Critique and revision loop
Improve any draft in controlled rounds: critique against a rubric, prioritise, revise only the top issues, and see exactly what changed.
Fill in your details
Values are inserted into every prompt below. Nothing you type leaves your browser.
4 field(s) left — unfilled variables stay as {{placeholders}}.
1Critique without rewriting
Separating critique from rewriting stops the model from smoothing your voice away before you agree on what's wrong.
Critique the draft below. Do not rewrite it yet. Purpose of the draft: {{purpose}} Reader: {{reader}} Rubric (judge against these criteria): {{rubric}} Draft: {{draft}} For each rubric criterion, give a score from 1 to 5 and quote the passages that most affect the score. Then list the five most important problems in order of impact on the purpose. For each: the problem, where it occurs (quote), why it matters to this reader, and a suggested fix in one sentence.2Revise the top issues only
Pick the issues you agree with, usually 2-3. Changing everything at once makes it impossible to judge whether the revision helped.
Revise the draft to fix only problems I list here: [write the numbers, e.g. 1, 2 and 4]. Keep everything else, including my phrasing and structure, unchanged where possible. Return the full revised draft.
3Show the changes
Use the change log to accept or reject edits. Repeat steps 1-3 for a second round if scores are still below 4.
List every change you made as: original text -> new text -> which problem it fixes. Then re-score the revised draft against the same rubric and note any criterion that got worse.
Example input
A 720-word internal memo proposing a support chatbot pilot, for a sceptical CFO and COO.
Example output
Scores: Clear ask 2/5 ("we should explore..." appears in paragraph 4) | Evidence 3/5 | Risks 1/5 | Length 2/5 (720 words) | Plain language 4/5 Top problems 1. The ask is buried. Quote: "In light of the above, we should explore..." Why: a CFO reads the first paragraph and decides. Fix: open with the budget, duration and success metric. 2. No risks are discussed... Change log (after revising 1 and 2) "Customer expectations are evolving rapidly..." -> "I'm asking for £18k to run a 90-day chatbot pilot on billing questions." -> fixes #1
Troubleshooting
- Every criterion scores 4-5: the model is being polite. Add 'Assume this draft will be rejected; find out why.'
- The revision changes your voice: paste two paragraphs of your own writing and add 'Match this voice.'
- Changes aren't all listed: ask it to compare paragraph by paragraph.
Known limitations
- Scores are subjective and drift between runs; use them for direction, not as a measure.
- Very long drafts get shallow critique; work section by section.