Lesson 7 of 32 · Part One, the principles

The Virtue of Persistent Effort

Iteration

Your first prompt is a hypothesis, not a finished article. Iteration means reading the answer like a detective, naming one thing that is wrong - the prices, the tone, the length - and fixing only that in your next message, so each round keeps what already worked and the good parts survive.

Reach for it whenWhen the first answer is nearly right but wrong in one way you can name, and when the thing you are making is something you will send, print or reuse rather than glance at.

How to use The Virtue of Persistent Effort in ChatGPT

Canvas is the iteration surface here, and asking for it early changes the whole exercise: instead of six copies of a price list scrolling up the chat, you have one document you keep correcting, and you can select a single package and ask for that line alone to change. Know the difference between the two buttons as well. Regenerate rolls the same prompt again and shows you a different attempt, which tells you how much of the first answer was luck. Editing your own message rewrites the prompt and branches the thread, which is what you want when a round went badly and you do not want it colouring the next one. Standing background - who you are, what you sell, how you like answers - belongs in Custom Instructions so the iteration starts from round two.

  • Ask for Canvas on anything you will refine more than twice.
  • Edit your message to branch; regenerate only re-rolls the same prompt.
  • In Canvas, select one paragraph and change only that.

From GenAI for Everyone: Prompt Engineering Handbook, First Edition 2025, by Raja Rajeshwari Seshadri and Shreeram Iyer.

Canvas is the iteration surface here, and asking for it early changes the whole exercise: instead of six copies of a price list scrolling up the chat, you have one document you keep correcting, and you can select a single package and ask for that line alone to change. Know the difference between the two buttons as well. Regenerate rolls the same prompt again and shows you a different attempt, which tells you how much of the first answer was luck. Editing your own message rewrites the prompt and branches the thread, which is what you want when a round went badly and you do not want it colouring the next one. Standing background - who you are, what you sell, how you like answers - belongs in Custom Instructions so the iteration starts from round two.

From the book Use regenerate to see another attempt, or edit your message to change the instruction. Canvas lets you mark a single paragraph and ask for changes to just that part.
Worth knowingRegenerating is not refining. It feels like progress because something new appears, but you have not told it anything, so you are gambling again rather than steering. The other thing to watch is inertia in a long thread: after ten rounds it clings to the shape of draft three, and new instructions get bent to fit the old draft. When that happens, edit the original message instead of adding another correction on top. Saved memory can also surprise you - a tone you asked for once in a different chat can come back as a standing preference, so if a loop starts behaving strangely, check what it thinks it knows about you.
  • Ask for Canvas on anything you will refine more than twice.
  • Edit your message to branch; regenerate only re-rolls the same prompt.
  • In Canvas, select one paragraph and change only that.
  • Name one flaw per round: price, tone, length - not "improve it".
  • Put your shop or trade details into Custom Instructions once.
The One-Shot Surrender vs. The Hammered Price List

Instead of

Make a Diwali offer list for my parlour.

ChatGPT version

I run a small beauty parlour in a residential area of Indore. Most of my clients are working women and college girls.

**My normal prices:** facial ₹800, threading ₹60, manicure ₹400, hair spa ₹900.

**Make me a Diwali offer menu for WhatsApp** with five packages - combinations such as threading plus facial, or a bridal trial.

For each package show: package name, what is included, normal price, offer price in ₹.

**Rules for this first round**
- Every discount between 10 and 15 percent of the normal total. Show the arithmetic beside each offer price so I can check it.
- Friendly and festive, the way I would speak to a regular customer.
- Put it in Canvas. I will be correcting this several times and I want one document, not five copies.

Before the menu, give me two lines on which combinations you think will sell best near a college and why - I may disagree.
Open ChatGPT 870 characters

Then loop it

  1. The packages are good but the offer prices are too low. In the Canvas, keep every package and recalculate so each discount lands between 10 and 15 percent. Change nothing else.
  2. Now select the package names only and rename them with short, catchy Hindi names. Then add one line at the end about booking on WhatsApp.
  3. Last thing: shorten the whole menu in the Canvas so it fits one phone screen without scrolling. Then make a second Canvas for New Year using exactly this format, so I only swap the festival name next time.

Why it worksEach follow-up fixes one flaw the first answer revealed - price, name, length - until the menu is ready to send, and the last step turns it into a reusable format.

The Abandoned Story vs. The Refined Product Tag

Instead of

Write about my Pochampally saree.

ChatGPT version

I am a third-generation ikat weaver in Pochampally, Telangana, and I sell cotton and silk sarees online.

Write a short story card, about 60 words, to tuck inside each saree package.

