Exercise 1: Why an Image Model Needs a Genuinely Different Prompt Anatomy — Possible Solution ==================================================================== THE TWO ANATOMIES, SIDE BY SIDE ------------------------------ prompt1-2's own anatomy — Clarity, Context, Constraints, Output Format — is built around making a REQUEST unambiguous to a listener capable of understanding it. Clarity ensures the request itself is well-formed; Context gives the listener background needed to interpret it correctly; Constraints bound what an acceptable compliant response looks like; Output Format specifies the shape the listener's response should take. Every one of those four pillars presupposes a listener who parses the prompt, forms an understanding of what's being asked, and then acts on that understanding. This chapter's own anatomy — Subject, Style, Composition, Lighting, Camera/Lens, Medium — is built around describing a FINISHED IMAGE directly, with no listener in between. None of the six categories asks the model to interpret anything; each one just adds another descriptive anchor pulling generation toward a more specific target. WHY THIS ISN'T JUST A RELABELING ------------------------------ Per imgai1-1, an image model has no representation of a prompt as a request at all — only as a conditioning signal. A relabeled version of prompt1-2's own anatomy would still be built around the assumption that some part of the prompt is "instructing" the model and another part is "constraining" that instruction — a request-shaped structure. But per this chapter's own mechanism (imgai1-2), there is no request being parsed anywhere in the pipeline — only a text embedding steering a denoising process. A prompt anatomy for that kind of system has to be built from descriptive categories that map onto genuinely independent aspects of a finished image (what it shows, how it's styled, how it's framed, how it's lit, what device "shot" it, what physical medium it resembles) rather than from categories built around clarifying or bounding an instruction, because there is no instruction being followed in the first place. WHY THE TWO ANATOMIES SHARE NO REAL STRUCTURAL OVERLAP ------------------------------ None of prompt1-2's own four pillars maps cleanly onto any of this chapter's six — Clarity isn't "Subject," Constraints isn't "Composition." They're not the same categories wearing different names; they're built for two different jobs (bounding a request vs. describing a target image), which is exactly why this course couldn't simply borrow prompt1-2's own framework with new labels. WHY THIS WORKS AS AN ANSWER ------------------------------ It lays the two anatomies side by side, identifies what job each one is actually built to do, and uses imgai1-1's own "no request representation" claim to explain why the image-model anatomy had to be built from independently descriptive categories rather than adapted from an instruction-bounding one.