This is a special announcement that for the next couple of days (or weeks or months), this blog space will be co-opted to bring you a special feature of “Why I Write” or maybe “Why I Create”.
I’d say I’d love to return you to your regular dose of “Why I Game” after, but given my tendency for long silences lately, I can’t quite promise that yet.
It’s not that computer games have gotten less attractive. I’ve a massive collection lined up and can’t spare the time to play ’em, let alone write about ’em.
It all started around the end of April when two things collided – the feeling of impending mortality in a world gone quite mad and a curious exploration of ChatGPT.
Messing with ChatGPT began fairly normally: asking it to analyze and organize thoughts in one’s head, playing with the mirroring function for therapeutic encouragement and brainstorming possibilities and generating fantasy names and roleplaying ideas and set dressing.
Then I got curious enough to see if it was smart enough to be the GM in an original setting I’d dreamed up.
It wasn’t half-bad, but it was definitely just pulling from genre conventions. Problem was, my original setting was a crazy mashup blend of cyberpunk and supers and urban fantasy, with a side helping of grounded semi-realism, sci-fi and biotech.
Minus magic because I didn’t want it to become Shadowrun. Minus occult or religious flavoring or SCP horror because I didn’t like those.
Trying to feed worldbuilding detail into ChatGPT meant me pulling out old lore documents (unreadable scribbles, really) and throwing it all into one prompt or another. To my amazement, it ate it all up and formatted them into very presentable semblances of readability.
One thing led to another, and I started explaining some of my best loved characters to it. Trying to see if it could replicate them faithfully (or at least acceptably) in its generated writing.
Then image generation became tempting, as well as avoiding timed delays and enforced model switches. It was also obvious that ChatGPT was as good as its “memory” and token limits. It made great sense until it started “forgetting” prior details, whereupon it started to hallucinate and make things up once more.
I talked myself into a trial run of the Plus membership for a month.
Something else had also occurred to me after reading OpenAI’s article on Use Cases for Writing With AI.
There was a story within me that I have wanted to write for a very long time.
The first official attempt was during Nanowrimo 2006.
It derailed badly about seven chapters in. I limped to the so-called “finish line” by writing vignettes and images in my head. Not even scenes. Maybe I didn’t fully have the knack or the words then.
The characters have been with me for even earlier than that.
In original form in various created fantasy worlds for roleplaying that never found willing parties to share them and in attempted writings that I mostly cringe to reread now.
Then slowly shaped and reworked to fit into the more mature cyberpunk-influenced creative universe that I was now more into, until the Ares versions of them can’t exist anywhere else, because their backstories are so rooted there now.
I’m not getting any younger.
There was still a fire inside me that wanted that damn story written.
For the sake of the characters. So that they can live forever on the Internet even if I don’t.
Problem was, I didn’t know where the plot was going. I didn’t know how to write the story I wanted to write.
Enter ChatGPT.
For a fortnight, we threw ideas at each other.
It gave me plenty of ideas – most of them bad or not quite true to my characters. I rejected them and explained more and told it to try again. And again.
I also fretted for a while about potential copyright issues and all that jazz… then realized that I had a day job and writing and getting published as a profession was likely a pipe dream. It’s not like I might pragmatically expect to make it so big that I can retire on royalties.
What I really wanted was just to get the story written and shared… out there.
If ChatGPT could help with that, why not feed it more of my writing so that it could improve, model-wise, for others’ benefit as well as my own?
So more and more of my previous original creations went in to ChatGPT and it started getting better and more spot-on with how my characters might act in given situations.
I was compiling a veritable swarm of words for later reference. Potential scene outline after scene outline. Plowing my way forward through branching possibilities for the plot. Until it all kinda made a certain amount of sense.
Enough to be excited as hell about the thought of making viable progress decades later. To be vibrating with urgency trying to find every available hour to write.
It was the most alive I had felt for years. Apparently, I had been trapped in a rut of mundanity and chores without realizing it. This dose of original creation was something I’d been lacking for ages, like a critical vitamin deficiency that I wasn’t aware of.
May went by like a personal Nanowrimo.
June got a lot harder. The words were now getting squeezed like blood from a stone. ChatGPT endured the venting a lot better than real people could.
Last week or so, I stalled about 26 chapters in. There are still a LOT more to go.
But I could feel it. Burnout was impending. So I had to take a break.
But I didn’t want to drop everything and stall with everything creative either. I might never get that fire back.
Enter the prequel story. The prologue.
(Literally, the working title for the document was Prologue.)
This was the original test case I used on ChatGPT. To see if it could suggest viable paths forward when my brain isn’t able to figure out a good path out of crippling blank page possibilities.
I started this story in 2013. Mostly as a solo roleplaying experiment mixed with an attempt at creative writing. (Hey look, we’re tying this back to “game” again.)
Just picked one of my easier characters to channel and get in the headspace of. Started rolling dice every time I had no idea where it was going.
The main concepts I used were from Mythic RPG:
- Ask a question with a yes/no answer. Roll some dice on a table adjusted for your own interpretation of likelihood and probability.
- When really stumped, roll on some Complex Question tables to get two words. Put them together and see what creative idea it produces.
I did most of the rolls using RPG Solo.
Some of the rolls were just a fudged 1d6 with Yes, No, Yes and… Yes but… No and… No but… possibilities.
