On Building with GenAI
I created a simplistic choose your own adventure story with my son. I've been working on this idea for more than a decade and the tools in the past have been extremely limited. I feel like the GenAI tools are at a place where I can build something fairly quickly and focus on creation rather than getting mired in the minutia of building software.
The choose your own adventure format has been around for decades. You have these books where they would say "turn to page X if you wanna continue this way and turn to page Y if you want to continue this way." We have advanced quite a bit at leveraging software to do this but fundamentally what we're trying to do is enable readers and even creators to create a dynamic story and enable re-readability and promote a deeper engagement with the story and the characters.
You can take the simplistic choose your own adventure stuff like the text-based adventures where you could go left or right or into the forest or into the mountains, and the results are always deterministic and created by the story creator. You can also look at more complex stuff like video games like Detroit: Become Human or Mass Effect where your choices define what types of conversation and story paths you can open up.
There are some choose your own adventure projects out now, where some of the results are generated by the AI itself, so rather than having a deterministic flow of storylines and character arcs the results are determined by previous choices sort of on the fly within the LLM's context.
This project started out when I read a post about leveraging GenAI for software engineering workflow. I wrote my own post about it here which references the original post. I wasn't really sure what I was trying to achieve other than to see how the GenAI tools could help me build something. I've gone through this like I said for years in my head, and the complexity can get extreme when you're dealing with something this potentially dynamic.
I sought out to see how far I could take this with extremely ambitious ideas and then ultimately dialed it back quite a bit. Here's the story.
Here are the results: