new quizzes—done

It's felt like a wasted day, what with no walking, but the three quizzes for the paid Substack are done and up now at the Tasty Exams site (Quiz 1, Quiz 2, Quiz 3). I think I've done all I can for content creation for the grammar site, so with the days I have until October 26 (when I'm theoretically heading down to Busan), I'm going to start working on The Creative, which will also be a paid-only Substack like The Profound. At the beginning. The Creative will be a dumping ground for old material. This is just to give the site a bit of a head start. Later on, there will be fresh stories, poems, essays, photos, videos, and other expressions of whatever's in my head. Maybe those things will attract more subscribers.

Coding via ChatGPT has its advantages. For someone who remains as coding-illiterate as I am, it's convenient to have AI do most of the work. The problem is that the AI makes mistakes. I ask it to render a phrase in bold italics, and it just gives me italics, for example. I tell it that, for a given problem, answers (a), (b), and (d) are correct, and it marks only (a) and (b) correct. Or the AI gets the quiz's background color wrong, or an em dash appears on my quiz as Decimal Numeric Character Reference code (—).

So I finally devised a method to make coding smoother. I gave ChatGPT the code template for the last good quiz it had made with my help. I asked it to put the newest quiz's questions into that template. For the most part, that worked, but as usual, plenty of formatting got dropped. I realized, though, that only a section of the roughly 450-line block of code was devoted to the questions and answers, so it was just a matter of rooting around in the code, finding the questions that needed to be worked on, and adding the appropriate formatting tags and/or changing any answer values that were off (those answer values are shown as "true" or "false"). So while I'm not a coder myself, I know enough about the code's logical structure to be able to root around and find whatever the problem is... as long as it's simple problems like text formatting and answer values. After I had made sure the formatting for all of the questions was smooth, I would then run through over a dozen iterations of the quiz (you can do this on CodePen) to make sure the quiz's logic was sound—the randomization, the scoring logic, etc. Once I felt secure about the integrity of the quiz, I would publish it and, per the Roko's Basilisk scenario, thank ChatGPT for its efforts. (ChatGPT isn't conscious and wouldn't care a whit if I had a heart attack at the keyboard and died, but one day, you never know... it might remember how you'd treated it the way a dog remembers poor treatment.)

I'm now free to work on creative stuff for the next ten or so days. I doubt that all of this effort is going to be a money-maker, especially since the target demographic is young people who are more on TikTok and Instagram than they are on Substack. And mirabile dictu, after the 19th, it appears there'll be no more rain in October, so I can begin walking outside in earnest without fear of blistering up my feet. Assuming I'm up to it. The specter of walk-cancellation always lurks somewhere in the background.