Anatomie AI — the design engine¶
Anatomie AI is a system that designs clothes. Every day it generates editorial fashion lookbook images for the Anatomie brand, learns from every grade Shawn gives it, and uses what it learns to design better the next day.
It is three connected surfaces over one engine:
| Surface | What it does |
|---|---|
| The Engine | composes the design prompts, renders the images with Imagen 4 Ultra, segments every garment so it can be graded individually, runs the experimentation Lab, and learns from feedback |
| ReviewHub | where Shawn grades the looks — a whole image or a single garment on a single model — as outlier, catalyst, on-brand, off-brand, or ugly |
| The Strategist | where you control the engine and watch it learn: tune the daily run, browse everything it knows, and follow how winning designs evolve |
The one-paragraph version¶
Every day the engine composes a batch of looks from a curated roster of designers, colorways, and garments. It pushes new design language on every look — new collars, closures, panel constructions, fabric treatments — while a locked scene template keeps every image on the same clean white-backdrop stage. Shawn grades the renders. The winners become seeds that the Lab clones and evolves; the losers become bans that fence bad language out of every future prompt; and everything becomes notes that steer the next day's designs. The loop compounds: what graded well yesterday shapes what gets made today.
Where to start¶
- New to the system? Read The loop, then the grading guide — those two pages are enough to use it.
- Want to understand the design machinery? Generation pipeline, Innovation engine, and The Lab.
- Running it? Control reference, Observability, and Deploy & verify.
- Lost in a term? The glossary defines everything.