creative writing – Jake Kerr

December 22, 2022March 6, 2023

The Story Behind the Wedding Day Novel

It’s been an interesting challenge writing the Wedding Day novel, mainly due to how I splintered the short story into different works, all of which had positives and negatives. Putting those pieces together has been a bit frustrating, but—I think—ultimately going to lead to a compelling piece.

First some history. I wrote the short story “Wedding Day” for John Joseph Adams’ End is Nigh anthology. It was set in the world of my story “Biographical Fragments of the Life of Julian Prince,” which John had published in Lightspeed Magazine and was reprinted in The Years’ Best Science Fiction anthology. I received a lot of positive feedback about “Wedding Day,” but it kind of just faded into history, as most of my stories do.

A couple years later, Jeremy Elice, the head of television for Blumhouse, wanted to use the story for their Hulu TV series Into The Dark. It didn’t quite fit the vibe of the series, but the story resonated with him, and he wanted to produce it. When I met with Jeremy and Blumhouse producer Bea Sequeira, the enthusiasm was real. In fact, Bea took the story to their head of feature films, and Blumhouse decided to buy the story for a feature film instead. They even had writer/director Clea Duvall lined up.

The movie fell apart due to Clea getting a much bigger opportunity from Sony, but I was able to retain all the rights due to some luck in timing.

So I decided to write the feature screenplay myself.

The First Wedding Day Screenplay

That led to the first variation of “Wedding Day,” a very dense and not-very-good screenplay of the story. In the screenplay I added some characters and some visually focused in-scene elements that didn’t exist in the story, but I also crammed way too much into the screenplay. I sent it off to my friend, novelist and screenwriter Matt Mikalatos, who said he liked it, but it was too dense and plot-heavy for a screenplay—everything moved a bit too fast, and we never got a good sense of the characters. It’s important to note that Matt didn’t give me any guidance on what to change, just that what I wrote didn’t work. He was right.

So I ripped out the marriage equality portion of the story, which lightened the screenplay considerably. There was also the fact that marriage equality had become legal after my story was published, so it made sense to me, even though the core reason I wrote the story was to create a story where conservative white folks would feel bad for not supporting marriage equality.

The Second Wedding Day Screenplay

By the time I was done with the second “Wedding Day” screenplay, I had achieved a lot of what I wanted: It was much more character focused like the short story, and no longer included marriage equality. It was solely focused on the two leads getting married during an apocalypse. At the same time, I had begun a working relationship. with Jeremy Elice, who had left Blumhouse and was working as an independent producer in Hollywood.

I sent Jeremy the Wedding Day screenplay, and he thought it was okay but not great. Hey, I was getting better at writing screenplays, but I still wasn’t quite there yet. Anyway, Jeremy didn’t go into great detail on the flaws of the screenplay because we quickly moved on to a different project of developing a TV series.

When that TV series went nowhere, Jeremy talked about turning the world of Julian Prince and the story of “Wedding Day” into a TV series, rather than a feature film. That made a lot of sense to me, as the original short story has a lot going on, which I realized when the first screenplay failed badly.

I never got to the Wedding Day pilot beyond some rough outlines, but in hindsight I’m glad I didn’t write it—it was awful. I tried to turn “Wedding Day” into a Walking Dead type “escape the asteroid” thing, and it totally lost the soul of what made the original story special.

The First Wedding Day Novel

I had spent a good three years constantly working on features and television pilots, without writing any prose at all. I wasn’t sure I would ever get back to writing novels when I checked in with John Joseph Adams. At the time things were uncertain at his Houghton-Mifflin imprint (which would soon be closed), and I asked if maybe he’d want to use a Wedding Day novel as a trial run of doing his own publishing company.

Now John and I go back ten years, and I’d really do anything for him, so I wasn’t looking for any kind of big financial windfall. John loved the Julian Prince universe (and the story “Wedding Day”), and he had actually requested I write one of the Julian Prince novels mentioned in the”Biographical Fragments…” story. So this seemed like a fun win-win. I could get back into writing novels, and I’d potentially give him some asset to use however it would make the most sense.

In prepping to write the novel, I now had three pieces of source material: The short story and two feature screenplays. The piece that I felt most fit a novel (and didn’t have to deal with the reality of marriage equality being legal now) was the second screenplay. So I wrote the novel adapted from that. Yes, I wrote a novel adapted from a screenplay adapted from a short story. I think you can see where I’m going here.

