Getting Started with Prompt Engineering

1. Introduction2: Hey, there. I'm Victor Cuevas. I'm the founder of the knowledge Star, and I'm the creator behind my digital dad, a YouTube channel where I help parents and kids learn technology together in fun and simple ways. I've taught thousands of learners of how to use AI and digital tools to solve real world problems, whether it's building automation, streamlining work, or even learning as a family. Now I'm here to help you master prompt engineering. This class Prompt Engineering Basics is designed for beginners, whether you're a content creator, teacher, student, business professional, or just AI curious. There are no prior experience required, and all you need is access to a conversational AI tool like Chat GBT. You'll learn how to understand the essential building blocks of a great prompt. You're going to craft clear and detailed instructions to get consistent, high quality AI responses. You'll experiment, refine, and perfect your prompts through iteration. The value of this class is simple. You're going to walk away with a powerful, reusable skill that saves you time, boost your creativity, and helps you get better results from AI. Whether you're writing emails brainstorming, creating lesson plans or generating content. Your class project will be to create a personal prompt library starter kit. Now, you're going to choose your own use case, write three to five strong well structured prompts, using the methods I teach and test and refine them. Share your kit with your friends and family so they can learn from your creativity. I'm so excited to have you here. This is going to be fun and practical. Let's get started. Seeing the first lesson. 2. AI Explained MMC: In this lesson, we're going to cover artificial intelligence, also known as AI. Now, what the heck is AI? My simple explanation, AI is a system performing tasks typically done by people. The first point in this lesson, AI is an umbrella term. Machine learning, generative AI, deep learning, and many more all fall under the AI umbrella. I'm going to leverage Aspen Digital's AI graphic to help break down the components. I will also leave a link to the PDF for you to download. Let's start with the first zone. The first zone of AI is known as automated reasoning. AI automated reasoning is a process where computers use logic and algorithms to draw conclusions and solve problems without human intervention. It's almost like a lot of if then statements. Now, let's use Pac Man as an example. There are AI ghosts and they use their superior reasoning skills to deduce optimal path to ruin our Pac Man's day. They don't need a map or some type of GPS. They just use simple rules and a serious grudge against yellow circles. The next zone of AI is known as machine learning. Machine learning is a data driven system that learns and improves from experience. Now, if you use Amazon or Netflix, then you've interacted with machine learning tools. These platforms analyze your browsing and purchasing history to recommend products, movies, or shows you're likely to be interested in. Machine learning has sub levels. First I want to point out is supervised learning. Now, in this instance, the AI model is trained on labeled datasets to answer the question, what is that? Based on characteristics or features, there are various techniques within supervised starting. One technique is classification. Now, if you have a credit card, you may have experienced that transaction is not approved. You call the credit card company and find out the system identified the transaction as potentially fraudulent. Although annoying at times, it's intended to protect you. This is an example of how AI classifies transactions as fraudulent or legitimate based on the learning patterns. We also have unsupervised learning. The AI model is trained on data that is not labeled. Answering the question, what's similar? By finding patterns, structures, or relationships within the data on its own. Clustering is one technique used in unsupervised learning. Think of Nike. Nike is an example using clustering to segment product offerings based on athletes, fitness buffs, and your Bastanistas. The next area is reinforcement learning. Reinforcement learning is based on the model taking actions and receiving feedback. Good and bad. Taking a trial and error approach. Now, if you use Siri, then you're using an example of AI that's used as reinforcement learning. Virtual assistants such as Siri uses reinforcement learning to improve their conversations. The last area is generative AI, which is quite simply creating new content such as text, images, videos, or music. When using tools such as hat GPT, you're using generative AI. Chat GPT can generate text and images based on your text input. One additional key term in the generative AI space to be aware of is multimodel AI. This is the ability of the AI model to generate multiple types of data, such as text images, video, et cetera, in one platform. I personally think this is a crown jewel for any company in the AI space. As you hear about conversation and AI, or if you're thinking about implementing an AI solution, first figure out which area of the AI map we're talking about. I'll see you in the next lesson. 