Overview
For context I wanted to try out making a simple mind map of graphic novels: authors, publishers, genres and cover art. The goal was to visualize connections and future facing show suggested books and how they relate to the ones the user likes. I got a start, it improved, then when working through fixes, drastically got worse and continually lost initial instructions.
Potential Niche Recommendation
In my case, it’s possible having a visual snapshot of versions may have been helpful (to say, hey version 12 looked and functioned properly, let’s go back to that) although that’s highly specific and likely would add significant complexity to an already complex Firebase system.
Evolution of What Happened
- Used a GPT to help craft a detailed prompt for Firebase.
- Studio built the app and it was a great start.
- Worked on functionality first like adding new books, importing + exporting a simple csv to use and update example data. Worked through a few errors although overall it was working well.
- Then focus turned to the design. The current layout was much like a row column layout and behind the items were lines connected related items. I was aiming to have something similar to what you may have seen with D3JS creating a force directed graph, essentially meaning everything is kind of spread out but connected to the things it is related to. After a few revisions, the mindmap was looking like what I was aiming for although every couple revisions, items would overlap one another which was an explicit instruction provided to never overlap (and even stated always have a minimum of 5px separating each item, which was anticipated to have avoided overlap). It was getting there, even worked through scroll zoom in/out which hit the mark.
- Moved on from spacing to images because one of my initial instructions was provide some default examples and pull their images from the internet. It got a few but missed most. A variety of the following changes were dedicated to finding images for publisher logos, authors and cover art. At one point it had about 80% of the authors, 1 logo and 1 cover art. A variety of recommendations were given around where to source the images, searching wikipedia and the publisher websites. Every iteration would confirm that the updated code corrected the source of the issue although no images displayed. When that was going nowhere, I looked in the browser console logs to find a bunch of errors so my next chat prompt included fixing the errors the browser had. Fixes were made and that’s when context faded away. Over the course of the subsequent revisions to fix+improve, I went into a loop/downward spiral of design loss to include less images being displayed, all formatting removed, removal of connecting lines and reverting back to a column row structure.
At this point, it seems starting over will invest less time than trying to continue down the same path because some of the earlier criteria appears to be lost.
Does anyone have a good solve for getting the criteria back on track,referencing an earlier version or locking in unchangeable criteria? Wanted to check before deleting and starting over.
Thanks in advance to any and all feedback!