Lesson 9 of 32 · Part One, the principles
The Ladder of Logical Progression
Step-by-Step Thinking
Split the job into named stages and ask to see the working at each one. Be honest about why: the old advice to add "think step by step" does very little on today's reasoning models, which already reason before answering. What still works is naming the stages yourself - load, then AC load, then comparison - because that is what gives you a rung to point at when a figure is wrong, instead of a single answer you must accept or reject whole.
How to use The Ladder of Logical Progression in ChatGPT
Drop the phrase and specify the output instead. On a reasoning model, "think step by step" is already happening and saying it adds nothing - sometimes it makes things worse, because the model performs visible deliberation instead of doing the work. What genuinely helps is telling it the shape of the answer: "Stage 1, Stage 2, Stage 3, each with its own heading, the arithmetic shown under each, and the assumptions listed before the numbers." That is a format instruction, not a thinking instruction, and it is what lets you say "stage two is wrong" later. Upload the figures as a file rather than typing them, and when an input was wrong, edit your original message so the bad number never enters the chain.
- Skip "think step by step" - name the stages as an output format.
- One heading per stage, assumptions listed before the numbers.
- Upload the figures as a file instead of typing them.
From GenAI for Everyone: Prompt Engineering Handbook, First Edition 2025, by Raja Rajeshwari Seshadri and Shreeram Iyer.
Drop the phrase and specify the output instead. On a reasoning model, "think step by step" is already happening and saying it adds nothing - sometimes it makes things worse, because the model performs visible deliberation instead of doing the work. What genuinely helps is telling it the shape of the answer: "Stage 1, Stage 2, Stage 3, each with its own heading, the arithmetic shown under each, and the assumptions listed before the numbers." That is a format instruction, not a thinking instruction, and it is what lets you say "stage two is wrong" later. Upload the figures as a file rather than typing them, and when an input was wrong, edit your original message so the bad number never enters the chain.
From the book Upload a file with the figures, and ask it to work through the steps one at a time. If a step is wrong, edit your message to correct the input rather than starting a new chat.
- Skip "think step by step" - name the stages as an output format.
- One heading per stage, assumptions listed before the numbers.
- Upload the figures as a file instead of typing them.
- Fix a wrong input by editing your message, not by correcting later.
- Verify any arithmetic that money or safety rests on.
Instead of
Can this house take one more AC?
ChatGPT version
I am a licensed electrician. A customer's house has a single-phase connection with a sanctioned load of [kW]. They want to add a 1.5-ton split AC. **Existing appliances:** [list with approximate watts] I do not need you to announce that you are thinking carefully. I need the answer laid out in three separate stages, so that if one figure is wrong I can tell you which. **Stage 1 - Present running load.** List every appliance with the wattage you used, note any you have assumed rather than taken from my list, then total it. **Stage 2 - The new AC.** State the running load and the starting load you are assuming for a 1.5-ton split, and say where those typical figures come from. Keep them separate - starting surge is not running load. **Stage 3 - Comparison.** Total against the sanctioned load. Say in plain words whether it is comfortable, close to the limit, or over it, and at what point in the day the peak would occur. Under each stage, show the arithmetic on its own line. Then finish with the general points I should check on site. I will verify everything with my own meter - this is a desk estimate, not a certification.
Then loop it
- Stage 2 is wrong - you used a high figure for the AC. Redo stage 2 for an inverter AC, show the new running and starting figures, and carry the change forward into stage 3. Leave stage 1 alone.
- Now write a simple two-line note for the customer in Tamil saying whether they should apply for a load increase, with no technical terms in it.
Why it worksAsking for each step lets the electrician spot exactly where an assumption was wrong instead of trusting a single yes or no.
Instead of
Why doesn't my bank balance match the books?
ChatGPT version
[Upload the cash book extract and the bank statement as files, with account numbers removed.] I am an accountant at a small firm in Pune. For [month] our cash book shows ₹[amount] and the bank statement shows ₹[amount]. **Unmatched items I have found:** [list of cheques issued but not cleared, deposits not credited, bank charges, interest] Prepare a bank reconciliation as a running table, one row per item, so that a wrong line is visible as a line rather than hidden in a total. **Columns:** item, nature (cheque not cleared / deposit not credited / bank charge / interest), add or subtract, amount, running balance, one-line reason. **Start from** the cash book balance in row one. Apply one item per row. Do not net two items together into a single row, even where they offset. **After the table:** state whether the difference is fully explained, and if not, the exact amount still unaccounted for. If an item in my list is ambiguous - if you cannot tell whether a figure is a charge or a credit - say so in that row instead of choosing.
