The Ladder Rule
You know when the AI starts strong and by the third paragraph has already forgotten what you asked for? There's a reason this happens — and a technique that fixes it for any task, no matter how complex.
The problem with all-at-once requests
The bigger the task, the more likely the AI gets lost along the way. It starts strong, but by the third paragraph it's already forgotten what you asked for in the first.
This happens because the AI generates the entire answer in one pass. If the task has many parts, it dilutes its attention across them.
The Ladder Rule (in 30 seconds)
Think of your task as a ladder. Instead of jumping from the floor to the top, you climb one rung at a time. Each rung is a separate prompt.
Don't do this: "Create a complete content strategy for the next quarter."
Do this:
- First: "List the 3 main themes we should cover this quarter."
- Then: "For theme 1, suggest 5 article titles."
- Then: "Develop article 1 with an intro, 3 key points, and a conclusion."
Real example: product launch plan
Rung 1: "List the 5 most common risks when launching a B2B app."
Rung 2: "For each risk, suggest a mitigation action."
Rung 3: "Build a 4-week timeline covering all 5 risks and actions."
Rung 4: "Turn the timeline into a 5-slide presentation."
Four prompts, each one building on the previous. In the end, you have a complete plan — and every part got the AI's full attention.
The golden rule
If a prompt doesn't fit in three sentences, split it in two. Two precise prompts beat one confused one every time.
With practice, you'll learn the right size: not so small it becomes micromanagement, not so large the AI gets lost.
The Before/After Method
The mental model that makes AI outputs reproducible.
Read articleWant the complete method with a curated set of reusable examples?
Explore the articles