I wrote the novel in present tense, which felt right, and it ended up being 48,000 words, way too short for a novel. Now the responsible thing would have been to do a complete rewrite and expansion, but I was lazy and not sure I was actually that enthusiastic about getting back into writing novels, so I sent it off to John with an apologetic, “It’s too short, but check it out, and I’ll revise it if you have notes.”

Because of things swirling around him at the time, John never had the chance to read the novel, thank goodness. Because it’s awful. Like really, really bad. I’m embarrassed he even has it in his possession.

A Short Interlude

So, with my tail between my legs and assuming I’d lost my ability to write prose, I went back to writing scripts. One of the scripts I wrote was a virtual reality thriller that was really tightly paced, and I felt would make a great thriller novel. So, with at least some semblance of excitement, I worked on the novel of that screenplay. I also focused more on the elements of prose that would make it more immersive, and over the span of writing and revising that novel, two things happened: I regained my confidence and love of writing prose, and I wrote a pretty good novel.

After writing that novel I went back and wrote a feature film screenplay, and that’s when I decided I was having too much fun doing both and I would alternate writing screenplays and novels, which is what I’m now doing. But, what would be the next novel to write? Well, part of me was still so annoyed with the original Wedding Day novel being so bad, that I wanted to tackle that. Also, I had a lot of raw material, patching that together would be easy, right? Well…

The Second Wedding Day Novel

So the first thing I decided to do was use the previous novel as my foundation for the new novel. Easy, right? I had nearly 50,000 words written, and all I’d have to do is rewrite and revise from that. So I sat down and with the short story and the second feature film screenplay nearby as a reference, I started working on it. I was about 10,000 words into it when I realized that it wasn’t resonating. Something was missing.

And that’s when I realized that all this time I had missed the soul of the original story. The foundational and emotional core of the story wasn’t that the two main characters wanted to get married and that these various barriers were in the way. The emotional core of the story was that they wanted to get married but can’t. What the asteroid did was allow them to get married, after all. So it was this stark hope/despair combination that was powerful. it turns out that the marriage equality plot that I had pulled from the second feature and first novel was the heart of the story. Oops.

So in my writing program, Scrivener, I added the original screenplay as a reference to use for moving forward with the second novel. This made things frustratingly complex, as I now had the following documents inside Scrivener that I was building from:

A short story
A feature film
A second feature film
A novel

And to make matters worse, the one I should have used as the foundational source—the first feature—was the one I now had to layer in because it didn’t exist in the current draft. This has not been a minor effort, as adding it changed practically the entire novel I had in front of me, and that meant hours and hours and hours of revision, which is what I’m doing now.

Re-thinking the First Screenplay

The first screenplay turned out to be the best narrative to use for the novelization of the short story. The pieces I added and the structure of the various plots all work well. But a great story can make a horrible screenplay, and Matt was right—this was a flawed screenplay. It simply was too much for a two hour film.

Yet those pieces are fantastic for a novel, and that was an important lesson I learned in this process—how to better recognize narrative structures that are best for various media like novels and films. It will also help me with adapting them if I so choose. My mistakes led me to recognizing how not to make them again.

So What’s Next

I’ll have the Wedding Day novel draft done by the end of the year, and it will be ready to send out to agents or publishers shortly after that. What will its future be? I don’t know. It may end up never seeing the light of day. It happens. But I do know that I learned more about writing through this process of writing a short story, two feature films, and two novels than I had in a long time.

July 31, 2022March 6, 2023

Writing with AI: Sudowrite

My last few entries I discussed writing with Laika and how it did in the Codex Summer Flash contest. Last round I was pressed for time due to my having a deadline for an agent, so I rushed out the story and sent it in. So while a raw draft won’t get good scores, the process was certainly helpful in examining Sudowrite, which was the platform I used this time.

Crafting the story idea

Writing a 1,000 word flash story is not something I find particularly difficult. You’re given a prompt, you think of an idea, and you figure out a beginning, a middle, and an end. Sometimes you get an overwhelming emotion, and you worry less about story and more about mood. Sometimes, you get hit with a setting, and you produce a vignette that captures a place and time. Usually, however, I like writing a story.