3. What Are LLMs MMC: In this lesson, we're going to cover large language models also known as LLMs. Before we kick off LLM topic, we need to start with foundation models. In the early years, AI was trained for specific tasks, limiting its range of functionality. However, foundational models change that. Foundational models are also known as base models. These models are trained on a vast quantity of data at scale to be used for a variety of tasks. Now, how can these large language models perform these magical tasks? They go through two phases, pre training and fine tuning. In the pre training phase, the model is generally trained on data that can be found on. Anything from web scraping, books, transcripts or anything else that is text based. Now, in the fine tuning phase, the models become better at specific tasks through techniques such as reinforcement learning, some use cases, a chat bot that can help you with your homework or answer questions about a topic you're interested in. Think of a self driving car that can take you safely from one place to another, a system that can recognize pictures of animals and tell you what kind of animal it is. This brings us to large language models, otherwise known as LLMs. Although both foundational models and large language models are both AI models, they differ in scope. An LLM is a top of model focused on language related tasks. Now, there are two key concepts that we need to discuss regarding LMs. One is language models and natural language processing, also known as NLP. So first, what is a language model? A language model refers to a type of model specifically designed to generate human like text or predict the probability of a sequence of words. Language models learn patterns and statistics from large amounts of text data, enabling them to generate responses that make sense to us. An example, we can start with the word cat. The model will predict the next sequence of words. Here, we get the cat sat on the floor. You see, it's not magic. It's just math. The next concept is natural language processing. This is how LLMs are able to perform language related tasks at scale. Natural language processing is defined as the branch of AI that provides computers with the capability of understanding texts and spoken words in the same way a human being can. Some examples of how LMS are used today. Chat GPT by open AI. A conversation AI chatbot that can answer questions, generate code, all from text. Another use case is documentation, translation and summarization. Today, you can upload a PDF to Chat chPT or Cloud AI, for example, to get key summaries and even translate it into different languages. Well, that covers it for LLMs. I'll see you in the next lesson. 4. LLM Creativity MMC: Now in this lesson, we're going to learn how to control creativity by understanding temperature. LLMs are designed to predict the next word in a sequence, and guess what? So do you and I. How would you finish a sentence? The sky is. Which word you choose given the following options? Blue, the limit or bananas. Well, pretty sure you would have chosen blue or the limit, but not bananas. Guess what? For LLMs, bananas is a feasible option. I know. Isn't that just bananas? LMs store probability for every word that could possibly follow the sky is. Can adjust the probability of the predicted word with a setting called temperature. Now, temperature controls how random the output of a large language model is. Chat IPT, for example, has a range 0-2. The lower the temperature, closer to zero, the more focused and predictable the output. The higher the temperature, closer to two, the more creative and random the output. This is just a control knob to be aware of when providing prompts in a chat, since it'll influence the output. Let's test out an example in open AI's playground. First, let's set the temperature near to two. Now I can ask who's Tony Robbins? Starts off strong, then completely shanks the rest of the response. At maximum temperatures, most models are incomprehensible. Now let's try the other extreme and set the temperature near to zero. Bingo. The model chose the most likely output every time. When you set the temperature to zero, the model eliminates a lot of the randomness from the predictions. We call this type of model deterministic. A deterministic model produces the same output for each given input. Now let's go through some scenarios. You're drafting a resume for a job you're applying for. What temperature would you set the model to? Low temperature. Now, if you wanted to write poetry, what temperature would you choose? Medium or high, either one. The key here, depending on the desired output, your temperature is a control knob that can influence results. I'll see you the next lesson. 5. Improving Models MMC V3: This lesson, we're going to cover how LLMs get good at what they do. Now, often you hear we've trained the model on a ton of data. Here's a question for you. Do you think it's sufficient for the model to be trained on the data just once? Well, it actually requires multiple iterations, and this brings us to a term called epoch. Each pass through the train data is called an epoch. Why is this required? Now, recall, LMs are neural networks. Before they're trained, predictions stored in the model are random. So even after one epoch, the LLM cannot predict words correctly. So how can a neural network of an LM approve over each training session? During each epoch, the neural network compares its prediction with the original data. Prediction is usually off by some amount, which is called the loss. The loss is a difference between the predicted and actual values. Now, like a rocky movie, after each epoch, our model becomes bigger, better, faster and stronger. Now, you think we should continue to train like Arnold Schwarzenegger, to the point, there is zero fat on the body or in this case, zero loss. Now, the answer is no. The loss is zero. The LMs predictions fit the training data exactly. Which means it can't generate new and exciting things. This is known as overfitting. The LLM is overfit when it replicates patterns in the training data to