Ourbit Master AI — 15-Day Programme

Ourbit Master AI — Day 3: Prompting as Instruction Design

Download lesson PDF

Day 3 — Prompting as Instruction Design

Purpose: Use clarity, structure, audience, format and constraints to reduce avoidable guesswork.

Source/build status: Recovered Module 2 is the source. Preserve its dog-comparison example.

Learning outcomes

• Turn a vague request into a clear instruction with an explicit task, audience and purpose.
• Use structure and constraints to reduce avoidable model guesswork.
• Identify missing facts that should be exposed for confirmation rather than invented.
• Separate prompt quality from factual verification: a better instruction can improve usefulness without guaranteeing truth.

Core lesson — prompting as instruction design

A prompt is an instruction package that reduces avoidable ambiguity. Three elements do most of the work: clarity about the task, structure for the expected output, and constraints that define useful boundaries. Audience, purpose and relevant context strengthen all three.

Compare “Write about dogs” with “Write a 100-word comparison of the exercise needs of a Border Collie and a Bulldog for a first-time dog owner.” The second request does not guarantee truth; it removes guesses about scope, audience, length and purpose. Good prompting improves the setup. Verification remains a separate responsibility.

Worked example — diagnose before rewriting

A manager asks: “Write an email about the delay.” Before generating, identify what is missing: who will receive it, what delay is confirmed, what action is required, what tone is appropriate and what must not be invented. A stronger instruction supplies known facts and asks the model to mark missing information for confirmation rather than fabricate a reason.

Hands-on practice

1. Choose a vague request you might genuinely use.
2. Mark everything the model would have to guess.
3. Rewrite it with a clear task, audience and purpose.
4. Add the required format and useful constraints.
5. Compare both outputs for specificity, usefulness and unsupported assumptions.
6. Revise once more based on the weakness you actually observe.

Knowledge check

1. What roles do clarity, structure and constraints play?
2. Why can a specific prompt improve usefulness without guaranteeing accuracy?
3. What should a model do when a required fact is missing?
4. Rewrite “Tell me about marketing” for a defined audience and purpose.
5. If an answer is too broad, which instruction elements should you inspect first?
6. When should uncertainty be exposed rather than filled with a guess?

Teaching / video script — 6–8 minute target

Open: Put “Write about dogs” on screen. Ask what the model must guess: audience, purpose, length, format and angle. Then reveal the structured Border Collie/Bulldog request.

Teach: Build the improved instruction visibly in three passes—clarity, structure, constraints. Show that prompting reduces unnecessary guesswork; it does not turn generated claims into verified facts. Then use the delayed-email example to demonstrate how to mark missing facts rather than invent them.

Close: A stronger first instruction helps, but real work rarely ends with version one. Day 4 shows how to diagnose weak output, iterate deliberately and break complex tasks into verifiable steps.

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.

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.