By the end of this chapter you can read what the AI hands back the way an editor reads a draft, looking for what is wrong, instead of the way a believer reads it, looking for reasons it is fine.
You briefed the task well and let it run. Now it hands you a draft, and this is where most of the quality is won or lost. Not in the prompt. In how you read what comes back. There are two ways to read a first draft, and only one of them gets you good work.
The believer reads to be reassured. It looks finished, it sounds confident, so they accept it. The editor reads to improve. They assume the first draft is a starting point, hunt for the weak line, and send it back sharper. Same draft, two readers, completely different result. The whole of Part 4 is learning to read like the editor.
AVATAR OPENER · ~90s
Watch: the same draft read by a believer, then by an editor
HeyGen avatar · generated, consistent presenter
This is the most important reframe in the course after Chapter 1. A first draft from an AI is not an answer to grade. It is clay to shape. It will look more finished than a human first draft, which is exactly the trap: polish is not the same as right. Your job is not to judge it pass or fail. Your job is to push it from a six to a nine, one redirect at a time.
Read like a believer
It sounds confident and looks done, so you accept it. You look for reasons it is fine.
Result: you ship the first draft. Average, with your name on it.
VS
Read like an editor
You assume it is a starting point, find the weakest line, and send it back to be improved.
Result: you ship the third draft. Sharp, and recognisably yours.
Read what comes back to improve it, not to be reassured by it.
The first draft looks finished. That is the trap. Polish is not the same as correct, and confident is not the same as right.
WHAT AN EDITOR LOOKS FOR
Editing is not vague disapproval. It is a small set of specific questions you run over every draft. Learn these four and you can sharpen almost anything the model gives you.
What is generic?
Any line that could appear in anyone’s version. Cut it or make it specific to you.
What is unsupported?
A claim, number, or fact with no source. Flag it. Confidence is not evidence.
What is bloated?
The padding, the throat-clearing, the rule-of-three filler. The draft is usually too long.
What is not me?
The lines that sound like a press release. Your voice file says what you would never write.
SEE IT
Here is a real draft moment. You asked for a short product description. The first draft is the kind of thing that looks fine if you are a believer. Drag the handle to see what an editor does to it.
First draft (believer accepts)Our premium artisan candles are crafted with passion and the finest ingredients to elevate your space and create an unforgettable ambiance. Perfect for any occasion, they bring warmth and elegance to your home.
After an editor's passSoy wax, cotton wick, scented with real bergamot oil. Burns about 50 hours. Made in small batches in my kitchen in Burgas, so each one is slightly different. If you want a candle that smells like a perfume counter, this is not it.
generic, unsupported, not youspecific, honest, yours
The editor did four things: cut every generic phrase (premium, artisan, elevate), removed the unsupported puffery, trimmed the bloat, and put the real voice back. None of that required rewriting from scratch. It required reading critically and redirecting. Which brings us to the move: you do not fix the draft yourself, you tell it what is wrong and let it redraft.
Redirecting a draft like an editor
That draft is too generic and oversells. Specifically:
- "premium artisan" and "elevate your space" could describe any candle. Cut them.
- You claimed it is "perfect for any occasion." We cannot support that. Drop it.
- It is twice as long as it needs to be.
Rewrite it using only true, specific details: the materials, the burn time,
that I make small batches by hand. Keep my plain, honest voice. Add one line
that admits what it is not.
Ask Claude to write something real for you (a description, a post, an email). Resist the urge to accept it. Run the four editor questions over it: what is generic, what is unsupported, what is bloated, what is not me. Write a specific redirect naming the weak lines, and have it redraft. Compare the third draft to the first.
Right if your redirect names specific lines and reasons, not a vague instruction to make it better, and the redraft is shorter and more specific than the original.
Show the worked solution
The skill is specificity in the redirect. "Make it better" gets you a different flavour of generic. Naming the exact problem gets you a fix. Suppose the first draft of your bio says "a results-driven professional passionate about delivering value." A believer reads that and thinks, sounds professional, done. An editor reads it and marks three things at once: generic (results-driven, delivering value could be anyone), unsupported (what results? value to whom?), and not me (passionate is on your never list). The redirect writes itself: "Cut results-driven and passionate. Replace with what I actually do and one real number. Keep it to two sentences." The redraft comes back specific and yours. The lesson is that editing is not a feeling, it is four questions and a precise instruction. You are not a harsher reader. You are a clearer one.
WATCH FOR
✗You accept the first draft because it looks done. Looking done is the trap. The first draft is clay, not a verdict. Shape it.
✗Your only feedback is "make it better." Vague in, vague out. Name the specific lines and why they fail.
✗You rewrite it yourself by hand. Faster to tell it what is wrong and let it redraft. You direct, it executes.
✗You mistake confidence for correctness. It will sound sure of everything. Sounding sure is not evidence. Check the claims.
WHAT YOU LEARNED
The takeaways
Most of the quality is won in how you read the draft, not in how you wrote the prompt.
A believer reads to be reassured and ships the first draft. An editor reads to improve and ships the third.
Four editor questions: what is generic, what is unsupported, what is bloated, what is not me.
You do not rewrite by hand. You name the weak lines specifically and let it redraft. Direct, do not redo.
Your project · step ten
Take the last thing your thread project produced and edit it like an editor. Run the four questions, write one specific redirect, and keep the redraft. From now on, nothing your project makes gets shipped on the first draft. You read everything once with a cold eye. Next chapter goes deeper into the hardest case: when it is not just bland, but confidently, plausibly wrong.
The believer collects first drafts. The editor ships third drafts. The reading is the work, and the reading is yours.