Every time I wanted to experiment with another model on my video prompt, I had to explain my rules again, even with skills. So it was never a fair test.
This time I put 9 lessons from the last few weeks into
@evermind's EverOS Cloud: rough sketch sheet as the reference, style lock, three beat finishers, and first frame already moving.
Then I opened a new Claude and a new Codex chat, had both recall the same memory first, and gave them the same Halloween brief and the same reference.
Both pulled the same rules and used them. Claude gave me the better fight. Codex got closer on the paint style and the long finisher. Neither kept the ending to one second.
Which one do you like?
Try it here:
everos.evermind.ai (Pro is 3× monthly usage for 6 months right now)