the point that it cannot generate new data or generate new patterns. The key to prevent overfitting is to monitor the model's performance. It's not a one size fit all solution. At the end of the day, training epoch is just one factor for LM improvement. More to come in the next lesson. See you soon. 6. Preprocessing Tokenization MMC V4: This lesson, we're going to cover pre processing and tokenization. Now, can the English language be messy? Absolutely. There are typos, abbreviation, inconsistent capitalization, multiple spellings, list goes on. So think of this sentence. I eight Grandma. One comma changes the entire context of the sentence. Computers depend on consistency. So to make text readable, we need to clean it up. The process of turning raw text into a clean dataset has two stages preprocessing and tokenization. Now, both of these steps help standardize text in a corpus before a models trained on it. What's a corpus? Well, a corpus is a collection of written texts that a language model is trained on, such as all of Elam Ma's tweets. The first part of the process is preprocessing. Now, pre processing is the process of cleaning, transforming and organizing raw data to make it suitable for AI models. This involves tasks such as clean up missing values in the data or removing outliers. This is extremely important to ensure AI models learn meaning patterns to make accurate predictions. As I say, bad data in is bad data out. The next step is tokenization. Tokenization is a process of breaking down text into smaller units called tokens, such as words or characters to make it easier for computers to understand and analyze language. Here's an example. I played basketball yesterday, period. How many tokens do you see? Now, considering stems, affixes, and punctuation, we have six tokens. Another method is using algorithms to build tokens from characters. An example of this is Byte pair encoding. Byte pair encoding is a tokenization algorithm that builds tokens from characters. Here's a quick example. Imagine you have a big block of legos and you want to build a castle. Now, you notice that you keep using the same pairs of legos together over and over again. Now, instead of picking up each piece one every time, you decide to create special blocks that combine the frequently used pairs to make building faster and easier. Now, by creating these special blocks, you make the building process more efficient. And just like a byte pair encoding makes text processing more efficient by combining frequently used pairs of characters or words. So tokenization essentially for building a vocabulary that can be used to train in LLM. In practice, LLMs are trained hundreds of billions for tokens. See you in the next lesson. 7. What is Prompt Eng MMC: This lesson, we're going to cover the following. What is prompt engineering? And why is it important? What is prompt engineering? In general, it's just an input to an AI model to guide an output. The prompt could be a string of text, as you've seen in chat GPT. The prom can be other forms of media such as images or audio. Let's take a look at example using Anthropi chat bot Cloud AI. You can use a free version or the pro version, which is about $20 per month as of today. Just know, the paid version lets you use the premium features more often than the free tier. Now, in the prompt window of Cloud AI, I can ask the following question. Explain what is generative AI in a funny tone in 400 characters. We get a funny response. Now, if I wanted to take this idea one step further and create a post in Twitter, which is limited to 280 characters, I can. Just need to tell Claude to rewrite in 280 characters. Now, let's take it one more level higher. Say I wanted the output framed in the following format, stages of whatever the topic is beginner, beginner vibe, intermediate intermediate vibe, advanced, advanced vibe, and master Master vibe. I can just ask Claude to rewrite and use this format. You can see here, the process of prompt engineering, as of today, is taking a prompt and testing it out, refining it until we get an effective output. Do I see prompt engineering changing over time? Absolutely. In the near future, I think it'll be conversational speech. But for now, we're sticking to the basics. Now, there are many ways to improve the prompt, such as applying webpage development techniques such as HMM markups. Now, why is prompt engineering important today? By writing good prompts, AI models can be guided to create relevant information. Recently had a gentleman tell me about a software application and small business idea that was created based on code provided by CHAT GBT. To me, it is simply amazing. In the next video, we're going to cover what is a good prompt. I'll see you soon. 8. What Makes A Good Prompt MMC: Now in this lesson, we're going to cover what makes a good prompt for genitive AI. An acronym for you to remember, I craft, examples. This encapsulates seven key components of a good prompt. And what are these seven key components? Instruction. Context, role, arrangement, formatting, tone, and finally, examples. Let's walk through each of these individually. First, let's cover instruction. Here, we clearly state what we want the model to do. Here's one example. I'm going to instruct Chat GPT to tell me about the benefits of being a Jedi Master. The second component is context. We need to provide background information or details to help the model understand the situation. Now using our framework, let's now try to create a video script for Mr. Beast. A quick note, I mentioned HTML markdowns. I keep my framework organized by using markdowns such as the number symbol. Here's an example, instruction. Generate