A Song of Onyx (Storm) and AI
Ted Gioia recently highlighted that when it comes to media, abundance is the name of the game. He cited Rebecca Yarros’s Onyx Storm, a 544 page romantasy novel that is also the fastest selling book in twenty years, as an example. While Gioia sees Yarros’s latest entry in her Empyrean series as indicative of where art is heading in terms of scale, I see something else.
I see Onyx Storm as the first opportunity for AI and literature that might actually work. And this is because, for all of its enormous popularity, Onyx Storm is terrible.
My suspicion is Gioia may have hesitated to cite Onyx Storm had he, you know, read it. Reading Onyx Storm is, in terms of content, equivalent to a binge watch session of Love Island. Despite being technically ‘long-form’ content, one would hardly argue that binging reality TV is the stunning rebuke to TikTok culture he thinks it is. It’s a book comprised entirely of manufactured cliffhangers, sexual tension, and dragon-based drama, not deep thought.
While reading Onyx Storm, I found myself experiencing the anhedonia Gioia mentions—I just couldn’t bring myself to pay attention. I didn’t care. Those who read it are not reading it as literature. They’re reading it as entertainment and to be titillated. That’s ok! But let’s not pretend it’s the same as the semi-virality of Middlemarch among the Silicon Valley cognoscenti and those in their milieu.
But Onyx Storm could have and should have been good. I know this because I’ve read the entire Empyrean series, including the banger of an initial entry, Fourth Wing.
Fourth Wing is not my usual fare. The four books I read immediately prior were Roadside Picnic, The Luminaries, Persuasion, and Blood Meridian. I only read Fourth Wing because a member of my book club selected it. Our book selection process is simple: we trust one another, so we rotate book selection person by person and each person just picks the book they want us to read and discuss next—no debate, no voting, no criteria. In this case, she was so excited about Fourth Wing she bought copies for all of us as Christmas presents.
So I read Fourth Wing. It was a delight.
Was it tropey and was the prose mid and the characters a bit wooden? Yes. But it rips. Fourth Wing is something like the literary equivalent of Mission Impossible: Final Reckoning. You do not go to a Mission Impossible film hoping for sumptuous cinematography or subtextual dialogue and face acting. You do not read Fourth Wing for its nuanced interiority or elegant characterization. You read it to have fun.
Do I wish Yarros had spent a bit more time crafting the book? Of course I do, the same way I wish Tom Clancy had edited down Rainbow Six or the Renowned Dan Brown had shown a bit more than he told. I can read a book on its own terms. Yarros didn’t set out to write a global phenomenon. By all accounts, Yarros was a reasonably successful romance writer who wanted to write a fantasy story. Romantasy is big, the shift was natural. Intentionally or not, Yarros figured out how to blend in some Dark Academia and Khalisi-wanna-be elements into the rare alchemy needed for an absolute blockbuster.
Even on their own terms as fun romantasy, however, Iron Flame and Onyx Storm are a mess. We know there are two more novels coming in the Empyrean Quintet. Thus I have a proposal for Yarros: use AI to help you re-write your novels and re-issue them.
The Return of the Living Story: Video Games & Homer
I get it, the idea of major revisions to and reissuing of a published creative work feels… bad. We recoil at it, instinctively. Lucas’s edits to the original Star Wars trilogy spring to mind.
The exception that proves the rule (there’s always one!) is The Hitchhiker’s Guide to the Galaxy. After starting as a radio play, it transmuted into books, then to TV, text-game, and LP. The original author, Adams, saw each as an opportunity to expand and explore. Each is its own experience, benefiting from the previous.
Video games reinforce Hitchhiker’s lesson for literature here.
No Man’s Sky, a space exploration game, was released missing nearly every feature original promised. Destiny, an online shooter with space wizards, had Bungie’s rock-solid gameplay, but narrative, dialogue, and lore that, where it wasn’t paper thin, was hilariously bad. Both, however, had just enough to keep audiences interested long enough to begin amending their mistakes. After years of patches, updates, and expansions, both are now considered masterpieces.
This is normal in gaming. Narrative single player games like The Last of Us, Horizon Zero Dawn, or Spider-man, allow players to adjust difficulty settings—choosing “Story Mode” for those who just want, well, the story or options like “Nightmare” for those who want (need?) to test their mastery of the mechanics. Games like Fallout 4 and Baulder’s Gate 3 allow users to modify their characters, install user-created modifications, and toggle graphics settings, not to mention choose their own path through the story itself. The game adapts to the player, not only the other way around.
Video games as analogy present a bit of a problem, I recognize, because the overlap of gamers and people hemming and hawing about AI in literature and publishing is vanishingly small. You might ask: how does this apply to narrative literature? What I want to emphasize here is that all these games have phenomenal stories that could sit alongside some of our best modern novels. Some of the best stories are being experienced exclusively by gamers! That conundrum, though, is for another essay.
The good news is that this narrative adaptability isn’t solely a modern innovation. It’s ancient!
I’m not a classics scholar, but most people seem to agree The Iliad was oral tradition before being set into text. This means each bard very likely adapted it to their audience—emphasizing different heroes, expanding favorite scenes, adjusting the reactions of their listeners. Even once transcribed and later published, the way The Iliad was told wasn’t permanent. Consider the translations! I vacillate between Fagels and Lattimore. When Wilson came out, people had opinions. All three are quite distinct.
One Thousand and One Nights comes to us cobbled together and double-translated, Arabic via French, its stories shape-shifting across cultures and centuries. Shakespeare’s plays transform with each production—gender-swapped, modernized, recontextualized. These stories have endured, in part, because they changed.
