◑ Slot H
One real fashion brief a day, fired as a 2×2 grid: the full director prompt and a ~500-char
Opus compression, each on Flux Ultra and GPT Image 2.5. The same compressed string goes to
both engines, so nothing moves but length and engine. No critic score is recorded: the comparison is yours.
A vs B — does simplicity win? (isolates length)
ultra vs gpt25 — is “short beats long” a law, or just an Ultra limit?