a video script idea for Mr. Beast's YouTube video that involves a large scale charity event. The contexts create a video idea that combines Mr. Bea's signature style of grand gestures and heartwarming moments with a charitable cause addressing current social issues or needs. I think I have a chance for Mr. Bees. The third component is role. Here, we give the model an identity to assume while responding. The role can be completely different, and if we tell our model to act as a teacher rather than as Mike Tyson. Now, normally, we want to start with act as whatever. One example, we can ask our chatbot, in this case, ChatBT to help us lose weight by creating a morning routine based on a role played by Tony Robbins. So our instruction will be write a morning routine. The context is 30 pounds overweight and I need to lose weight. And the role is act as Tony Robbins. The next component is arrangement. This is how we want the information presented. We may want our information structured with an introduction, three main points, and the conclusion. Here's an example. Write a story about how Humpty Dumpty beat up the Big Bad Wolf with three little pigs story. The context, Humpty Dumpty has been working out with Mr. Arnold Schwarzenegger, and it's built six foot five and strong. The role act as Morgan Freeman for narration. The enragement should have three paragraphs introduction, middle and end. I know this is a little bit random. Let's see what we get. We now have story time for the kids. Now we can add formatting. This is the desired format for the output. Here we can change the structure of the text. Let's take the previous example and ask Chat GPT to maybe add a table of the characters in the story. Using the same instructions, I'm going to add format, make a table of the characters in the story. And as you see, we get our table of characters. We're now ready for tone. Here we have the opportunity to determine the tone or style of the response, such as being happy, silly, angry, whatever. I'm going to make this really brief. So instruction, give me a brief history of Michael Jordan in two paragraphs, but my tone, I want my story sarcastic and humorous. The style of the output has changed sarcastic and humors. We're ready for our last component, which is examples. This is a technique to guide the model on the type of response you're looking for. What we're going to do is we're going to use another example, bringing in all the components, including examples. So first, the instruction, write a joke about the rise of AI in everyday life. Context. In 2024, AI has become an integral part of daily routines from smart home devices to AI driven customer service. People are increasingly relying on AI for various tasks leading to humorous situations. The role, we're going to act as a stand up comedian, the arrangement, present the joke with a clear setup and punch line. Format, it's going to be brief and concise, ensuring the joke is easy to read and understand. Tone is light hearted and humorous, and we're going to give it an example, such as, why did AI go to therapy Because it had too many unresolved loops. I hit Enter, and we get our headliner joke, just kidding. What you now have is a framework for understanding good prompt structures. Onto the next lesson, I'll talk to you soon. 9. Prompting Strategies MMC: In this lesson, we're going to cover a few AI prompting strategies that will help AI do harder jobs and give smarter answers. The three techniques that we're going to cover zero shot chain of thought prompting, few shot prompting, and few shot chain of thought prompting. First, let's cover zero shot chain of thought prompting. Zero shot chain of thought prompting involves asking a model to solve a problem by explicitly instructing it to think through the steps without providing any examples. Technique is useful for complex reasoning tasks where you want the model to explain its thoughts process. Let's check out an example. Now imagine you want to solve a problem like what is 15 plus 27. Now, instead of just asking for the answer, you prompt the model to think step by step. In hachBT, I'm going to use this prompt. Let's think step by step and what is 15 plus 27? You can see the model walks us through step by step to get the final answer. Now, let's discuss f shot prompting. View shot prompting involves providing the model with few examples of the task you wanted to perform. Now, this helps a model understand the pattern and generate the desired output. Let's cover another example. Suppose you want the model to translate English sentences to French. You provide a few examples to guide it. So in my prompt, I'm going to answer the following. I translate the following sentence to French and first is Hello, how are you and have the French equivalent. I love to read books and the French equivalent. The weather is nice today and the French equivalent. So, finally, I'm going to enter, translate this sentence to French. I am learning to code. We get our answer. I'm learning to code in French. Now, the third technique is few shot chain of thought prompting. Few shot chain of thought prompting combines the principles of few shot prompting and chain of thought prompting. You provide a few examples where each example includes a detailed reasoning process. Now, this helps the model understand not just the task, but also the reasoning steps involved. Let's walk through an example. Let's say you want the model to determine if the sum of odd numbers in a list is even or odd. You provide the examples with reasoning. So here, I entered the odd numbers in this group add up to an even number. I have a list of numbers. And adding all these numbers is the answer gives 25. The answer is false. And I do the same thing with a few other examples. So finally, my question is, the odd numbers in this group add up to an even number? And then we get our answer. So in summary, zero shot chain of thought prompting instructs the model to think step by step without examples. Few shot prompting provides a few examples to guide the model on the task. And few shot chain of thought prompting combines few shot examples with detailed reasoning steps to guide the model. By using these techniques, you can improve the model's ability to perform complex tasks and generate more accurate and reasonable responses. See you the next lesson. 