Well, despite only being on a $20~ ChatGPT Plus plan, and being subject to 5hr usage limits, I’ve managed to make a lot of progress on the various parts of the tape.deck app. Last time, I only had some wireframes, and a simple spec, now I have a lightweight Jira clone running in ChatGPT Work that has split everything down into tasks (and sub-tasks) and stages, story pointed it all out, and figured out a means to estimate 5hr usage % for each given card relatively accurately ahead of time.

Because I’ve got a structured way to track and step through tasks in some kind of logical order, I was able to get up and running a functional Registry app view:

The Registry was its own set of tasks to extract, review, promote, or ignore both symbols (entities) and links betwen them from the 142 books in the Lorecore series. Now – as far as I can tell anyway – all the important bits and bobs in the multi-verse of my story now each have their own registry entries, which track who or what that thing is, where it is referenced in the series, and links out to other entities, etc. This runs either in the browser in ChatGPT (where it is small and cramped) or in another browser running off a port on localhost.

Then there is now a functional knowledge graph or “story space” view which shows those symbol units and their connections as clusters. It looks something like this:

It’s all evolving a bit, but implementing a Jira-like tracking system with numbered “cards” for issues with semi-accurate build estimates has really helped change the game for how this process can go. I still have to work around the 5 hour limits, but I can do so with a much clearer plan of what can be accomplished, and not have things get stuck in a void of “Thinking” forever, until its used up its allocation, without ever finishing anything… (One trick I devised for planning is instruction the system to decide what to focus on to achieve in the next five minutes that will have the highest overall impact for moving forward.)

I’m really really close to pulling the trigger on the next higher tier subscription so I can reduce some of this friction and waiting, but I’m scared that if I do that, I will just end up using it even more than I already am – and maybe my brain is already spinning fast enough! Plus, there’s the $$$ question too. But the more I use this, the more I think how absolutely insane it is to be able to do all this solo, with no engineer, for $20 a month. Even $100 or $200 for a month if you have a really targeted project you’re doing is maybe the deal of the century…