How To Build An App With AI (no experience required)

1. Introduction: Hello, everyone. I hope you are doing well. Welcome to our new course to learn how to create applications without having to develop. For those who don't know me, I am FIRs. I have a PhD in information system specializing in artificial intelligence. Today, I will guide you in creating a project where we will ask artificial intelligence to develop for us, to literally create all the code for our application using simple prompts, which we will launch together. The course starts first with some theoretical concepts to learn how to create your prompt, and then we will begin a practical workshop where I will show you step by step. How to create your project and how to simply have a functional web or mobile application. The course is designed for beginners. By the end, you will be able to create your own application or even your website using the prompts we will use together. I thank you very much and see you soon in our next video. 2. Claude 3.7 Sonnet & Claude Code: Hello, everyone. I hope you are doing well. In this video, we are going to look at the technological innovation created by Enthropic. It's simply artificial intelligence with the Cloud 3.7 Sonnet model and cloud code. Now if I start talking about Cloud 3.7, you should know that quite simply. It is an artificial intelligence that has the ability to respond very quickly and also have a deep reflection for questions and exercises that are complex. It's as if you have a brain in fact that can react in speed mode and also in deep reflection mode. If you send it a basic question, it will respond instantly. It's as if you had an assistant. If you are going to ask it, well, or send it us in more complex tasks, you will see that it will actually solve them. It will first think in detail and then show you the steps of its reasoning. But that's not all. We also have Cloud code. This is simply a feature that makes Cloud powerful because from the start, Cloud has been strong coding, but now today with Cloud code, it has the ability for example, to fix bugs, to develop, create code, generate tests automatically, and even in fact, it is capable of creating mobile applications or even web applications that are very advanced, very complex. You will see that thanks to what we call the API, it will simply allow Cloud to connect with other agents or code editors because today it is delivered with the functionality there, we can launch it in the terminal, but we can go very far with the API. It is what will allow us to fully exploit this coding. Truly, this new duo between Cloud 3.7 Sont and Cloud code, it really provides extraordinary strength. You simply needed to create a free account on the cloud.ai website. It's enough. Even with a free account, no need to have a pro account, we can test Cloud 3.7. To have exploited, of course, the coding, it will be much more interesting to use the API. The app, in fact, is what will allow us to go very far in coding and programming. 3. Software Engineering Performance: And there you go. Here we have a graph, here we have what we call software engineering performance. We see several bars and each bar. In fact, it represents a language model and we will try to see accuracy dom in solving concrete programming problems. Obviously, the more the axis or the bar is high, the better the results. Here we have the highlight of 3.7 bell. We notice that it far exceeds the other models. Here it reaches 62.3% and when we use what we call scaffolding. These are basically additional models that assist the intelligence to reason and test actually without code. Here we see it reaches 70.3%. Obviously, it is much superior to the other models, and if I look, for example, at a comparison with the code cloud 3.5, already the Cloud 3.5 is not bad with 49%, but in comparison with the Cloud 3.7, there is really an evolution and we see that today, in fact, this system is the most efficient today in solving problems related to programming and encoding. And it's very important because the Cloud 3.7 is designed to think both quickly some and more deeply. Of course, according to the needs the type of exercise, the type of problem to code. It's a bit like having a single brain that is programmed or operates in instant response mode and deep reflection mode. In fact, is very, very important because the more difficult tasks we have, the more the Cloud 3.7 will gain in precision. This shows that the Cloud 3.7 Sony is today the leader in solving problems in software development. 4. TAU-bench Testing: So here in this graph, we have another test. This is a test where we see how the models, the different artificial intelligence models react when they handle complex tasks. This is a test we call the test tor Bench, where the tools call upon other tools to solve complex problems or scenarios. If we look here, obviously, the more the bar is high, the higher it is performing, it's more interesting and efficient for solving task by actually using the different tools. Here, Cloud 3.7, as you can see, it shows again that it handles complex scenarios better with more advanced reasoning and interaction capabilities. Clearly, we can conclude here that Cloud 3.7 actually surpasses the competitors. Here, if compared with open AI 01, also with the older version of Cloud 3.5, we see that it actually surpasses, especially in what we call realistic situations. Is it thinks it uses multiple tools. The objective is to arrive at a solution and that in fact simply confirms that this model is not only fast and accurate, but it's a system that is designed and adapted to solve practical cases that companies can encounter on a daily basis. 5. Claude 3.7 Sonnet (Extended Thinking): So here, this is a new board. In fact, here, we are going to see how Cloud 3.7 Sonet is a model today of the new generation. It is capable of performing various tasks in several themes, and there you go. It can solve, in fact, simple questions up to advanced programming. In this board, we will see that we are comparing Cloud with version 3.7. T Bel, this is extended thinking. In fact, this is simply the extended option. If we go to Cloud, we will see that here. There is this option. Of course, it is in promde. You had to pay a subscription of $20 per month to access this option or stay with the classic model of 3.7 sont. If I come back here, I will see the two models here. And we have actually tested them in several benchmarks. That means these are tests, for example, here. In terms of coding, we find tests in mathematics, university tests, and in all these tests, we saw that 3.7 ranks first or also is very close to the best result in several categories. Above all, it is compared with openiA. It is compared with Grog three from LM Mask, from project from Xcom actually. Also, in comparison with the Chinese Artificial Intelligence Deep Seek, it is either in the first rank or very close to the best results. This is due to its ability to switch from mode, quick response to mode, deep reflection. This balance or rather this way of doing things makes it and it gives importance to Cloud 3.5, which is today a very strong new generation, especially we program on JRPeln today, in fact, it even had simulations with the tests it conducted on the game. Pokemon. In fact, it showed that it is the most creative and there it actually proves its versatility and robustness. This is the chart that showed that today, Cloud 3.7, especially the extended version is the best artificial intelligence on the market. 