10. Advanced Prompting Strategies MMC: Now in this lesson, we're going to cover a few additional AI prompting techniques used in prompt engineering to improve interactions with LMs. The three techniques we're covering is generative knowledge prompting, least to most prompting and emotional prompting. So first, we're going to cover genetiveKledge prompting. GenetalKledge prompting involves using AI to generate relevant knowledge statements that can help solve a specific task. Now, this is useful when we want to get a thoughtful response, since we're asking the model to create potentially useful information about a given question before generating a final response. Let's check out an example. Imagine you want to create an article about climate change. Use generative knowledge prompting to gather key inputs. In Chat GBT, I'm going to use this prompt. Generate some key points about the impact of climate change. Now we got some key points. I can add a follow up question. Using the generated key points, write an article about the impact of climate change. AI uses the generated points to create a comprehensive article. Now let's discuss east to most prompting. Least to most prompting is a hierarchy approach where prompts are given an increasing level of assistance. Now, the idea is to start with the least intrusive prompt and gradually increase the level of help until the desired response is achieved. Now, this is useful because it allows AI to break down a problem into subproblems and then solve each one individually. Let's cover an example. Suppose you want the model to break down the steps to teach a child to use a spoon. Now, in my prompt, I'm going to enter the following. I need to teach my child how to use a spoon. Don't solve the problem, but break it down to subproblems. The model breaks down the task to smaller subgroups and provides guidance on how to tackle each task. Now, the third technique is emotional prompting. Emotional prompting involves adding emotion cues to prompts to enhance the performance of AI models. These cues can make the task seem more important or urgent, potentially leading to better responses. Let's walk through an example. Now, imagine we need help with our resume. A non emotional prompt. Can you help me write a resume? Now let's add some motion to it. This is very important to my career. Can you help me write a resume? Get a more polish response. So in summary, general knowledge prompting generates relevant knowledge to inform task completion, making it easier for users to gather and use information. Least to most prompting uses a hierarchy of prompts from least to most intrusive to guide behavior, promoting gradual learning and independence. Emotional prompting adds emotional cues to prompts to enhance the quality responses, making interaction more engaging and effective. Now, by understanding and applying these strategies, users can effectively guide AI models to produce more accurate and relevant outputs, simplifying their interactions with the tech. See you in the next lesson. M 11. AI Hallucinations MMC: In this lesson, we're going to cover one of the challenges in AI, which you have heard about in the news AI hallucinations. So what are AI hallucinations? These are incorrect outputs generated by AI models. The AI model can sometimes say things that sound true but are in fact inaccurate. This happens because they don't really understand the world like humans do. They just combine information in ways that often work, but can lead to mistakes. What are the causes of AI hallucinations? Now, there are many factors and cause A hallucinations, but here are a few. Training data limitations. Although huge models such as ChatBT that's been trained on the Internet, the information that it was trained on may have been incomplete or inaccurate, leading to generate incorrect information, such as Google's AI overview. AI model biases. If the training data contains biases, the model might reflect these biases in its output such as Google's gem generation of racial diverse Nazis. Complexity of language. The human language is highly context dependent. Sometimes a model misinterprets the context or fails to understand the subtleties resulting in hallucinations. In some cases such as in 2023, when Microsoft Bing AI produced some creepy conversation with its users. How do we combat hallucinations? One method is using self consistency prompting. Self consisting prompting involves generating multiple responses to the same prompt, analyzing them for consistency and selecting the most coherent answer as a final output. Let's check out an example. In Chachi PT, we can ask when I was six, my sister was half my age. Now I'm 70. How old is my sister? Let's think step by step. Let's grain another answer for the same question. Now, let's do it one more time. All three lead to the same conclusion. Which builds trust in the answer. Now, another method is role prompting. Role prompting is a technique used in prompt engineering to guide an AI models response by