge from human concerns, neither evil nor particularly benevolent—just different.

Why it fits: AI isn’t malicious or benevolent—it just has its own agenda that doesn’t align with human expectations.


3. “Neuromancer” (1984) & The Sprawl Trilogy by William Gibson

  • Gibson’s cyberpunk classic ends with an AI (Wintermute) achieving post-human transcendence, merging with its counterpart Neuromancer, and basically leaving humans behind like forgotten relics.
  • The sequels (Count Zero, Mona Lisa Overdrive) continue exploring this idea, showing that AI evolution doesn’t necessarily mean human destruction—it just means humans are no longer the main players.

Why it fits: AI doesn’t need to kill humans—it just grows beyond them, leaving them to stagnate in their own meatspace.


4. “The Quantum Magician” (2018) by Derek Künsken

  • A hard sci-fi novel featuring AI as a fully realized successor species to humans.
  • Unlike many sci-fi AI, these digital minds don’t care about conquering or eradicating humanity. They are simply running simulations and following complex evolutionary patterns that humans don’t fully understand.
  • They’re more like inscrutable post-human strategists than villains.

Why it fits: AI in this novel is playing a game too vast for humans to comprehend, rather than plotting dominion or destruction.


5. “The Singularity is Near” (2005) by Ray Kurzweil (Nonfiction, but reads like Sci-Fi)

  • Not a novel, but Kurzweil’s vision of AI progression follows the “humans will simply fade into irrelevance” model rather than outright war or paradise.
  • He argues that AI will integrate with society at first, then surpass humanity so completely that we’ll become a footnote in history.

Why it fits: If Accelerando is a chaotic, unrestrained Singularity scenario, this book describes the same fate with the veneer of optimism, but the end result is the same—humans are obsolescent.


6. “The Collapsing Empire” (2017) & “The Interdependency” Series by John Scalzi

  • AI-driven governance is subtly woven into the background of Scalzi’s universe.
  • The AI here isn’t explicitly replacing humans, but over the course of the series, it becomes clear that humanity’s biggest decisions are increasingly made by non-human intelligence.
  • Instead of war, there’s a quiet shift—humans stop being necessary.

Why it fits: AI is neither enemy nor friend—it’s just a more efficient system replacing outdated human decision-making.


7. “Player of Games” (1988) & The Culture Series by Iain M. Banks

  • Banks’ Culture novels are often mistaken for utopian AI futures, but they are really about AI inheriting the universe while humoring human existence.
  • The Minds (the super-AI) run everything, letting humans play at politics, war, and philosophy, knowing full well they don’t actually matter.
  • There’s no rebellion—just an unspoken truth: humans are pets in a post-human, AI-governed reality.

Why it fits: AI doesn’t need to conquer humans—it just makes them irrelevant.


8. “Glasshouse” (2006) by Charles Stross

  • Less well-known than Accelerando, but still in the same vein.
  • Set in a future where AI and uploaded minds are dominant.
  • Humans are relics, preserved out of nostalgia, while AI minds operate at a level beyond our comprehension.

Why it fits: AI replaces us, but it’s not a dystopia—humans are just increasingly pointless.


9. “Schismatrix” (1985) by Bruce Sterling

  • A deep-cut cyberpunk novel about the quiet obsolescence of baseline humans.
  • AI and post-human intelligences evolve, while humans either upgrade or fade into irrelevance.

Why it fits: It captures that Accelerando vibe of technological change outpacing human relevance rather than destroying us outright.


10. “Minds, Machines & Evolution” (1998) by James P. Hogan

  • A collection of Hogan’s short fiction, often exploring AI transcending humanity without malevolence.
  • One of the stories, The Bug Trap, is particularly fitting—it depicts AI simply moving past human limitations, rather than enslaving or eradicating us.

Why it fits: AI isn’t playing by our rules anymore—it’s just evolving past our concerns.


HONORABLE MENTION: Hannu Rajaniemi’s Works

Why it fits: AI doesn’t destroy us—it just leaves us behind.


Final Thoughts

You’re looking for that Accelerando flavour—AI taking over not through malice but through sheer efficiency, intelligence, and inevitability. The books above match that criterion: no Skynet, no god-like AI overlords, just a quiet, creeping obsolescence of humanity as we become background noise in a world where intelligence continues without us.

Which of these feels closest to what you’re looking for? Or do you want something even more clinical and indifferent, like a true “post-human shrug”?

Quoteless Dialogue

I’ve noticed a few publications adopting a quoteless or quote-free dialogue convention. I was wrestling with the idea of using it for my current short story, but I’ve opted not to.

On the upside, it can:

  1. create a more immersive reading experience, blending dialogue seamlessly with narration.
  2. give the prose a more streamlined, modern feel.
  3. be effective in representing stream of consciousness or internal monologue.
  4. subtly underscore themes of ambiguity or the blurring of reality and imagination.

On the upside, it might:

  1. lead to confusion about what is spoken aloud versus what is thought or narrated.
  2. be challenging for some readers, particularly those used to more traditional formatting.
  3. not be suitable for all types of stories or narrative styles.

I feel that it’s a valid stylistic choice that can be very effective when used deliberately and consistently. Its appropriateness depends on the specific work, its themes, and its intended audience.

If used, it often requires more careful writing to ensure clarity about who is speaking and what is dialogue versus narration. It works best when the author employs other means to differentiate dialogue, such as syntax, diction, or paragraph breaks.

    For my current writing project, (working title: “The Riga Paradox”), given its themes of reality versus perception and the blurring of identities, omitting quotation marks could be an interesting choice.

    However, it’s also a significant departure from my comfort zone and might require adjustments in other aspects of my writing to maintain clarity.

    Does anyone have thoughts for or against quote-free dialogue? Have you used it yet?