**It must get across**
- The saree is handwoven.
- The tie-and-dye pattern is planned thread by thread before weaving begins.
- One saree takes about [number] days.

**Voice:** warm, as if I am speaking to the buyer myself.

Give me three attempts that differ in approach, not in wording - label them A, B and C. One plain and factual, one that leads with the time it takes, one that sounds like a note from my family. I will pick one and we will work on that one only.

Then open the one I choose in Canvas, because I will be adjusting it word by word.
Open ChatGPT 737 characters

Then loop it

  1. B is closest, but it reads too much like an advertisement. Take B into Canvas and make it humbler and more personal - a thank-you note from my family, not a sales line.
  2. Better. Keeping that voice, give me three versions in the Canvas: one for silk, one for cotton, one for someone buying it as a gift.
  3. Now add a Telugu translation of the cotton version underneath, and list any lines that may sound unnatural in Telugu so I can check them with my daughter.

Why it worksThe weaver treats the first draft as feedback, steering tone, then variety, then language, instead of giving up when the first card sounds like an advertisement.

The Quick Answer vs. The Tested Plan

Instead of

How to save money as a cab driver?

ChatGPT version

I drive a cab full-time in Bengaluru through ride apps, six days a week.

Build me a simple weekly table I can fill in each night, with columns for: date, hours driven, trips, total earnings, fuel or CNG cost, app commission, money taken home.

Below the table, explain in five short points how to read it at the end of the week to spot my best and worst days.

Use simple English. Put the table in Canvas - I will be adding columns and then pasting my real numbers into it, and I want one sheet that keeps improving rather than a new one every time.
Open ChatGPT 550 characters

Then loop it

  1. In the Canvas, add two columns to the same table: one for toll and parking, one for small expenses like tea and food. Leave the five points as they are.
  2. Here is last week's data: [paste numbers]. Fill the Canvas table with it and tell me which day, and which hours of the day, gave me the best take-home per hour.
  3. Good. Now turn the five points into a short checklist I can follow every Sunday evening, and add a simple Kannada version of the checklist below it in the same Canvas.

Why it worksThe loop moves from an empty template to his real numbers to a weekly habit, each round built on what the previous answer got right or missed.

Questions

What is The Virtue of Persistent Effort in prompt engineering?

Your first prompt is a hypothesis, not a finished article. Iteration means reading the answer like a detective, naming one thing that is wrong - the prices, the tone, the length - and fixing only that in your next message, so each round keeps what already worked and the good parts survive.

When should I use this technique?

When the first answer is nearly right but wrong in one way you can name, and when the thing you are making is something you will send, print or reuse rather than glance at.

How do I apply it in ChatGPT?

Canvas is the iteration surface here, and asking for it early changes the whole exercise: instead of six copies of a price list scrolling up the chat, you have one document you keep correcting, and you can select a single package and ask for that line alone to change. Know the difference between the two buttons as well. Regenerate rolls the same prompt again and shows you a different attempt, which tells you how much of the first answer was luck. Editing your own message rewrites the prompt and branches the thread, which is what you want when a round went badly and you do not want it colouring the next one. Standing background - who you are, what you sell, how you like answers - belongs in Custom Instructions so the iteration starts from round two.

What are the limits of this technique in ChatGPT?

Regenerating is not refining. It feels like progress because something new appears, but you have not told it anything, so you are gambling again rather than steering. The other thing to watch is inertia in a long thread: after ten rounds it clings to the shape of draft three, and new instructions get bent to fit the old draft. When that happens, edit the original message instead of adding another correction on top. Saved memory can also surprise you - a tone you asked for once in a different chat can come back as a standing preference, so if a loop starts behaving strangely, check what it thinks it knows about you.

How do I improve the answer after the first try?

Iteration follows this book's five-step loop: Aim, Prompt, Read, Refine, Reuse. After the first answer, read it like a detective and name one specific flaw - too long, wrong tone, a price that is off. Fix only that in your next message: "Keep everything but make the discounts 10 to 15 percent." Fixing one thing at a time keeps the good parts. When the answer is finally right, ask the AI to rewrite your original prompt with all the fixes included, and save that improved prompt for next time.

Loop it

Iteration follows this book's five-step loop: Aim, Prompt, Read, Refine, Reuse. After the first answer, read it like a detective and name one specific flaw - too long, wrong tone, a price that is off. Fix only that in your next message: "Keep everything but make the discounts 10 to 15 percent." Fixing one thing at a time keeps the good parts. When the answer is finally right, ask the AI to rewrite your original prompt with all the fixes included, and save that improved prompt for next time.

This lesson on the other tools

The technique is the same; the way you write it is not. Claude, Gemini.Perplexity.