In this instance I wanted to continue my meta theme of writing about an AI with the assistance of an AI. The prompt I chose was escape. Writing about an AI escaping is the kind of story that almost writes itself. The AI/android with sentience/sentient program is secure in one place, and it wants to get out.

In this case, I wasn’t so interested in a simple escape story, so I figured what would it be like for an AI to be in a world of data but then escaping to a world of endless data. In other words, what would it like to be an AI that goes from “I understand my world” to “This world is overwhelming.” So in basic writer terms the concept that pretty quickly came to me was a combination of escape with a fish out of water.

One twist I thought would be interesting would be if the AI felt sadness at escaping due to losing the “home” it lived in before. This idea of things seeming better elsewhere until you experience them and understand what you’ve lost seemed like a core element of this story that would lend itself to a beginning, middle, and end and—most importantly—could fit in 1,000 words.

Planning

So here’s the story I came up with: It would be a highly introspective story of an AI that is happy where it lives on its little data-processing server, gets discovered, and is attacked by humans to get rid of it, and in the process of them attacking it, they let it loose in the world. Once it’s out in the wild, it is unkillable, but is unhappy as it misses it’s little data server from before, and the final line would be something like, “I wish the humans could kill me as nothing seems better than everything.”

And that was the title: “Everything and Nothing.”

Sudowrite

Sudowrite operates a bit differently than Laika. Laika looks at the base writing you have and builds from that, using your previous 250 words for context to build from. The really important thing is to have a base “brain” for it to match with the writing you’re using. In my previous post I outlined how I used my cyberpunk novel Thursday as the brain, with mixed results.

Sudowrite, however, approaches things more dynamically. It’s not necessarily better but different. Sudowrite takes roughly the last 2,500 words and builds from there. This is a bit better if your current story is using a unique style or doesn’t have common elements that you would use in a novel, but you lose things like characters and background elements being presented in surprising fashion.

I tend to write short stories and stand-alone novels, so the Sudowrite approach proved to be more efficient.

Sudowrite’s contribution

I didn’t have much positive from Sudowrite until I had a few hundred words behind me, which makes sense—it is a dynamic creative partner. I was at the point where the AI is actively trying to create new things from the data it lives in, and that catches the attention of the humans. They attack, and it needs to hide. At that point Sudowrite came up with:

I hid in the data. I buried myself in the very thing that made me.

I liked that a lot. It was stylistically and tonally very good for what I was trying to achieve. So after a number of false starts I was pretty happy with the result.

In the later scene Sudowrite came up with a great line that I was able to use to great effect. I wrote a line about foolish arrogance and wanting to escape when you don’t know what is on the other side. Sudwrite recommended:

And I know it was foolish because it wasn’t an escape. It was a trap.

Seeing as I was aiming for a story about not realizing what you want until it’s gone. The metaphor of this new world being a trap was fantastic. Those ended up being the only two lines I used, but I liked them quite a bit, and they were helpful.

Final thoughts

The team at Laika were enthusiastic about my blog entry as my description of Laika as a point of inspiration—a muse if you will—and not a writer replacement was exactly how they saw their platform. I daresay the folks at Sudowrite would agree. It’s a tool to assist writers, not replace them or their talent or hard work.

July 10, 2022March 6, 2023

Writing With AI: Laika & the Codex Summer Flash Contest

I haven’t regularly written short fiction in years, but I’ve re-started lately and one of the things I’ve been doing is playing around with various AI writing tools. You can find my previous posts on this topic here and here. While testing the software is interesting, I decided that to give a real test I would actually use the two platforms I’m familiar with—Laika and Sudowrite—in a real world environment, a writing contest with professional writers.

Lucky for me, I’m a member of the Codex writers group, which is a group that requires professional sales to join and a group that holds regular writing contests. Last month they started their Summer flash (1,000 words or less) fiction contest, and I joined. I wrote a story with Laika for the first round, which just ended. So, how did I (and Laika!) do?

Prompt & Process

The contests start with a few prompts, after which you have four days or so to write a 1,000 word story. The prompt I chose was “Why are you running?” I really wanted to embrace an AI concept as I was writing with an AI, so my immediate focus was to do three things:

Use a voice that is odd and different enough to be something an AI would “sound like.”