6. Introduction to Prompt Engineering with Claude 3.7: So very well, before entering the prompt here to send the request to our artificial intelligence, it was simply necessary to know how to write this prompt to get the best results. For that, there is a method, it's a method that allows for six steps. This method will simply allow us to have an optimized prompt for any search and question that we can send to the artificial intelligence, especially for Cloud. What's the idea? The idea is first to know that Cloud needs precise and specific information to be able to give you the right results. That's why from the moment you understand how Cloud reasons and senses the information it needs, you simply need to get used to respecting the art of creating prompts. Next, I will show you here. The precise method is one that you simply need to apply once applied, you can easily achieve an optimized result. On roles and on roles and on roles and on roles and on artificial intelligences, especially with Cloud. To start, we have what we call at the beginning, precision and clarity of the information. This is a very important step. Here you simply need to formulate your request with clear and specific instructions. You should not provide very broad information and avoid anything ambiguous or vague formulations. Also, I'll take a very simple example. Instead of saying, tell me about AI, use a phrase, where you ask or you say, explain the recent advancements in generative artificial intelligence in three to four paragraphs for a non technical audience, and you will see that the result it will give you will be very different from the result of a simple or basic prompt. Here we will tell it. Tell me about AI in general. 7. Defining Role and Context: Very well. Now we are going to move into the section role and context. This is very important quite simply. When I input my prompt, it's very interesting to start assigning a role to my prompt. I'll take a very simple example. Imagine that I want to get information about what a local network is. Instead of asking it the question, what is a local network? It would be very interesting to tell it. You are a cybersecurity expert. I want you to explain to me the details of a local network, how to secure a local network. You see there, when I start with you are an expert or I give it reactors, we have made it easier for artificial intelligence in the field and it will simply take on that functionality. It will try to be much more expert and precise in the result it will provide. Suddenly, whenever I define the role, I will help artificial intelligence give me the best results. 8. Using Optional Parameters in a Prompt: Very well. Now we are going to talk a little about the limit because with artificial intelligence, you need to set its limit. For example, when I ask it a question, it would be very important to tell it. Listen, the answer, I want it to be long, short, and by giving it this limit, you will see that the quality of the result will be adapted to your need. Because sometimes when we try to use print, we need, for example, quick short information and a quality response. If we don't specify, you will see that the intelligence can write an enormous amount of data and information and it will overload us with that data. Sometimes we have the opposite case. Sometimes we really need to create a report to do a complete detailed analysis. At that moment when we specify that to it, you will see that it will activate its analysis mode and delve deep into its analysis. That is very important. For example, I can give an example of a prompt where I tell it. Listen, I want you to summarize a document and the result. I want it to be in a maximum of two paragraphs with accessible language. Once we have finished the limit part, you will see that everything we will see here will be optional. If you include it, it will provide even more precision. If you don't include it, it won't significantly affect the result. I Ibn say the format, sometimes when we are developing, we need feedback in a specific format like XML. Sometimes we give it an Excel file for analysis and we will get a result. This format, for example, of a table or certain columns. When we specify the format, you will see that the AI will adhere to that format which interests us. Next, we have the feedback part. This is very important because with artificial intelligence, when you input an initial prompt, if the prompt does not meet your needs, the result at that point, you can tell it, Listen, I want you to redo the response, but with simpler language, not using, for example, technical terms or conversely, I want a more precise response or take that point and elaborate on it a bit more. As a result by iterating and giving it feedback, it will better understand your way of doing things. In the end, the example, the example is very important. I take a very simple example. Imagine I want to create a LinkedIn post. For me, the LinkedIn post, imagine I am used to posting on LinkedIn. I have my own way, my own style. There is a structure that I always want to keep. At that moment, it's very simple. I can give it five or six linked in posts that I have already published, and I tell it, Listen, take this format respects this way and I want you to write me a new post on such and such a topic. You will see that when you give it examples, it will understand you better. It will work much more and process your data and feedback. So when we apply the six steps, you will know that you will have a truly optimized output and that the AI will try to go as far as possible to meet your needs and provide you with an optimized response for your output. 9. Step-by-Step Reasoning: Here's an example. This is simply a basic print. I am actually asking it to explain the concept of reinforcement learning in AI. However, look at it when I actually improve this print. Here, in fact, I start assigning it a role. I tell it that you will act as an AI trainer for beginners so it will understand a little bit that it will facilitate understanding. Afterwards, I ask it to explain to me the concepts of reinforcement learning in three to four paragraphs. At this point, I will give it a little bit of guidance that I don't want it to be too slow, not too short, three to four paragraphs. I will ask it to explain with simple words. With concrete examples and that the structure of your response should include an introduction, key points, and a conclusion. In any case, this is a point where it will give me a response, more detailed, easier to understand for a beginner. A little bit regarding some tips to better move forward with plant engineering, it's that if you simply have a complex problem and you want Cloud to help you solve it, try asking Cloud to go step by step in its reasoning. This means that each time it receives your problem, it will break it down into several steps. It will start sending you back each step with an explanation of its reasoning. There it gives much better results than just giving it the print all at once and asking it to give you the entire result directly. Imagine you have an Excel file with banking information. You want to conduct an in depth analysis regarding the sales trends in the company. Instead of asking it directly to give you a summary or a complete report, try to proceed step by step to conduct the studies and break down your problem or your request into