But what if we didn’t wait centuries for translations or adaptations. More over, what if we admitted to ourselves that not all novels are capital-L literature. Not everything deserves to be treated as a precious cultural commodity. Some things are good but uneven and could be made much better. Novels have, thus far, been immune to the Directors Cuts, extended editions, and remakes of film. Part of this is the sheer difficulty of writing. Some of it is the cost exposure of publishing. But some of it is just norms. It Is Simply Not Done.
The good news is we can just do things! What if we used AI to make books that could be good, but for whatever reason weren’t, actually good?
Adjustable Art
The same year we read Fourth Wing, my book club also read, among others, Persuasion by Jane Austen, Titus Groan by Mervyn Peake, and Autumn by Ali Smith. Imagine if I could take Fourth Wing and combine it all of those—not rewrite it, but tune it.
I’d first layer in Peake’s atmospheric descriptions and unsettling character details. I want to feel medieval society breaking under the weight of war and conspiracy. Next, I’d dial down the repetitive “he growled/she gasped” dialogue tags, smooth the clunky exposition, and reduce the smut from explicit to suggestive (or vice versa, depending on your taste!) with Austenian nuance, deft, and humor. Also, why not bring in some of her subtly searing class analysis to boot. Finally, I’d salvage the sparse character interiority with Smith’s knack for commentary and introspection, her eye for the critical quotidian and her ambiguous humor to be carried along by magic, dragons, and death.
AI could make this possible. While AI is, for the moment, quite terrible at original creative writing, it is impressively good at re-writing. AI is particularly good at translating among styles. The Ghilbifying craze is a great visual demonstration of this: take a picture, apply Ghibli’s style. Voila! Your memory is now kawaii. The AI couldn’t create a picture of you or invent Ghibli’s style, but it can quite readily combine the two.
AI could also allow authors to revise far more easily and efficiently. The current costs of editors, publishing, and the rest make it difficult for an author to face the daunting task of replying to legitimate criticisms of their work that appear well after publishing. With AI, however, a writer could revise their work as often as video game makers do. Yarros, it seems, didn’t have the time she needed to write well and hit the publisher’s deadline. Fine— then write well after the deadline!
There is no reason it needs to be Yarros herself either. An assistant, similar to one of the many minions James Patterson deploys to churn out a dozen books a year, could help re-write after rather than before. Compress this sequence, clarify this scene, sharpen these characters, add nuance to this dialogue, move this information up and this information back. As with Patterson does with drafts, Yarros could review the adjustments and approve an updated version. I’ve no doubt her obsessive fan base would—like Swifties who buy vinyls or gamers who pre-order for cosmetics—jump at the chance to have an improved re-read experience.
The Death of the Author… Again?
“But whose book is it then?” critics will cry. “If you change Yarros’s prose, it’s not her novel anymore!”
I mean, it’s Yarros’s. Yes, it’s kind of a Book of Theseus problem, but let’s not pretend like we haven’t solved attribution issues before. Somehow, Apple managed to adopt Murderbot and Amazon Fallout and our heads didn’t explode. We know how fan-fiction works. Is it going to be a little murky for a bit? Sure probably, but that’s not a reason to not pursue something that could both help a good author be great and give readers a better version of something they already want.
AI-assisted reading and writing could help every book find its optimal balance. Not by homogenizing everything into middlebrow mush, but by letting each reader find their personal sweet spot. Borrowing further from video games, readers could get a kind of tutorial or on-ramp to great works.
Moby-Dick is my favorite novel (I own four copies, have a tattoo, etc), but I was hugely skeptical of it at first. Thankfully, our professor had us read Typee, Oomo, Redburn, Bartleby the Scrivener, and Billy Budd before we dove into the main text. With that scaffolding of understanding (honestly Typee and Bartleby would have been enough) Moby-Dick was much easier to grasp in all its sprawling strangeness. Works like the Mahabarata or The Tales of Genji might benefit from similar such context building and reading training.
Finally, AI could also rescue overlooked works. How many brilliant stories are trapped in impenetrable prose? How many addictive plots are dismissed because of clunky writing? AI could be literature’s great democratizer—not by making everything the same, but by making everything accessible in multiple ways.
Great novelists must be masters of every element of their craft: voice, dialogue, setting, plot, pacing, character, action, prose, and description. Compared with nearly every other medium, such as film, TV, and video games, where creatives have entire teams that bring their vision to life, this seems incredibly unfair. AI could allow authors to act as directors, bringing specialized skills to assist where they themselves struggle.
The death of the novel has been heralded more than a few times. The growth of Amazon and the advent of ebooks were supposed to be a death knell of the small book store and the physical book. Yet both are thriving. Now I get it, there are lots of questions with the big one being how does the money part work? I’m not sure, but copyright has proven surprisingly flexible. Maybe we have some kind of watermark or soft DRM, where I can’t sell a remix without an author approving or getting a cut. Maybe there are video game style mod(ification) market places where readers can download customizations built by others. Maybe publishers charge more (or less?) for various versions. Who knows! The market will figure it out (to your delight or chagrin as the case may be).
My point is not that this is definitely how it should or will happen. If I knew that I’d be building or investing instead of opining. What I am arguing is that one possible alternative to seeing AI as a threat to literature is to see it as an opportunity to elevate the craft. Used well, AI can improve the quality of popular works and the popularity of quality works. How we tell stories has changed over time. Maybe the novel as static artifact will go away, but, as Gioia notes, long form as a whole will surge ahead at the same time.
As someone interested in literature, I want the form to succeed. I love stories and am often saddened when something with obvious potential is hamstrung by unevenness in the writer’s craft. That AI might bolster the form and bring more stories to more people is something we should be exploring more. I’m not entirely sure how it will work, but I’m optimistic we will. If we’re lucky, we’ll figure it out soon enough for me to enjoy and be enriched by finding out how Yarros’s five-part story ends.