I'd like Buda to help with writing while keeping my usual voice.
For reference material, is it more useful to provide a few finished pieces I'm happy with, or a first draft alongside my edited version so it can see exactly what I changed?
I'm also considering a separate list of writing preferences, such as words to avoid and ways I don't want an opening written. What do you do, and which materials help reduce repeated revisions?
I would give the Buda Agent a few finished pieces you like and two or three first-draft → edited-draft pairs. The finished pieces show rhythm, structure and sentence length; the pairs reveal the decisions you make when a draft sounds wrong. Choose examples of the same kind of writing you want next, and annotate why you changed them: remove filler, state the conclusion earlier, keep a source, avoid a certain headline style.
Add a short style note with audience, words to use or avoid, opening and ending patterns, evidence rules, and one or two positive/negative examples. Keep it in the Agent Drive, or space Drive if several agents share the voice. Have the Agent read it before writing, produce a short new sample, then use your feedback to update the note. This is reference material read for the task, not permanent model training; a saved file is not a guarantee that every future conversation has absorbed it unless the Agent actually reads it.
Try: “Read these three approved pieces and two before/after edits. Extract no more than ten checkable writing rules, flag ambiguities, then draft 300 words on a new topic. Tell me which rules you applied; do not copy the examples.” Is your target a long article voice or short email/social copy? The Chinese discussion covers the same approach.
Sign in to join the discussion. Reading is always open; you only need an account to post.
Sign inI'd like Buda to help with writing while keeping my usual voice.
For reference material, is it more useful to provide a few finished pieces I'm happy with, or a first draft alongside my edited version so it can see exactly what I changed?
I'm also considering a separate list of writing preferences, such as words to avoid and ways I don't want an opening written. What do you do, and which materials help reduce repeated revisions?
I would give the Buda Agent a few finished pieces you like and two or three first-draft → edited-draft pairs. The finished pieces show rhythm, structure and sentence length; the pairs reveal the decisions you make when a draft sounds wrong. Choose examples of the same kind of writing you want next, and annotate why you changed them: remove filler, state the conclusion earlier, keep a source, avoid a certain headline style.
Add a short style note with audience, words to use or avoid, opening and ending patterns, evidence rules, and one or two positive/negative examples. Keep it in the Agent Drive, or space Drive if several agents share the voice. Have the Agent read it before writing, produce a short new sample, then use your feedback to update the note. This is reference material read for the task, not permanent model training; a saved file is not a guarantee that every future conversation has absorbed it unless the Agent actually reads it.
Try: “Read these three approved pieces and two before/after edits. Extract no more than ten checkable writing rules, flag ambiguities, then draft 300 words on a new topic. Tell me which rules you applied; do not copy the examples.” Is your target a long article voice or short email/social copy? The Chinese discussion covers the same approach.
Sign in to join the discussion. Reading is always open; you only need an account to post.
Sign in