several subproblems. Sometimes you can ask it that and tell it, Listen, my goal is to have a complete report. Try. To fix or distribute the problem into sub problems and send me personalized and precise reports as you go along. You will see that it will propose the five or six steps to take to ultimately find the complete report. This helps artificial intelligence to think better and provide you with better solutions to your problem. Also, it is very important to use the reasoning mode. It is very important to also go with this extension option where it will favor the thinking mode the deep reasoning mode, especially if you have slightly more advanced problems, and that is very interesting, especially, in fact, if you are studying and conducting analysis. Consequently, having a pro version can be very interesting. Finally, don't forget that there is official documentation sent by the official site. This documentation perfectly explains the art of creating plants with Cloud. Consequently, there we will obviously find several examples, several details regarding the functioning of Cloud. Generally, this method is one that is based on this documentation to simply get a very optimized response. Finally, when you follow this method, you will optimize your interactions with Cloud and you will simply obtain more precise responses and above all, useful responses that meet your needs. 10. Guide - Coding Prompts for Claude 3.7: So Cloud is known today as a model of artificial intelligence with coding power. It is actually capable of coding with enormous precision, today we can actually use this intelligence, especially Cloud 3.7 to Mast create But code. That's why the use of Cloud has two levels. Either we are experts, so we will need Cloud, for example, to correct code or fix bugs. On an already existing code, that's what experts use it for today or sometimes to give it a complex problem in a code and Cloud will simply solve it here. There is also a second use where beginners will simply ask it to create a code or to set up a learning system. Because thanks to Cloud, we can also learn programming languages, do exercises and corrections, step by step, explanations and even real time corrections. To start, it's very important to see that here, Cloud in terms of code, we have a whole documentation, in fact, that talks about Cloud code. This is, let's say, artificial intelligence that is specialized in coding. If I come here then in the official article 0N x.com in Tropic, they have highlighted this documentation on this Cloud code. Cloud code, we can launch it on the terminal so that it can simply respond with precise development. Also we can connect it with what we call a code editor, where it will use the Cloud API to really leverage this Cloud code. Don't forget that there's a whole comparison that has been made here on software engineering. This is a very well known test where they will test the models to see if these models can solve complex codes. I I compare it with Deepsec with open I, whether it's the oh three mini high or the 01 cloud is truly the most accurate and it obviously comes with a precision rate of 110.3%. As a result, the coding precision is very high. Today, it is the most powerful model in terms of coding. To understand how to set up the prompt, whether you're an expert or a beginner, you need to follow a structure because Cloud actually has a whole documentation on how we set our prompt on the system. That is to say, we won't just tell it. I want to create a Pacman game, hit Enter, and it will develop it. It's true that it can give you a game, but if you want more precision and code quality and especially a system that is capable of providing you with precise code, you need to develop your prompt well. And don't forget, especially when you are going to work with what is called a code editor, another intelligence, another tool that will use the cloud internally to unlock IA code and fix errors that comes at a cost. So if we use prompts that are very general, not precise and not subject to specific requirements, it will be very expensive and will always result in codes with bugs to fix. On the other hand, if we know how to write the prompt well, the result will be perfect. Based on this documentation, we took into consideration all the recommendations and simply grouped them into the method called codex. This method is one that shows the best practices for writing an optimized prompt for code for programming. And of course, it consists of five step five recommendations. This documentation is also available in the core support in French, it's the same thing. You can view and follow it in French. But here we will just stay focused on the one in English, the original to perfectly understand how to launch a prompt. This applies to both beginners and experts. At the end of this documentation, we'll see what we call templates for each domain. You can simply copy this template, replace the information with your specific needs, and you will see that the result is truly optimal, and it simply gives you high quality code that meets your request very quickly. 11. Prompt - Context and Constraints: Well, then to have quality code, we actually need three very important elements. Let's start to see first, a code that is as relevant as possible. That's why we actually need to have a relevant point. As we need to have a functional print and a print that is well structured. If we actually provide these three elements, we can ultimately have quality code. The more details and precision you give to your imprint, the artificial intelligence or the model will simply react in a much more precise manner. So if I start with at the level of codex method, we can begin by talking about context first. It's important to actually know at the level of context here who generally generally for people who are knowledgeable, those who are not knowledgeable cannot know the context. They do not know programming languages. It's optional, but if you are knowledgeable, you can provide the context with which the intelligence will react. If we say context, what is it? It means we will precisely describe the context of code usage. We will give it, for example, some constraints like versions, a specific environment, specific dependencies. Imagine that you want to A code or a program that is already developed. You want to continue the programming, you no longer want, for example, a developer to do it manually. You want assistance with AI and you want Cloud to finish the programming. At that point, we need to provide it with information and constraints and explain the environment. Example, I tell it, listen, I am using Python to on version 3.10 and as a result, it will continue the programming with this specific version, so it won't develop a little with something else because afterwards, it might cause a conflict in the code. We can tell it. Listen, for example, I don't want you to use external libraries like Pandas, for instance. This is for programmers. They understand this. As a result, it will adhere to the current code you have. Generally, this mode context, we will need a lot. Otherwise, we want to develop something specific with a precise language or it's something from A to Z. That means we will do, program everything with Cloud, we will require it to comply with that. If we already have existing code that we want to process, it's very important to provide it. In fact, the constraints, the versions, the environment so that it can really execute your code correctly and be precise. If you don't do that, if you don't provide the context, be sure that Cloud will look for the latest versions, the best environment. But what happens is there can be a conflict with your existing code. Especially if you have code that is already programmed, already developed, perhaps with another AI. To have continuity, it was really necessary to specify the context. If you don't know if you are really a beginner, you will at that point provide the following information without specifying the context, and at that point, Cloud will look for the appropriate context or the context that best fits your need and request. 