Exercise 1: Why the Capstone Stopped Fighting the Chalkboard Text — Possible Solution ==================================================================== WHAT THE CHAPTER'S WARN-BOX SAYS ------------------------------ Per this chapter's own warn-box, "this is a structural limitation, not a wording bug... Rather than fighting it further, this workflow makes a deliberate choice: keep the chalkboard visually present as background texture... but don't attempt to force specific legible text onto it." WHY THIS IS A STRUCTURAL LIMITATION, PER imgai1-8 ------------------------------ Per imgai1-8, "nothing in the training signal ever represents letters as discrete symbols with a required exact sequence, only as visual textures that loosely correlate with certain image regions." No amount of rewording the subject clause changes what the model actually learned about text during training — the model was never taught the concept of a discrete, orderable character sequence at all, so no prompt phrasing can supply a rule that was never encoded into the model's own representation of "lettering" in the first place. WHY MITIGATION AND ELIMINATION ARE GENUINELY DIFFERENT GOALS ------------------------------ Per imgai1-8's own closing distinction, "'less frequent and less severe' is a genuinely different claim from 'solved.'" Continuing to reword the subject line in search of legible chalkboard text would be chasing elimination of a structural limitation through a technique (prompt rewording) that this course has already established doesn't reach that layer of the problem at all — the same category of mistake imgai1-8 already warned against for hands (weighting a hand descriptor amplifies an existing fuzzy pattern; it doesn't add a missing rule). Prompt-level rewording is, at most, a mitigation lever — it cannot manufacture the discrete-symbol representation the model was never trained to have. WHY THE DELIBERATE WORKAROUND IS THE CORRECT RESPONSE ------------------------------ Recognizing this, the workflow makes a genuinely different kind of choice: instead of continuing to spend iteration passes chasing an outcome the mechanism can't reliably deliver, it redefines what "success" looks like for that one element — a chalkboard that reads convincingly as "a menu board" from a design standpoint, without requiring specific legible words, still serves the original brief's own actual purpose (a warm, plausible coffee-shop scene) without pretending the underlying limitation had been solved. WHY THIS WORKS AS AN ANSWER ------------------------------ It explains, using imgai1-8's own mechanism, why no amount of prompt rewording can supply the discrete-symbol representation the model never learned, and distinguishes the chapter's own mitigation-vs- elimination framing to justify redefining the goal rather than continuing to iterate against a structural wall.