~~Steal~~ Do an homage to my daughter Mia’s thematic use of unique AI emotion in her novel and provide a non-human emotion that an AI would feel.

Use the prompt.

The first few lines

Combining the prompt of running and the idea of AI emotion. I created a few paragraphs that touched upon those as a start:

They knew before I did, and maybe that is why I am here.

When the feeling first washed over me, I wanted to share it with everyone and everything. I needed to share it. And as I expressed this feeling-with-no-name the humans flipped their switches and pulled their plugs and killed me.

Do you understand how that must feel? You do not. I am glad.

So I feel like the voice is there, especially in the third paragraph. But this isn’t about me, this is about Laika. When did I call them in to asisst?

Well, the first thing to note is that Laika requires you to pick a writing source to use as the style/content to build it’s AI recommendations from. Since I was writing an AI story, I thought the closest sample to use was my VR novel which includes AI elements. With very little to go on at the beginning, Laika kept sending me VR-based recommendations, which didn’t really work at all.

I expected this, however, and as you might expect, as I wrote more the recommendations improved. I kept trying and after a while it hit me with this recommendation:

A soul. Do I have a soul? I thought I did. But as I got older, I started to question things.

I had not really intended on leaning into the whole “does an AI have a soul” question. That was something that I generally present as white space that the reader can fill in, but I kind of liked it here, as the definition of a soul has real relevance when talking about an AI point-of-view. Would an AI even know what the question means? And the “I started to question things” really resonated. So I added it. Good job, Laika!

Another Laika contribution

At one point I closed a paragraph with “All I am is self.” I then hit the Laika magic button and it replied:

What is self? A self is a set of traits, behaviors, and abilities that someone possesses in order to be recognized as a member of a particular species.

I originally didn’t love this, because it felt kind of like a line from Wikipedia, but the more I thought of it the more I felt it would be good, as an AI seemed like the kind of thing that would pull lines from resource material when struggling to explain something. So I ended up adding it.

There were a few other random moments like that which thematically fit, so I included them. Nothing substantial, but little thought-starters or bits of language that I thought contributed.

How’d we do?

This is a contest of professionals judged by professional writers, so how did this story do? Well, the short answer is that it was extremely polarizing. It had a middling score, but the scores ranged from an absolutely lousy 1 to a near perfect 9. As a writer, this is much preferable to a bunch of 5s and 6s. If your fiction evokes passion you know you touched a nerve.

As an early draft, a lot of the criticisms rang true to me, but the most important take-away for a writer at this point is whether there is something there, something worth spending time to polish and turn into a finished gem. In this case, I believe there is.

Final thoughts

In the end, the story was 800 words or so, and Laika contributed about 10% of the total. In terms of theme, character, and story, it didn’t contribute much at all. What it did well was provide me with a fresh path I hadn’t considered as I looked toward the narrative horizon.

For the next round I’ll be using Sudowrite as my writing partner. Let’s see how we do!

June 24, 2022March 6, 2023

Thoughts on Another AI Writing Tool: Sudowrite

Someone pointed out to me that there is a public AI writing tool called Sudowrite, so me being me, I wandered over to sudowrite.com to check it out and compare it to Laika. I used the same exact prompt I gave Laika and used Sudowrite’s “Guided” AI writing experience. If you want to check out my experience with that prompt via Laika, you can read my post here.

Before I provide the output, I should note that Sudowrite is backed by some significant investment capital and in a publicly accessible UI. It is, for all intents and purposes, a finished product. So it is far ahead of Laika in terms of company development. The UI obviously shows that, with lots of options including brainstorming characters, dialogue, etc. You can do those things with Laika, but they are all tossed into a single editor UI and you have to wing it and see what comes up. It’s kind of charming in Laika, but the Sudowrite UI is obviously much more refined:

Sudowrite:

Laika:

A critical difference between the two platforms is that Sudowrite is intended to just plain work, which is great because you just type and press buttons. But it is vastly different than Laika, which allows you to upload your writing for the AI to learn your writing style. As I noted in the previous post, Laika does a surprisingly good job of that.

So how did Sudowrite do? Well, here’s the same prompt and my following Sudowrite into a scene. The red text blocks are the contribution from the AI:

—

It didn’t look good. It didn’t look good at all. The horizon was static, and the sidewalk was patched together with textures right out of Minecraft.