12. Desired Prompt: Well, so here we also have the objectives. When I launch a plan at the coding level, it was absolutely necessary to set the objective clearly. Of course, you must give it the functionalities you want it to develop. Above all, you must set the priority and that is important. Imagine you want to create a website. If your priority is to work on the design, work on the homepage, that comes first. Then you can tell it that you are working on, for example, electronic PMO, that you want to integrate functions like stripe, external integrations, that you need databases, so you will list the functions and actually set their priorities at the beginning and it will understand the sequence of development and programming. And also we have a section called data examples. Here, you need to provide what are called inputs and outputs. You want to create an application. The application will receive user information and provide an output. The data and outputs is very important to provide them. The more details you provide, the better it will perform the programming. Then we have a very important section, which is the desired explanations. Do you want Cloud when it develops you want it to give you explanations regarding what it has done? Because here you need to be careful. There is a difference between explanation and Cloud's reasoning because Cloud when it develops, it will provide you with its reasoning. But you can ask it for explanations about that and then it will give you comments actually where it will explain, in fact, in a way precisely what it is doing. When we ask it for explanations, you will see that it can actually give you questions. At that moment, it will show the locations or the modules that it did not understand or it will ask you what your decision is regarding certain choices. As a result, asking it for explanations will create a bit of interaction between you and the programming with Cloud and in the end. That of course is somewhat at the level. Of the experts. These are simply specific requirements. If you have, for example, simply requirements, of course, not negligible like you're looking for a specific style, you're looking for specific performances, then it is very important and you can of course request unit tests here if necessary. For example, there you will tell it. The code must follow the convention if there is a convention, for example, PPI. It's a well known convention in programming and you want that, there you go. It includes, for example, three unit tests when you provide specifications, it will execute them, it will adhere to 13. Comparative Examples: So here's an example of a simple prompt. If I look at the prompt there, I tell it, write a program to sort a list. Look, when I try to apply this method, where I give a much more detailed prompt. Well, obviously, the result is going to be really different. In this prompt, I simply start by putting the context at the beginning. For example, there, I tell it, I am working on a Python project and I provided the Python version. Only, for example, there, I give standard libraries. This will prevent me from later installing other specific libraries which could create problems with hosting or Bahas or execution. Then I tell it, here is the objective. I want to create a custom sorting function for lists, and I specify here, these are lists of dictionaries containing user information like name, age, score. As a result, I provide the exact context. Therefore, for it, it will perfectly understand the context of my programming. Afterwards, I provide it with the inputs and outputs. That is very important. As a result, it will perfectly create the database containing these input information and it will understand the function, how it will simply take its parameters. Then I provide additional explanations. I tell it that I am well versed in Python, but not an expert in sorting complex objects. I want you to add comments on the functions beyond the coding. For me, when I receive a code, it will still be commented that code, and that is very, very important. Those comments will be useful for humans. Later, if I want to continue updating this program and also don't forget those codes, they are codes, that cloud, when it implements them, it will help me later to ask questions if I want to ask them, if there are functions I want to improve, if there are functions I think are not well done. At that point after testing, I can read the comments and I can simply ask it to do something else because reading the comments is of course, much simpler for a human than understanding code. I can as you see there, at the end, I gave some requirements. In fact, I tell you that the code must be efficient, even on large lists. I tell it that here, I can send you a list of more than 10,000 elements. As a result, even at the programming level at the database level, it will take that into consideration. It will work on indexing functions so that it is fast and all. There you go. Include error checks. If I give it an input that contains errors, it will give me messages to tell me, there you go. There are errors in that input. As a result there I have given all the information and the result will be quite simply different. From the result of the first print. This first print is too general. The second one is very specific to a specific need. It's normal this method, it's a method that corrects that gives me an optimal print to find a code and the result work that is simply optimal. 14. Ready-to-Use Prompt Templates: Here are some prints that are ready to be used in various contexts. For the first context, imagine I already have code that contains errors and I want to fix those errors to do what we call debugging the code. For that, I first need to give it the context. Tell it here, I use and I put the name of the language or the framework with the version, and then I tell it here. You objective is to debug the following code which should, for example, I actually give it the function I expect, the expected function, the objective function. I tell it here. In my program, it must do this and that, but it produces an error message and it would be very interesting to copy paste the error message or take a screenshot and attach it. Afterwards, once I actually input the error code, I can give it either the output or the error. Also, can receive it when I have changed other information about the code. As a result, I can give it a little bit because sometimes I try to correct the code, so the error might change. I can give it the maximum amount of information about that error. Afterwards, I can ask it for explanations. It, explain to me the source of the problem and how to solve it. Then I can tell it, look, I want a solution that must be simple. That does not require completely