The story was written by Claude. The image was generated by ChatGPT.
The grief audit flagged Sarah for insufficient mourning. Her husband had been dead six months; she’d smiled twice in public. The tribunal gave her a choice: pharmaceutical correction or emotional retraining. She chose the pills. Now she cries at everything. They’re very pleased with her.
I love listening to AI podcasts. Recently there has been a lot of discussion about Claude’s (Anthropic) growing expertise in creating code. Some experts were suggesting that by the end of the year AI will be writing 90% of all code.
Now, I know absolutely nothing about coding but I am forever curious. So, I asked Claude, “Can you write a piece of code that can generate pieces of flash fiction?” I was expecting a simple answer, instead it replied, “Absolutely. Given your flash fiction expertise, I’ll build something with the kinds of constraints and parameters you’d actually use – genre, word count limits, tone, specific prompts or themes.”
Continue reading “Flash Fiction Generator”
Yesterday I used an image created by Midjourney as a prompt in Claude for a 200-word short story. Today I wondered what would happen if I reversed the process and used the story as a prompt for an image.
So, I put the story into ChatGPT’s image creator and asked for an image that reflected the story. This what I got…
I’m not sure which is the more melancholy, the story or the image? What do you think?
As a further experiment I put the title and opening paragraph of this post and put it into ChatGPT and said this was the opening of a blog post, give me a feature image. I liked the output.
For this experiment I thought I would begin by asking AI, in this case ChatGPT, to give me a prompt. ‘Write a prompt in the form of a dystopian monologue.’ I did not say what I was going to do with the prompt, but notice how it has decided it is going to be a story. This is what I got…
I am required to explain what you will produce.
You will call it a story, though that word is no longer used officially. It must be short. Length beyond necessity is considered inefficient and may be interpreted as avoidance.
You will describe a world that resembles this one closely enough to be recognisable, but not so closely that responsibility can be assigned. Characters may exist, but they should function rather than desire. Motivation is optional. Compliance is preferred.
If something troubling occurs, present it calmly. Raised voices suggest instability. Questions are permitted only if they resolve themselves quickly.
You may end the piece without closure. In fact, that is encouraged. Endings tend to invite reflection, and reflection has a habit of spreading.
Remember: this is not a warning. It is a demonstration. Begin when ready.
This prompt is giving ChatGPT some very clear guidance as to how the story should be written. I wondered what would happen if I then simply pasted this prompt into ChatGPT? Here is the resulting story…
Continue reading “A Prompt to Prompt a Prompt…”
This experiment began with a simple question: what changes when the centre of a story is shifted away from the people in it?
Rather than asking AI to invent a character with feelings, motivations, or a backstory, I asked it to write from the perspective of a door. Not a symbolic door. Not a magical one. Just a door that opens, closes, and stays where it is. The constraint was not technical. It was perceptual. The door can only know what passes directly in front of it.
That immediately removes some of the usual narrative shortcuts. The door cannot explain why someone hesitates. It cannot interpret an argument, a departure, or a return. It can only register patterns. Who arrives at the same time each morning. Who lingers. Who does not come back. Meaning has to emerge indirectly, through repetition and absence, rather than through insight or emotion.
Continue reading “What Happens When the Door Tells the Story”
The AI Grandad 