Welcome to Prompt, Tinker, Innovate—my AI playground. Each edition gives you a hands-on experiment that shows how AI can sharpen your thinking, streamline your process, and power up your creative work.
This week's playground: use a second model to review, not rewrite.
Here's a pattern I see constantly, and I used to do it myself. You draft something in one AI tool, go back and forth until it feels done, and call it final.
But "final" often just means you stopped noticing the problems.
By that point, the draft has already had two opinions: the AI that helped create it and yours. The problem is that neither of you has much distance anymore. You both helped make the choices, revisited the same paragraphs, and shaped the same story.
That's when I bring in a third opinion: a different AI that hasn't seen the draft before.
It comes in cold. It never wrote the thing, so it has nothing to defend. And that distance makes it much better at spotting where the story sags, where the logic gets muddy, or where a reader might tune out.
When I bring in a second AI
A quick email and a $100K proposal don't need the same review process.
- Everyday content: a quick client email, internal update, or routine post—draft it with AI, polish it myself, send it.
- Higher-impact content: a major proposal, executive presentation, or important external piece—draft it with AI, get it into great shape, then hand what I think is the final version to a different AI with one narrow job.
The second AI isn't there to catch typos—my original AI tool and I already handled that. It's there to catch the things we've stopped seeing after reading the draft fifteen times: the paragraph that buries the actual ask, the opening that takes too long to get moving, or the section that reads fine to me but falls flat for someone else.
Your AI experiment: Try this prompt
👉 Time to tinker: The part people get wrong is asking for a general "review." That invites a rewrite, and a rewrite in someone else's voice is not what you want. Be specific about the job.
📝 Prompt:
Please review this near-final [content type] written for [audience]. Don't rewrite it. Please review it for three things only: clarity, storytelling, and reader momentum. Tell me where it drags, where the story loses the reader, and where you'd stop reading if you were skimming this. Please provide recommended fixes if you have them.
Narrowing the job is what makes the feedback useful. Without those constraints, you get a second draft. With them, you get a fresh opinion—which is what you actually needed.
💡 Pro tips: Make this a standard practice
- Vary who gives you that fresh review. ChatGPT, Gemini, and Claude each have different reflexes. If you always use the same second model, you'll eventually just be optimizing for what that one model likes.
- And keep the review short. If you get back three paragraphs of praise and one buried critique, ask again: "Skip what's working. Tell me the one thing most likely to make someone stop reading."
What did you discover?
Try this on something you already consider done—a proposal, a post, an email you were about to send. Give a different model the narrow job above and see what it flags. Reply and tell me what it caught that you missed.
📩 Not subscribed yet? Hit the button at the top. You won't want to miss what's next.