rewriting the function or my code. Here's another example. This is for learning, a concept. There I tell it. Listen, I want to learn, for example, a programming language and I can tell it here, this is my level and I started, for example, two months ago or three months ago. Then I can tell it. Listen, I want to learn how to implement or program or create a function and I give it exactly what I want to do. Then I can give how I have already tried to do it, the approaches I have tried, for example, I watched videos on YouTube or I bought courses, or I can even simply give it the link to that course. Then I can ask it that I need step by step explanation so it can help and guide me. At that point, I can tell it Listen, I also want detailed feedback, so I can learn and then it will literally create a step by step course for me to learn how to develop or program something specific. There's also to optimize existing code, sometimes you might have code that when you run it, for example, consumes a lot of memory or is too slow to produce results. At that point, the same thing, I give it the context, the objective. I tell it here. I want, for example, to improve performance or the readability or the code output. There you go. I give it all the current code. Then I tell it, Listen, you're going to detect the problems and explain them to me, the optimizations I can tell it Listen, I want you to maintain compatibility with the existing stack because sometimes it can improve, it can have new codes and new errors, and it's very important that it keeps your program, that it keeps it functional. Too, it's a little bit very interesting. It's the one that most people, especially beginners and those who want to create the program from A to Z, here's how I can make a SAS that is simply applications so that I can provide an online service or an online application. For any SAS application, we first need to give it context. I tell you that I am developing a SAS application and I provide information about the sector, the technology I want that you use, and Also I tell it what infrastructure I want you to aim for in the creation then I said the objective. Quite simply, I tell it that I want to design the architecture and the base code for this application that allows a user, then I give it the entire list of functions and it's very important to specify that context, objective. After that, I start giving it. In fact, the data meaning the functionalities, but be careful in order of priority. I tell it here are the functionalities in order of priority, and I list all the functionalities. Then I can actually give it the description. Here is your user who will use the application. Then I provide explanations. In fact, at the architectural level here, if there are diagrams, well, this is for people who have done programming, it creates what we call the concepts, the diagrams. I can upload them and it will simply code and program them. In the end, I can provide requirements, for example, on the side, security, for example, authentication, security. These are simply requirements that I can make and finally, we can also see templates for creating applications whether web or mobile applications. If I take the case of web applications, here, for example, is an example of a prompt, so I tell it that I want to create a website, I actually provide the type of site, the objective to say here is the audience I want to target, the technologies, whether I want it to be done with HTML, CSS, JavaScript, or if I wanted to use a specific framework, then I get into the content, meaning that I will give it an objective. What will the site be used for and therefore, I will need it here? To mention, as you can see there, the information on the homepage, the other pages, if I wanted to develop them, the aspect of accessing resources. If I have images, APIs, I want to give it the information here in the prompt right from the start, then I give it explanations. For example, start by presenting the HTML structure, then the CSS, meaning I give it a bit of guidance on how I want it to program for me. In the end, I can ask it for the quality requirements of the result for this site. I can even ask it from the beginning. Look here. To write very good SEO code for me, so as a result, it programs and does this to ensure it is optimized for search engine optimization. In any case, these are a bit of the preferences and requirements that I can set. The same applies to the mobile application, so it's very, very similar to the web application. But here, I simply specify for which platform, whether it's IOS or Android. As a result, for him, he will prepare the structure from the beginning. But the rest is practically the same as a web application. I provide the explanations, I set the requirements, and there you have it, I will simply have quality code because I have simply set up a quality prompt. 15. Introduction to Claude 3.7: So here are some recommendations that I can put at the end of my prompt. You will also see it makes a difference with those recommendations. At the beginning, I can actually ask it to develop or write the code and explain it. Rather, explain its logic step by step before starting the creation. This actually improves the final result because I actually require Cloud to start simply at the beginning by explaining to me what it is going to do. Everything from A to Z and afterwards, it will start applying it and that improves the quality of the there is also the possibility to add this, which is that I can ask it to start with a simple structure, then request specific improvements. That means if I want, for example, to create a program, I will give it a list of features and it will start working on them. Those features without delving too much into the complexity of each one. Then I will ask it to take those features one by one and to start developing them in a precise manner. If I tell it, start this way and create a simple structure, you will see that each time it opens the code for you, it will tell you, here is my next step or what I propose to you. Do you want us to start developing the first function more broadly or in depth? So you will follow the coding step by step. Each time you say, you finish this function, then you tell it to move on to the second one, then you tell it to move on to the third one. At the beginning, it simply creates the overall structure and starts developing there is also the possibility, in fact, this is to ask it, for example, the format of the artifacts that in a way, helps, especially for code that is very complex. This helps later with reading and reusing the code, and there is the possibility, in fact, to ask it to explore other alternatives, and of course, you need to know those alternatives. You need to be a developer to know the different alternatives you tell it. You're going to develop this test with two technologies and tell me which is the best approach, and of course, it will make a comparison for you. It will help you make the best decision. Developers actually use this a lot, especially on complex projects because the choice of structure from the start is very important, as you can program for several days and in the end when you have a very large code base, a very large site, sometimes, the perspectives will prevent us from programming if we have started with a certain structure or technique or a library that is weak or not well developed enough. That's why we sometimes suggest comparing approaches. Finally, in conclusion, this method will allow for the formulation of plans that are precise and complete the goal, in fact, is simply to have better code provided by Cloud. By following this structure, you will save time and you will actually obtain solutions that are more suited to your specific needs. You understood the difference between putting it, actually giving it a simple print and there you go in a detailed point. The difference is clear in the result. That's why it was necessary to adhere to certain information to provide it with the maximum data using this structure, context, objectives, the data we are dealing with, explanations and requirements. You will see the results will be of very high quality. 