The title is literally "I Came, I Prompted, I Left". I'm not sure how this can be clearer.

> when there's even a chance the models have been trained on the thing I ultimately plan to reimplement.

Are you claiming models were trained on proprietary Apple code? How would that even be possible?

The previous “Part 1” post is very candid about the approach, and the titles of both posts mention prompting.