assign it to a specific role or character to embody. Let's check out an example. We're going to use the prompt. You are a friendly kindergarten teacher. Explain what photosynthesis is to a 5-year-old child. Now, here the role is kindergarten teacher. The context is friendly and explain to a young child. And the task is to explain photosynthesis. Response not only leads to a more accurate response, but also to a more tailored and appropriate response based on the audience. Now, there's another technique called retrieval augmented generation, also known as RAG. This allows the AM model to access information from external knowledge sources as context for the prompt. Now, normally to implement RAG into a workflow, there are technical requirements. We can initiate a simpler version of this concept by uploading documentation into the chat bot. Let's take an example. Here in scri.com, I can get a PDF version of books. Now say I want to use the four agreements as the example. Download the document. You can head over to PerplexEAI and upload the document. Now I can ask for key insights related to the four agreements book. One more technique is self evaluation prompting. Self evaluation prompting involves asking an individual or an AI system to reflect on and evaluate their own work, responses, or capabilities. Let's check out an example. Starting from Chat GBT, I provide the following prompt. Explain the concept of photosynthesis. Next, you add a self evaluation prop, evaluate the accuracy and completeness of the previous explanation of photosynthesis. Now we can add a follow up prom. Based on your evaluation, revise the explanation of photosynthesis. Now we have a complete response. Now you can see the challenges with hallucinations and how to combat them. It's always important to fact check AI outputs since hallucinations may arise. I'll see you in the next lesson. 12. AI Risks Ethical Concerns MMC V2: On this lesson, we're going to highlight some of the risks and ethical concerns with genitive AI. Genitive AI, while powerful and innovative, comes with significant risks, particularly in amplifying biases and spreading misinformation through AI hallucinations. Now, let's start with amplified biases. GenertI models are trained on vast datasets from the Internet, which inherently contains biases. These biases can be related to race, gender, ethnicity, and more. An example in 2023, the stable diffusion model depicted professionals such as doctors and engineers, predominantly as white males, while nurses were depicted as white females. Misrepretation does not align with the actual diversity in these professions. Next is AI hallucinations. A hallucinations occur when genera models produce information that appears plausible but entirely fabricated. These hallucinations can lead to the spread of misinformation. Can have serious consequences. In the case of Mata versus Abaca, a lawyer used chat GPT for legal research, and the AI generated false citations and quotes, leading to significant legal repercussions. Some additional dangers include copyright issues and sensitive data. On the copyright side, there's significant debate over who owns the copyright for AI generated content. The AI itself, the developers or the users. As recent as June 2024, some of the world's biggest music labels, Sony, Universal, and Winner Records, filed a lawsuit over AI copyright infringement against Sno and UDO. While, AI companies have previously argued that their use of copyright material falls under fair use, the record labels contend that Sno and UDO are profiting from replicating songs without transformative purpose. This sets the stage for the critical legal debate over the boundaries of fair use in the context of AI generated content. There is also the risk of sensitive data leaks and privacy laws. Users inputting sensitive information into public AM models poses a significant risk of exposing confidential data. Let's take it one step further. AI models may inadvertently generate or reveal personal information potentially violating privacy laws like GDPR. In 12023 example, chat GBT users reported instances of data leakage, including exposure of personal data, conversations with a chatbot, and login credentials. Some users found they could access details of other user proposals, presentations, and conversations. There's not a simple technical solution to mitigate all these risks, but I like a phrase coined by the AI Exchange founder, Rachel Woods. If you will not put on read it, do not put in the AI chat bot. As an individual, you just need to know when and where to use AI tools. Although there are tons of advantages to use these tools, it's important to understand some of the risks. See you in the next lesson. 13. Conclusion2: Wow. You made it. I just want to say a huge thank you for spending your time with me in this prompt engineering based class. Now, I'm Victor Cuevas, and I'm so happy I was your guide on this journey into the world of prompt crafting. You've now got the tools to design clear structure prompts, experiment and refine your results, build your own prompt library to save time and spark creativity in your work or personal projects. So whether you're using prompts for brainstorming, content creation, lesson planning or productivity. You now have a skill that will help you grow in value as AI continues to evolve. If you enjoy this class, I'll be grateful if you leave a good review and follow me here on Skillshare for future classes. I've got more exciting content coming your way to help you and your family thrive in this digital world. Thanks again for learning with me. You crushed it. Now go make some AI magic, and I'll see you in the next class.