16. Understanding and Using Claude Code: Very well. Here we are going to try to talk a little about Cloud code. This is the part. In fact, this is where Cloud code will try to help us, in fact, to create code or simply to fix existing code. We simply find a whole documentation there that explains to us how to work, how to leverage this code power. It's a very, very long documentation. There is a lot of information and details about installation and code manipulation. Generally, there is the possibility to do it with a classic and simple terminal to be able to interact with Cloud simply by installing this intelligence on a simple terminal. We can simply access a folder and launch Cloud to start asking it to create code. But in reality, it will be much more interesting to use what we call a DI agent or a code editor agent. That's why we're going to try to use Klein quite simply. It's a coding agent and there we will simply attach it, access to the Cloud API, and as a result, it will develop for us while using the power of Cloud. That's how it will work. We will try to see the installation of Klein and the manipulation of it. Of Cloud as the best artificial intelligence, don't forget with the tests that have been done recently. Here we find that Cloud is much more powerful than Deep Seek, more powerful than 03. Mini, the same as one, even when comparing it with the latest version of 3.5, it is the top. Cloud 3.7, Sony, it is the strongest coding and programming. Let's first take a quick look at the official website. Of cloud there. I will show us the use of this API, the power of cloud. You will see what it has here. It actually gave us the ability to do. Here is the API connection with several types of intelligence. There is one that is a bit light and fast. The one that is very, very strong in management and creation, in fact, analyzing towards tasks and tasks that are complex. But for us, we are much more interested in this one. Bell that is very strong in coding and programming. In terms of cost, it's more manageable and as a result, that's what interests us the coding part, the programming part. If we also look at the price, it's very important to know how much the Cloud API dog will cost. The Cloud API here, of course, we will use this version 3.7 bell. For everything that is input, it costs $3 per million tokens, and for everything that is output, it's $15 per million tokens. It's let's say, a bit more expensive than at the penny level. If I look at the gear, I can compare it with this one in terms of water three mini because this one is very strong, also in coding, it's much better than the others in terms of coding. So here the prices it's 1.1 dollar per million tokens for the input. And for the output, it's 4.4 per million tokens. So we understand that there's a price difference. We can say that here for the input, it's almost three times and the same thing. Here, for the output, we have 15 and we have four. So we can say that it's three times more expensive. But the power, you will see that it's cloud 3.5 it's much stronger in terms of coding and quality. Honestly, if you do a project with the 03 API, it will cost you just as much because there will be a lot of error fixes that you will request. You will consume a lot of the API, unlike here, the API is a bit expensive. But the quality of the results is better, in a way, the price, it is justified so that we can simply make these installations there and start launching, first, we need something called the API. We need to create our API key and of course, purchase some credits. We will try to look at this in detail. 17. Getting and Setting Up a Claude API Key: Very well. What we're going to do first is simply go to this site here, console log intropic.com. Here you will see in the setting section there. I can simply go to Billing and I can click to simply purchase credits, so the credits when you see it, when I enter my credit card, when I click here, you can choose $5-100, of course. As you work, the limit can increase from $100. But at the beginning, I can, for example, enter here, add $5 and click on By credits. The the result quite simply, I will have credit here. This will give me the possibility to click afterwards on happy Keys and there I simply have the possibility to create my key happy. I will click here. Of course, it will ask me to give a name to my key. I will name it Klein quite simply, since we will do it with the Klein tool, so here I will decline. I will click on AD and you will see that now I have my key. This is a key, of course, that you should not display, you should not share it. It's a secret key because if someone uses it, they will consume your credit. I will create it just for this training, for this video. Afterwards, we will destroy it, so there's no point in copying it. Then I copy it, I finish this and so I simply have here my key that has just been created correctly. 18. Installing Visual Studio: Quick Guide: Very well. Once our API has been created, we are ready to use it. Quite simply, we will need our Visual Studio code. Here, the website is code.visualstudio.com. I enter here. In fact, this is a code editor quite simply where I can open files, I can open sessions. I can actually install extensions here, it's thanks to this tool that I can actually install the client extension inside. Quite simply, it's an interface, if you will, that will actually help me install libraries and execute certain codes. To download it, it's simple, it's free. You will simply click here on Download. You will choose your operating system. Personally, I work with Mac, so I can install Mac. Even within MAC, you can select your chipset. Is it an Apple silicon or a tel ship to have maximum compatibility with your system. You also have those who work with the Linux version and of course with Windows. Once you have completed the installation, you simply need to click Download and then click next, next for the installation. You just need to launch it 19. Installing Cline and Configuring Claude 3.7: Here is our visual, the code studio which is installed, so to install the client in the world, I've already installed them. But to install it for the first time, you will click here. So in what we call the store and there, you will simply type Klein like this and you will simply find therefore the lien tool, I remind you of that. This is the site of Klein and the possibility to actually install it from the site, but it will be much simpler to go directly from your visual and install it when you click for the first time, you will see that you need to of course, check that you have the very latest version because there are always versions that update it has just updated automatically, as you can see, as a