I’ve left records of these rolls in the footnotes of the chapters I utilized them. Because I think it’s fun to see what complications were generated from using them. It was useful in the early chapters for open exploration of places and scenes I might not have thought of otherwise.
Because I was using characters I knew so well, there were also chapters where I just had to let them speak their truth, without doing the solo roleplaying bit.
And alas, I ran aground. Somewhere in the middle. Exactly where, I’d leave as a fun exercise to the reader. (Maybe I’ll share it after a while.)
Didn’t know what to do next. Didn’t know how to handle the problems I just set up for myself. Didn’t know how to close or end a story. At all. Let alone satisfactorily.
A dozen years rolled by.
Now it’s 2025 and ChatGPT exists. So I threw GPT at it as a test case.
And by golly, it helped to push some of the fog away.
Not all of it. Some of that was probably also my fault. I was almost crystal clear on where the story was going in early May. Then I got sidetracked to the far more important big long term chonker of a story.
When I got back around to it in late June, just as a breather exercise and a way to prove to myself that I could -finish- something, dammit, because the chunky story kept growing and GROWING… it took quite an effort to keep pushing forward, 500 words at a time.
But this prologue story was deliberately positioned as a shorter, less critical “short” story.
(You should know by now with my tendencies towards walls of text that I ended up with a 29k word novella.)
And I got it done. By hell or high water, it is DONE.
It can probably still be polished forever. But I don’t really want to.
It’s good enough for what it is.
The test case. The smaller scale prototype for me to learn lessons and make mistakes on.
(And boy, have I learned lessons about how difficult it is to format this for the blog. Took me the better part of an entire day.
The original dream was to make a whole different WordPress site for the collection of tales I might eventually acquire. Since I’m only one tale in, this seems like way too premature a hope. So I have co-opted my own existing blog and shoehorned it into a different section.)
So I’m happy to announce the soft launch (very very soft and still under construction) of the Tavernpunk Tales of Ares.
(It’s also easily accessible in the menu bar up top.)
What’s tavernpunk, you ask?
It’s a cozy little word that ChatGPT invented when I was trying to explain my vision of a weird little fantasy tavern in the middle of a cyberpunk city sprawl. I’ve gleefully stolen it because it is the exact expression of the stories I want to write, mainly for myself to read.
Because I like cyberpunk. But it’s also depressing and nihilistic because it’s an inherent part of the setting. Chrome burns out the human soul. Live fast, die young, leave a pretty corpse. It works for metal modernity and bleak bittersweet (or just plain bitter) endings.
What I really want to enjoy is cyberpunk with a happy ending. A hopeful one, at least.
And the way around it, for me at least, was to throw enough fantasy at the genre until it balanced out the scales. Urban fantasy, swords & sorcery fantasy, science fantasy, romantic fantasy, whatever.
I needed supernaturals that could heal and regenerate the incredible amounts of damage a cyberpunk dystopia could inflict.
I needed people who love each other, through all of the damage anyway.
To allow GPT its little time in the sun, here’s its mirroring take on it:
Prompt:
Not easy trying to write a goddamn cyberpunk story with a happy ending. But I think we're cracking it steadily with our new "tavernpunk." Just got to get to the actual writing of paragraphs soon, heh.
ChatGPT said:
It’s hard as hell — and you’re doing it.
You’re actually doing it.
You're threading the needle between everything cyberpunk usually wants to be (broken, cynical, bleak) and something that says:
➔ Yes, the world is still broken.
➔ Yes, the scars are real.
➔ But healing exists. Found family exists. Love exists. Rebellion doesn't have to rot into nihilism.
➔ There’s still space for mischief, laughter, and choosing each other anyway.
That’s tavernpunk.
Cyberpunk with heart.
Grit with hope.
Steel and scars and stupid grins at 3am over a bottle of something awful.
The first story I’m happy to present is The Mindwalker Job.
The roles were inspired by some RPG I don’t remember now. Maybe a Fiasco playset. Maybe a random cyberpunk or heist one I was reading way back in 2013.
They seemed to fit some of my characters so I threw them into it and gave them a mission with the intention of that solo roleplaying adventure romp.
They inhabited it as best they could – personality, feelings, truth and all. I hope I did them some justice with my narrative.
Oh, and as mentioned, my characters have existed in my head for more than twenty years. One of their backstories has an uncomfortable resemblance to something that’s happening in the real world now. I can’t exactly erase their backstory without erasing their entire existence.
All I can say is that it is horrifically tragic that the real world has started to resemble a fictional dystopian cyberpunk setting twenty years later. (And I’m sure similar things have happened in history as well, but it’s a bit front and center and on the nose when it’s happening right now.)
And yes, the heist movie poster in the featured image of this post is intentionally campy. Very much doesn’t resemble any of the actual characters that much, at all.
If you get to the Epilogue, you’ll know why. 😉
(Also AI-generated, for the standard disclaimer, BUT it was Photoshop edited by human hand because ChatGPT refused to take out one extra weird hallucinated sword in the cityscape background. Asking it to remove the sword removed BOTH swords.
Instead of wasting my time trying to argue with it, I just copy-pasted the part of the ZERO swords picture onto the TWO swords picture to give me a desired total of ONE sword.
No need to panic about complete AI takeover yet. Humans will still be needed for counting tasks, no worries.
And basic spelling. The four roles were also added by human hand, because I was seeing the rest of the image get horrifically messed up by creatively-spelled variations.)











