Ourbit Master AI — 15-Day Programme

Ourbit Master AI — Day 4: Prompt Iteration and Task Decomposition

Download lesson PDF

Day 4 — Prompt Iteration and Task Decomposition

Purpose: Improve weak outputs deliberately and split complex work into verifiable steps.

Source/build status: New gap extending Day 3.

Learning outcomes

• Diagnose a weak AI output against explicit success criteria instead of repeatedly adding instructions.
• Revise one meaningful variable at a time and explain what changed.
• Decompose a complex task into stages with checkpoints before dependent work continues.
• Recognize when a failed upstream assumption requires rollback rather than cosmetic rewriting.

Core lesson — Prompt Iteration and Task Decomposition

A weak first output is diagnostic information, not a reason to keep adding random instructions. Compare the result with the intended outcome, identify the largest gap, change the instruction or evidence deliberately, then test again. For complex work, decompose the task into stages whose outputs can be checked before they become inputs to later stages.

Worked example

A small course launch plan can be separated into audience and success criteria → offer assumptions → outline → draft assets → verify factual claims → review against criteria. If the audience definition is wrong, correct it before generating every downstream asset.

Teaching / video direction

Show one overloaded prompt becoming a five-stage workflow. Demonstrate one-variable-at-a-time iteration and visible checkpoints. Close by explaining that longer workflows create a new problem: keeping the right information available—Day 5.

Hands-on practical exercise

1. Take a multi-part task you actually understand and write its success criteria.
2. Ask AI for a first attempt and identify the single largest gap.
3. Change one instruction, input or constraint that addresses that gap; compare the result.
4. Split the task into at least four dependent stages and define what must be checked at each handoff.
5. Introduce one deliberate upstream error and identify which downstream work becomes unreliable.
6. Record the iteration history and explain why the final workflow is more dependable.

Knowledge check

1. Why is a weak first output useful diagnostic evidence?
2. Why can changing several prompt variables at once make improvement harder to understand?
3. What is task decomposition intended to achieve?
4. When should downstream work stop because an upstream stage failed?
5. Give one example of a useful checkpoint between two dependent stages.
6. How does deliberate iteration extend Day 3 prompting?

Assessment and remediation contract

The scored daily quiz uses 10 questions; 7/10 (70%) is a pass. Missed concepts map to exact lesson anchors for targeted remediation, including on a passing attempt. Attempts 1–9 below threshold return to remediation and a fresh eligible question set; an unsuccessful tenth attempt restarts the day in a new learning cycle while preserving history. The 50-question validated bank is a learner-release gate and will be scaled only after the Day 1 runtime passes acceptance.

Teaching / video script — 6–8 minute target

Open: Start with the worked scenario, not a generic definition: Prepare a launch plan for a small online course.

Teach: Demonstrate the method visibly: First define audience and success criteria; then outline; then draft; then verify claims; then review against criteria. Pause at the judgment point and explain why the system should not simply continue on autopilot: Change one important variable at a time, inspect the result, and keep checkpoints between dependent steps.

Practice bridge: Ask the learner to reproduce the method on a different low-risk example, mark the verification/control step, and explain the reason for it.

Close: Day 5: manage context so important instructions remain available.

Production / QA notes

• Use the worked scenario as the primary visual sequence; avoid generic AI-brain imagery.
• Keep terminology vendor-neutral and define unavoidable technical terms in context.
• Do not equate fluent output with verified truth or AI assistance with autonomous authority.
• Before learner release, fact-check the final lesson and validate every scored item against explicit teaching.