What it learns from your edits
Which corrections it notices, when they carry into later drafts, and how to correct what it learned.
When it learns
When a draft goes out, ReadyToEdit compares the text that went out with the text it wrote. Publishing its draft untouched teaches it nothing, because agreement is not a correction. What you changed is the signal.
One edit is evidence, not a rule. A correction starts to shape later drafts once you have made it across several of them.
What it notices
It learns from what it can see in your edits, not from a guess at your taste. These are the corrections it looks for, and the sentence it writes down for each.
| Your correction | What it writes down |
|---|---|
| Taking the emoji out | “You take the emoji out.” |
| Removing exclamation marks | “You’d rather state a thing than exclaim it.” |
| Removing hashtags | “You don’t post hashtags.” |
| Rewriting marketing language | “You cut the marketing words out of what I write.” |
| Making it shorter | “You write shorter than I do.” |
| Making it longer | “You add more than I give you.” |
| Replacing the opening | “My openings do not survive. You write your own.” |
| Replacing the ending | “You end pieces differently from how I do.” |
| Moving parts around | “You move things around — my order is not yours.” |
| Rewriting almost all of it | “For pieces like this, my draft is a starting point rather than a draft.” |
Correct what it learned
What it learned is listed on What I’ve learned, each statement with the edits behind it and a count of drafts for and against it. Only sometimes weakens a statement; Not true? retires it.
One list per workspace
What it learns applies across your whole workspace: all of its projects, clients and channels share one list. A rule only one client has, such as a word they never use, belongs in that project’s instructions in Style & context instead.
It never improves anyone else’s results, and your material is never used to train shared models.