“Make it better” is a weak instruction. “Keep the opening, remove the sales language and ask one clear question” gives the writer something useful to change.
Nynch’s email draft feedback separates a one-off correction from a preference you want remembered. That gives you control over what the system learns from the way you edit.
Choose the right kind of feedback
Open Improve draft on a supported email draft. Explain the change in plain language.
- Update this draft revises the current message.
- Update & remember also saves the explicit writing feedback for future work.
- Learn from my edits is available when you have changed an AI draft and want those edits to inform the writing style.
Keep reusable preferences distinct from facts about one recipient. “Use British spelling” can apply broadly. “This buyer needs the contract by Friday” belongs to that conversation.
Review the result before sending
Compare the revised message with your intent. Check names, links, promises and the actual ask. Learning your preferred wording does not remove the need to verify the content.
The interface also provides Undo writing-style change for a saved style update. If a remembered rule was too broad, correct it rather than compensating in every later prompt.
The value compounds through explicit choices
You spend less time repeating the same editorial instruction, while keeping the decision to remember it visible. Merely opening, reading or sending a draft is not the same as explicitly teaching a writing preference.
The screenshot shows the saved-preference area, not a completed feedback event. For broader reusable instructions, see Preferences and Playbooks. Learning here means saved guidance used by the writing workflow, not private retraining of the underlying AI model.