Coach a note
Ask a model for a closer read of a draft, with a spending limit and a Stop button.
The coach reads the whole draft against your profile. It marks the places that need work, gives a short summary and a score out of 10, and hands its marks to the same card as the local checks. It runs only when you ask.
Before you start#
- Model Hub is installed and enabled, with a provider set up in it. Learning Editor needs Model Hub's API 0.10 or newer.
- A model is set for the coach. The coach's role in Model Hub is
learning-editor.coach, on the careful tier unless you change it. To pick a model, open Settings → Learning Editor and select Choose the coach’s model. - A profile helps but isn't required. Without one, the coach uses plain defaults.
Steps#
- Open the note.
- Run Coach this note from the command palette, or select Coach in the panel.
- Watch the status bar. It follows the run: for example Coach: step 2 of 8, counting habits, $0.01 of $1.00.
- When the run ends, a notice says how many places the coach marked and its score. The marks appear in the note.
To stop a run, click the status bar item or select Stop coaching in the panel, where Coach becomes Stop coaching while a run is going. A stopped run saves nothing.
What you get#
- The coach's marks, in the note, on the card. Apply and dismiss them as any other mark: see Work with marks.
- Its summary and score in the Coach’s last read section of the panel.
Every mark is checked against your note before it's saved. A mark whose words can't be found in the note is left out, and the notice says how many couldn't be placed. Where a coach mark overlaps a local one, the coach's mark is shown.
What it costs#
Each run is one model run through Model Hub, billed by your provider. It stops at Spending limit per run (US$1 unless you change it, in Settings) and after 8 steps. Model Hub's own limits on the tier apply too. If the run reaches a limit after it handed in marks, those marks are kept; if not, nothing is saved.
What is sent#
The note's text and your profile, to Model Hub, which sends them to the provider you chose. Nothing else from your vault. See Privacy and costs.