At the end of a lesson, I was discussing with my tutor about where to best direct my Japanese learning efforts for the most gain right now. He suggested the “Instantaneous Composition Method” as the likely best candidate to help improve my speed in the production of speech.
I did a short bit of research on the method, and given that I already made a quiz app (here) that generates sentences of different JLPT N-levels, it seemed obvious that I had 90% of what I needed to apply the method.
I used that app as a base and introduced a horizontal card carousel UI that I thought was better fitting for the style of app.
In about 30 minutes I had it at a functional level, and then spent probably another maybe 1-2 hours of prodding around with refinements. These included refinements such as queuing up multiple sentence cards, rather than hitting the LLM each time (and then grabbing new sentences in the background as required to keep the queue healthy).
Next steps
I think some worthwhile next steps would be to expose some more configuration options for the types of sentences being generated. For example, selecting JLPT-level-appropriate grammar patterns to focus the generation on etc.