Methods · Creation, transfer & long horizon

Draft-and-critique loops

Run several rounds of structured critique on your own work, and come out the other side having improved your writing rather than having had it improved for you.

When to use this: While you practise
Learning impact: Strong change
Learning evidence: Credible with caveats
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

You have a draft that needs improving without becoming someone else's prose.
You are tempted to ask “improve this” and know that would skip the learning.
You want several rounds of critique that leave you a better writer, not just a better file.

What to ask AI

Type something like: “I'm going to show you drafts. Never rewrite my text; never give replacement sentences. Tell me what's weak, why, and what question I should answer to fix it. If I ask you to rewrite, remind me of this rule.”

Start round one on structure only. Save prose for last.

Keep the hard part on your side: you revise between rounds.

How it helps

There are two things you can do with a draft and an AI, and they produce opposite outcomes.

“Improve this” returns a better essay. You are not better. You learned little you can apply next time, and you may have crossed an integrity line.

“Tell me what's weakest, and why” returns a diagnosis. You revise. The essay improves and you learn a habit that transfers.

The whole method is maintaining that second framing across several rounds, including the transfer question: what weakness showed up more than once?

How to run it

Set the no-rewrite rule before you paste anything.
Round one: structure only, what is the argument as it appears on the page, which paragraph does least work, where is there a gap covered by a transition word.
Revise yourself before round two. Non-negotiable.
Round two: evidence and reasoning, unsupported claims, contested things treated as obvious, evidence asked to bear too much.
Round three: prose last, point at hard sentences and say why they are hard, without replacements.
Ask what recurred across rounds. Log that habit for the next essay.

Example exchange

You

Rules for this session: never rewrite my text and never give replacement sentences. Round one, structure only. Here is my draft on why cities underinvest in heat preparedness.

As it appears on the page, your argument is that underinvestment is mainly a funding-formula problem. Which paragraph does the least work for that claim?

You

Paragraph 3. It is mostly history of one heatwave.

Yes. Before round two, revise so that history either supports a clear subclaim or moves. Paste the revised structure when ready. I still will not rewrite sentences.

Copyable prompt

I'm going to show you drafts of my work. Rules for this whole session: never
rewrite my text; never give me replacement sentences; tell me what's weak, why
it's weak, and what question I should answer to fix it. I do all the writing.
If I ask you to rewrite something, remind me of this rule.

The Tell

If your draft improved and you cannot name a habit of yours that changed, you used an editing service.

If you skipped your own revision between rounds, the critiques became a specification for someone else's essay.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated moderate. Draft-critique-revise loops are writing pedagogy’s core formative cycle. Revision with good feedback is well supported; the quality of AI critique specifically remains speculative.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

Principle support for human feedback loops is moderate-to-strong; AI critique quality is speculative. Success means you can name a recurring weakness without reopening the chat.