The memory that makes everything else personal

It learns from decisions, not declarations

When the AI generates options A, B and C and you pick B, that selection is recorded. If you edit B before using it, the original and your edit are both kept — the difference is signal. If you export the result, confidence in that direction goes up.

• Viewed — weak signal
• Saved — moderate signal
• Selected — strong signal
• Edited and exported — very strong signal
• Explicitly marked “I like this style” — the strongest signal there is

Confidence you can inspect

No preference forms from a single action. Evidence accumulates and confidence saturates gradually — a preference like “maximum 3 words of thumbnail text” only reads as strong after enough consistent decisions back it up.

On the Creator Brain page every preference shows exactly what it's based on: how many decisions, of what kind. If the system thinks you prefer dark thumbnails but that's only true for your AI videos, you can scope it — preferences can be conditional on topic.

Your memory, your controls

Everything the system learns is yours to manage:

• Correct — confirm a preference so it's never overwritten by new inference
• Remove — delete a preference outright
• Disable learning — stop recording decisions entirely, at any time
• Rejected ideas are remembered too, so the AI stops regenerating concepts you've already said no to