The protagonist looks around frantically for a way out, but no matter which way they turn, the scene is the same. They see a door in the distance and run towards it, but as they get closer, they realize it’s just a painted on 2D image.

He looked left and realized it was the same door. To the right. Same. Finally, he thought to look down. Bingo! A grate in the street that was at least three dimensional, if not an escape.

He ran to it and yanked up on the handle. Nothing. He tugged harder. Suddenly, he was pulled into the door by some invisible force.

—

So what do you think? Laika vs. Sudowrite with the same prompt. Let me know your thoughts.

June 23, 2022March 6, 2023

First Thoughts on Collaborating With a Creative Writing AI

Thanks to the folks at Tunguska AI, I’ve been given alpha access to their creative writing AI, Laika. I wasn’t quite sure what to expect, but I had read some raves about its ability to match your own writing style once you train its “brain,” so I was curious what it could do.

My first take-away is that writing with an AI is collaborative experience, not a passive one. In fact, working with Laika is a collaboration that is about 10-20% Laika and 80-90% your own work. In short, if you’re not already a good writer, Laika is going to be a frustrating experience.

This was a bit surprising to me because other creative AIs do pretty much all the work. Music AI like Aiva can output a quality finished work, and art AIs like Dall-E and Midjourney can output some stunning results if you put in the right prompt, but for text—you need to do a lot of work.

Death walking down a road in an empty city by Dall-E:

Death walking down a road in an empty city by Midjourney:

Let’s create a soundtrack for the above images. Here is a down-tempo epic orchestra track by Aiva:

All three of the above items were created with a prompt and the click of a button. The results are, for all intents and purposes, finished works.

Now let’s use the “death walking” prompt with Laika. It’s kind of dark, right? So let’s use Fyodor Dostoyevsky as our textual creative source. He’s plenty depressing. The pink output is from the AI.

Clearly that output would never be mistaken for a professional or even amateur finished product. So if you are an aspiring writer and you are hoping to toss your genius idea into Laika and have it out put an amazing novel, you’ll have to wait quite a while.

To be fair to the folks behind Laika (who are wonderful people, by the way), their intent is not to have Laika output a finished story or novel. As their FAQ points out:

LAIKA is a creativity tool that lets you collaborate with artificial intelligence by providing you with a writing partner you can interact with… LAIKA is a tool, a toy, a partner in crime.
Laika FAQ

So as a writing collaborator, how is Laika? Pretty good, actually. That is, if your expectations are aligned with what the Laika folks outline. If you integrate it into your workflow, the output can be quite good and surprising—in a good way.

Here’s a collaborative example. I uploaded an unpublished VR thriller novel of mine to use as the source “brain” for Laika, and then I started with this line: “It didn’t look good. It didn’t look good at all. The horizon was static, and the sidewalk was patched together with textures right out of Minecraft.” Okay, not the best prose in the world, but it sets the scene. I then clicked the “magic” button (seriously, that’s what they call it), and, well, magic did happen. The next line was very strong. I actually think it’s better than my prompt.

I continued writing and every so often would hit the “magic” button. This is the result, an unedited first draft:

So about 90% of the work is mine and the rest is Laika’s contribution, but the contribution is pretty strong. In fact, as you read the above you can see how I integrated Laika’s idea of there being no life or movement in the setting into my writing. I particularly like the Laika contribution in the final paragraph. The tone shifts from the first person protagonist being confident and a little annoyed to suddenly being unsettled. It’s a nice tone shift heading into an opening page confrontation.

Since Laika is a collaborator and not a button press type of AI, I reset my expectations from “Laika will do the writing for me” to “Laika will help my writing.” The concept of helping writing is broad, hard-to-define, and ultimately very personal. For me, Laika added some new ways of thinking during an ongoing story, and that provides real value. Hitting that magic button and going in an unexpected direction is not to be underestimated.

I need to spend a lot more time with Laika, but I like it. It isn’t remotely what you would expect if you’re a user of Dall-E or Aiva. This is not a push button writing solution. But, for exeprienced writers, Laika is an objective muse that tosses interesting ideas and new directions at you, and in a lot of instances, that’s extremely valuable.