Video summary

Prompt engineering basics | هندسة الأوامر

Main summary

Key takeaways

Educational

Main Ideas & Concepts

  • Prompt/Program Engineering as “the key” to AI results

    • Strong outputs depend on how well you craft inputs/requests (commands).
    • AI does not “know your mind” without your wording—so word choice improves results.

  • A 4-axis approach to working effectively with AI models (high-level criteria)

    1. Focus on strategic outcomes
      • Start from the goal/result you want, not vague requests.
      • Example pattern: specify the format and components of what you want (e.g., a lesson plan outline with approved sections).
    2. Clarity and effective connection
      • Instructions should be clear, specific, concise, with no ambiguity.
    3. Systemic inputs
      • Use known frameworks or structured approaches to build requests.
    4. Know the model’s capabilities/limitations
      • Compare models (the video mentions several) and choose the right one for the task.
      • Don’t ask a model for something it can’t do (e.g., requesting an image from a model that doesn’t generate images).

  • Why prompt engineering matters (practical benefits)

    • Leads to accurate, unambiguous, complete outputs.
    • Avoids wasting time repeatedly fixing weak prompts.
    • Structured prompts help you reach goals faster because fewer revisions are needed.

  • The “driver seat” mindset

    • You define tone, format, depth, and output structure.
    • The objective is to prevent AI from “driving” and you being stuck “patching holes.”
    • You still must review/validate outputs, especially for critical or educational use.

Methodology / Instruction-Like Guidance

A) How to Design Prompts (core structure with key elements)

The speaker presents a design-focused prompt structure using five key elements, with a later addition of a 6th.

  1. Define the role

    • Assign the AI a specific persona/expertise relevant to your needs.
    • Include additional constraints (e.g., years of experience, ministry/region, relevant context).
    • If you want results for another country, change the geographical scope accordingly.

  2. Provide context (described as major—~“80%”)

    • Include what surrounds the task:
      • environment, country, curriculum setting, audience (students/educators), community culture, etc.
    • You cannot fully separate the AI from your reality; you must embed your reality into the prompt.

  3. Set tone

    • Specify style/emotional quality and how the AI should write/respond.
    • Emotional phrases can act as AI “triggers” (learned patterns from training data), even though the AI has no feelings.
    • Goal: guide the AI toward the desired urgency/attention level.

  4. Mental frame

    • Specify the mindset the AI should operate in (e.g., educational vs business mindset).
    • Use wording that signals the desired approach, since phrasing can change output behavior.

  5. Output format

    • Explicitly state what the output should look like:
      • article, paragraph, table, bullet list, email, image, presentation (e.g., PowerPoint)
    • Specify expected depth (quick summary vs analytical).

  6. Add examples (the “6th point”)

    • Provide example artifacts so the AI mirrors your preferred style/quality.
    • Examples reduce the time spent describing everything from scratch.
    • Examples include:
      • uploading an approved lesson plan model
      • using reference visuals (e.g., design inspiration from sites like Pinterest/Google)
    • Techniques include:
      • upload a design and ask AI to place your content into that design
      • reverse engineer an image (turning/copying the structure/style back into a usable template/instruction)

B) Quality Assurance & Safety Rules (explicit ethics/instructions)

  • Always review AI outputs

    • Even small errors (misspellings, wrong diacritics, incorrect content) can be disastrous in education.
    • Don’t treat AI output as final without verification.

  • Verify Quranic material carefully

    • If the AI includes Quran verses, check them carefully.
    • Prefer providing verses with correct diacritics from reliable sources rather than trusting the AI to produce them accurately.

C) Using a Framework to Structure Prompts: “RHO…DES” / “Rhodes”

The speaker introduces a reusable framework with components:

  • R = Role
  • H = Objective/Goal
  • O = Objective/What you need from the AI
  • Details (parameters/requirements)
  • Examples
  • Sense check
    • Require the AI to confirm understanding before proceeding.
    • If unclear, the AI should ask 2–3 questions about missing/unclear parts.

Key emphasis:

  • The order of components is not mandatory, but the content must be present.
  • Use “Sense check” to prevent the AI from starting the task incorrectly.

D) “Teach-back” / Step-by-Step Mode During AI Work

During AI work, instruct the AI to:

  • break tasks into steps,
  • show how it is doing each step,
  • explain its reasoning flow (described as a “thinking” metaphor), so you can analyze and correct the process.

Examples Mentioned in the Video (Prompt Use-Cases)

  • Lesson plan / learning plan creation

    • Specify role (e.g., expert Arabic teacher), context (e.g., Egypt ministry context), tone, output format, etc.
    • Use model lesson plans as examples/reference.

  • Social media promo creation using the Rhodes framework

    • Use an image and ask the AI to create a complete promo using the framework (not partial pieces).
    • Include role/persona + goal + details + example + sense check.

  • Summer plan for children

    • Requires a diagnostic interview via questions first, then plan generation.
    • Includes constraints such as target age group, country-specific resources, budget, and activities.

  • Converting an idea into a prompt using a dedicated “Prompt Engineer” GPT

    • Example idea: a 60-second cinematic ad for an Arabic course for non-native speakers.
    • The prompt includes:
      • role (ad director/scriptwriter)
      • goal (ad output)
      • audience demographics
      • style (modern/contemporary)
      • target learner profile (beginners; travel/culture/religious/work motivations)
      • optionally tools for video generation (e.g., Runway)

Tooling / Platform References

  • Model/platform examples

    • ChatGPT variants and model names like ChatGPT/Charge BT, Gemini, Claude/Cloud (subtitles were inconsistent, but the intent is comparing capability sets).

  • GPT/program references

    • A Custom GPT called “Prompt Engineer”
    • Another GPT/program labeled “Program Engineer” (and video tool examples like Runway / others)

  • Video-generation tool references

    • Runway (and other tools with unclear subtitle names)

Speakers / Sources Featured

  • Primary speaker (unnamed)

    • Instructor/host presenting “prompt engineering basics” and teaching frameworks (subtitles reference names like Ms. Iman, Ms. Asmaa, Ms. Ghada, and Dr. Tamer/Dr. as participants or addresses).

  • Participants mentioned by name

    • Ms. Iman
    • Ms. Asmaa
    • Ms. Ghada
    • Hassan
    • Mr. Khaled
    • Ms. Hanan
    • Dr. Tamer
    • Dr. (another doctor reference)

  • Other source/provider mentioned

    • EduCareers

  • Tools/GPT sources (referenced)

    • GPT named “Prompt Engineer”
    • GPT/program named “Program Engineer”
    • Runway for video generation

Original video