result, you will have a button called instore. It's the first time if it's already done. You will simply click on Auto Update so that it updates automatically, and that's very interesting. It allows you to have the very latest version of Kline and once it is installed, you will simply click on Restart extension so that this extension restarts for your system to be ready to launch, now what have I done? I've installed visual, I've installed Kline as a result, when I clicked here on the left on Kline, I have my support which is now installed. So here I actually give it the authorization, and now what am I going to do? I am going to install, in fact, the connection with Cloud 37 directly here with his input. To do this, you will see down here. In fact, there is the possibility to select your coding model. I click here and well, I work, for example, with Open AI. Now we are going to switch together to another tool which will be clouded, so I select here. You will find anthropic there. I click and you will see that there, I need to enter my key. Remember that we created a key together. As a result, I'm going to copy and paste my key there you go. It's a test key that we created together. Now I am going to paste it here, so it pastes in a way, encrypted, that's very interesting. I won't appear in your code and there you will see that there are several models. I am going to take the very latest model, which is 3.7 so that I can simply use Clouds power to the fullest. When you see it there here, I simply have the Cloud model installed with line on the visual site. As a result, I can code, I can program without any problem thanks to this artificial intelligence. 20. Create Code with a Single Prompt: Well, here, I'm actually going to do a test so I can simply put my prompt here. But before putting the prompt, we are going to create a folder so that simply we will open folder. Quite simply, I will simply show it a folder on my computer, so we will simply name it. Here we will create a web page, for example. We will open this here and there. We will actually launch our prompt. There incline, I would actually need to launch my prompt, so I will paste a prompt. We will increase the size a bit to read the prompt. So at the prompt level, I will tell it that there you go. You are an expert in development and I would need it to write for me. There you go. An HTML web page with an animation and I wanted to return just a single page where there is HTML, CSS Javascript. The idea is simply to make a weather card, where we will find an animated weather map, there is wind, rain, sun snow I want to have a small button to interact with these maps. Quite simply, it's a test to see the platform's capability and the power of the code and if it is capable in fact of generating code with an animation with audiences and HTML. As a result, you need to see that I have put my prompt, I'm simply going to do it here. Press here to launch it. Don't forget that we are on Cloud 3.7. There we are indeed with our model and I start the programming. You will see what is going to happen. So it will actually start to develop, create, in fact, the information, the data. As you can see here, it's really very fast and don't forget that it actually has the capability to reason quickly and to actually move towards very advanced reasoning. Now, as a result, it is in the process. There you go creating the code, honestly, in terms of sending the code. It is very, very fast if I compare it with the others, intelligence, therefore, we'll wait a little bit to actually see the final result. So here, it is actually creating this file, so there is the possibility to create what in want. Do you know New file. Now with the creation of this new file, we'll give it a little time to finish. The development and complete programming. Now it has actually finished the coding and you'll really see it has typed almost 794. Lines of code, that's huge. It has done all the functions and now here it is actually trying to execute. In fact, it's on the internal browser of the system. Either I will actually wait for the complete installation or I can simply execute it to see how it actually works. In fact, it works, so we'll wait a moment until it finishes. In fact, the execution of its code here and its internal browser and there you go. It has just finished displaying. Now I actually have my application and here I can manipulate. Can see it here, but we can do better to execute it and view it directly on our browser. If I scroll down a bit here, it has done the recap made its explanations, and now it officially shows me that it has actually completed its task. And so I can scroll down a bit here to simply execute this code. If I click here, we will try to launch, actually enter the terminal. When you see here and there, it will try to send me the work. I'll show you right away. We will simply here. There you go. This is the page that just sent it to me. It's a page, therefore, that was executed on this URL internally within the folder I created. And you will see that there you go. It's a very beautiful map. In fact, where I see an animation of the sun, rain, wind, and snow and it's very beautiful. There it's flawless. It's an explanation actually, that is simple. Maybe I will need to make other modifications. If I want to, in fact, modify it or if I want to improve or even create an application to make it much more developed. But still, as a first result, I find that the work has been very well done, so it accomplished its task. I have my entire file here. And I can, of course, continue the work. As you can see, of course, there is a cost associated with it. Every time I launch the application, it will consume. I I come back here, in my opinion, if I refresh this page, I will find in fact, there is consumption at the user level. This is the number of tokens it has consumed. I can obviously see how much money I have left. Obviously we started with $5 and now there is 450 left so it has consumed, let's say, $0.43. It's quite costly, that's for sure because you can feel it when you compare this work with the work done with hat DJPD using the OpenAI API. But still, in fact, the result is a very good result of high quality, and that actually explains why. There you go. The cloud, there you go. It does a lot of iteration and reasoning to give us the most correct code possible. Not to mention that it also does a whole part at the code level. I will explain. I have asked there you go. I will actually demand that it gives me a very clean code adapted so there you go, what it is, that means so that they are in extension so there you go implemented in another application. Therefore, the way which it implemented. This code is a very optimal way 21. Generate My Documentation Using PLAN Mode - Part 1: So if we also take advantage of the feature offered on Klein, it's the one that is available here. In fact, if I switch to this interface, I will ask Klein here to do it for me, what we call the complete structure of this project and this will actually be based on documentation where it will simply provide me with all the details, in fact, all the guides, the necessary articles to understand this code, and also it will organize it in the format of a project. It will collect, there you go, the information, the files, the folders and generally, this will help me enormously. If I want to continue developing this project with another company or with another artificial intelligence, and