promptengineering101

Meet Your Sous Chef: What is an LLM?

• Large Language Model
• A predictive text assistant trained on vast data - like a sous chef trained in every recipe.
• You give it a prompt (your recipe) and it prepares a relevant, customized response (your dish).

LLM Strengths: What Your Sous Chef Does Well

• Language Generation: Can whip up clear, fluent responses across formats
• Pattern recognition: Synthesizes information like combining ingredients into something new
• Rapid Response: Instantly brainstorms, drafts, or ideates on cue
• Flexible use cases: Can support any station like marketing or legal or education.

LLM Weaknesses: Limitations of Your Sous Chef

• Hallucinations: May invent facts or substitute ingredients incorrectly
• No True Understanding: Predicts words but doesn't grasp meaning
• Limited Memory: Can forget earlier steps in a long recipe
• No Real-Time Awareness: Doesn't know what's in the fridge today unless you tell it
• Bias and Tone Drift: Can add seasoning you didn't ask for unless guided

Common Pitfalls - and How to Stay in Control

• Hallucinations: Ask for sources or verify manually
• Missing context: Add background information or define the audience
• Off-brand tone: Include tone/style guidance in prompt
• Bias or sensitive content: Review critcally before sharing externally
• Outdated knowledge: Pair with real-time sources or apply human judgement

5-Step Process for LLMs (Cooking Workflow)

• Input: Supply the ingredients - words, context, instructions in the prompt
• Tokenization: The sous chef (LLM) breaks your prompt into smaller chunks, like chopping up ingredients into usable parts
• Encoding: The AI organizations those chunks and prepares the kitchen station - assigning numerical meaning
• Processing: The LLM processes everything by drawing on training to assemble a cohesive response.
• Output: Finally, your sous chef presents the finished dish tailored to your recipe (prompt).

What is a Prompt?

• A prompt is the text or question you give to an AI model to guide its response.
  • It is like a set of instructions, a creative brief, or a recipe.
• Prompting is the recipe, and the LLM is your sous chef.
  • The better the recipe (prompt), the more effective the sous chef (LLM) can be in helping prepare the final dish (output).

The 4 C's of Great Recipes (Prompts)

• Clear: Precise language, avoid ambuguity
• Concise: Use just enough detail, don't overexplain
• Commanding: Use imperatives like "Explain", "List", "Compare"
• Contextual: Provide relevant background as needed

Prompt Template

• As a role, your mission is to task.
• Deliver this as a format, writing in the style of tone.
• Focus on achieving objective, but don't restriction.
• Additionally, aim to incorporate desired skill/technique.

Yes Chef: From Recipe to Result

• Prompt Recipe -> LLM Sous Chef --> Output Dish.
• Summarize this product in one sentence....
• Summarize this product for a blurb on the product launch webpage, include a one sentence tagline, product description, and mockup a blog post too.
  • This one needs a lot of structure, steps, and context.

Menu of Dishes: What Your Prompt Can Cook Up

• Light Appetizer: Summarize with quick bite-sized insights
• Side Dish: Compare and contrast to enhance understanding
• Hearty Starter: Explain and Analyze to warm up thinking
• Main Course: Strategy and Planning with rich layered decisions
• Chef's Special: Action-Oriented Practical Execution
• Dessert: Creative and Storytelling that is light, fun, or emotionally engaging

Plating the Dish: Formatting Your Prompt

• Simple Lists
• Markdown Tables
• Freeform Text
• JavaScript Object Notation (JSON)