Then loop it
- You treated the bank charges as a deposit. Correct that row and recalculate every running balance from that row down. Rows above it should not change.
- There is still ₹[amount] unexplained. List the common causes of a leftover difference, in the order I should check them, with what to look at in each case.
- Put the corrected reconciliation in Canvas so I can keep editing it as I find the remaining items.
Why it worksA running balance after every line makes the error visible at the exact rung where the reasoning slipped.
Instead of
Plan a Kashi trip for us.
ChatGPT version
My sister and I, both in our sixties, want to travel from Madurai to Varanasi and Prayagraj for about seven days next February. We prefer trains, cannot walk long distances, and want real time at the ghats and temples. Do this in three named stages, with the reasoning for each stage written before its conclusion - I want to be able to reject stage one without losing stages two and three. **Stage 1 - Getting there.** The main realistic ways to reach Varanasi from Madurai, with the pros and cons of each for two women in their sixties. Total door-to-door hours, number of changes, and how much standing or walking each involves. Do not recommend one yet. **Stage 2 - Order of places.** Given the route you consider best, a day-wise order for Varanasi and Prayagraj. Explain why that order and not the reverse. **Stage 3 - Rest and bookings.** Where the rest breaks go, and what must be booked early and how far ahead. Mark anything you are unsure of as [check] rather than giving me a number that sounds right. Then give the final plan.
Then loop it
- In stage 1 the overnight connection looks tiring for us. Show an alternative with one night's halt on the way, and set the two options side by side on the same criteria. Stages 2 and 3 can wait until I choose.
- Good, we will take the halt option. Now put the final plan in a simple day-by-day table, and add a short list of questions to ask the travel agent.
- Keep that table in Canvas so I can move the days around as our dates firm up.
Why it worksBreaking the trip into route, order and rest lets her question each decision before it is buried in a finished itinerary.
Questions
What is The Ladder of Logical Progression in prompt engineering?
Split the job into named stages and ask to see the working at each one. Be honest about why: the old advice to add "think step by step" does very little on today's reasoning models, which already reason before answering. What still works is naming the stages yourself - load, then AC load, then comparison - because that is what gives you a rung to point at when a figure is wrong, instead of a single answer you must accept or reject whole.
When should I use this technique?
When the answer rests on a calculation or a chain of assumptions - an electrical load, a reconciliation, a route - and you need to audit the reasoning rather than trust the conclusion.
How do I apply it in ChatGPT?
Drop the phrase and specify the output instead. On a reasoning model, "think step by step" is already happening and saying it adds nothing - sometimes it makes things worse, because the model performs visible deliberation instead of doing the work. What genuinely helps is telling it the shape of the answer: "Stage 1, Stage 2, Stage 3, each with its own heading, the arithmetic shown under each, and the assumptions listed before the numbers." That is a format instruction, not a thinking instruction, and it is what lets you say "stage two is wrong" later. Upload the figures as a file rather than typing them, and when an input was wrong, edit your original message so the bad number never enters the chain.
What are the limits of this technique in ChatGPT?
Be sceptical of displayed arithmetic. Named stages make a calculation auditable, which is not the same as correct - the working can look immaculate and still contain a slip, so check the sums that matter yourself. Avoid stacking corrections, too: if the load figure you gave was wrong, editing your first message is far cleaner than six messages of "actually it was 3 kW", because the old figure keeps resurfacing in a long thread. And on anything touching money, electricity or safety, treat the staged answer as a draft to verify on site or with a professional, never as the finding.
How do I improve the answer after the first try?
Step-by-step answers are easy to loop because you can point at a single rung. Read each step and ask: Is this number right? Is this assumption true for me? When one step is wrong, do not ask for a fresh answer - say "Step two is wrong because…; redo from step two." The steps before it stay, and the fix flows forward. For money, health or safety work, treat the steps as a draft to check yourself or with a professional, never as the final word.
Loop it
Step-by-step answers are easy to loop because you can point at a single rung. Read each step and ask: Is this number right? Is this assumption true for me? When one step is wrong, do not ask for a fresh answer - say "Step two is wrong because…; redo from step two." The steps before it stay, and the fix flows forward. For money, health or safety work, treat the steps as a draft to check yourself or with a professional, never as the final word.
This lesson on the other tools
The technique is the same; the way you write it is not. Claude, Gemini.Perplexity.