this helps me to continue the development. Therefore, if I come here, I will simply paste the prompt here. This is a prompt that I will ask it to work on, which will act as an expert in documentation. Don't forget that Cloud has a very advanced reasoning capability which is very sophisticated. I will try to launch this prompt right here in this full mode so that it gives me the perfect structure of this project. I launch it here it will take a few moments to go through again all the code so that it provides me with what we actually call a plan, a general and global project architecture. We'll actually leave it in execution mode here to see the result display afterwards. There you go, it just finished. I gave me information about the architecture of this solution. If I scroll down a little here, you'll see that it will draw for me. In fact, a diagram. This is what we call the workflow. Here it will simply provide me with, in fact, the the projection and the flow of information on my project and this will greatly help, in fact, to understand projects. Don't forget, in fact, that sometimes I can have a very complex project and this project, I can simply in fact it, understand and determine it in a much simpler way thanks to what we call workflows that will simplify. In fact, the flow of information that will exist for me, in fact, it explains to me especially for the ownership of the code, for another developer or even for another there you have artificial intelligence in case I ever continue with another intelligence. 22. Generate My Documentation Using PLAN Mode - Part 2: So I have the option, as you can see here. It offers me either to confirm the existing code status, or I can still complete the documentation. If I simply want to copy it here, ask it to continue the documentation, you will see it will do what is necessary to have the maximum reports on this documentation, the maximum diagrams, workflows, this remains very interesting, especially in large projects if I really want to sell the project or simply put it in deployment mode or scalability mode. Uh to add several other modules. There you go. It really switches to reasoning mode. That's actually the second functionality of the cloud which is step by step reasoning for understanding and for me, I can take advantage of it. There you go. It gives me the project architecture, as you can see here, it really provides me with all the files, the functions that exist in this project in a very, very precise manner. It gives me the explanations. It actually provides me with the levels, the information on the diagram. Look here, it's really creating a visual. Really, this is what makes it powerful when I combine Klein with Cloud because Cloud alone, with the terminal, it's a bit complicated to execute even to see the code execution. But here, it really becomes something very interesting. Here it suggests that I continue towards the documentation, but well, we can stop at this level. Understood that with the plane mode, it's a very interesting mode with Klein, which allows me to work on the architectural side and I can at any moment switch to the action side to actually return to executing commands and simply work on my code in a much simpler way. Here I switch to action mode and in action mode, it just needs me to input the prompts. 23. Creating a PDF Generator Project: Very well. Now we are going to do an exercise where I will try to create an application with Klein using, of course, the Cloud 3.7 API. But what is the idea? It's simply that I want an application. It's a SAS where I will input my resume in PDF format. This application will read it and I will also add as input in the entry, a job offer description. What is the objective? It's that this artificial intelligence will read the resume and the job offer description, and it will create a new resume for me that is more suited to that job offer while of course staying focused on the resume I initially sent, my original resume. Afterwards, I can download that resume. I'm going to try to create a new folder here. Here we are going to create a new folder on the desktop, so we will call it PDF. There you go. This is the new folder. I will simply copy paste I created a detailed printout so that I can run it here on Klein. Here I put my print. This is the detailed print out. I explain to it step by step what I want to do. I tell it here in the print that I will also use the Cloud API. So that it can perform the analysis and generate the new PDF or the new resume. Now I launch it. Of course, it will take a few moments to complete the reasoning, the experience, the understanding of this project, so that it can simply proceed to launch the programming and coding. Generally, it takes 3-4 minutes for it to finish the reasoning and complete the initial coding. You will see here, of course, it tries to learn my confirmation, there you go for executing a few command lines like here actually for PDF processing. I tell it yes. Now it is simultaneously installing libraries so that they can really get into the code and develop. Which is really impressive actually with Claude, he takes his time for reasoning to understand the entire project to set up. Here are the resources, the libraries they want to use. Afterwards, there you go it's off. He will actually start programming. Now he will develop and we'll actually program the new files and we will actually let it run. And I will get back to you as soon as he finishes or as soon as he displays an error message to see a little bit how we can react in the face of these messages, 24. Live Testing My SaaS: So they are very well then. He tells me that the application is ready and that I would simply have to click here to actually see it run. Of course, if you noticed here, it has created several files at the index level, the CSS JavaScript. I even created for me, this is very interesting. I like it the red me, so it's a little guide that explains to me, here are the details of the application. I really like it. It's actually here's the architecture. Therefore, the mode of the USI link. The documentation was done here and that's very, very interesting. I find that it has thanks to the plants. In fact, he managed to create a clear simple project. And now we are going to click on a year to execute it. We will simply need to see the execution now when you see it there, so it will run it on a local machine. We will see the result. There always the execution remotes, we are waiting for the server to open to see the application in reality. Here I see then the execution of this project. It's a simple thing. Here, I can upload my PDF, here, I can add the description, and here I can actually ask it to perform the optimization for me. I'm just going to take a test resume that I downloaded from the Internet, a simple resume. I'm going to click here. For upload, we'll first see if it managed to successfully upload or not. Here I see that the upload is working correctly, so I can already see that it has detected here. Actually, the file, we are going here on the Internet to look for a job offer in accounting. We're just going to copy an offer just to test. We're going to take this offer. We are going to holy copy it and try to paste it here. All that's left to do is click on Optimize my resume to see a little bit if it is capable of optimizing using artificial intelligence. In reality, here, it gives me a message actually, file to process, CV. Then what am I going to do? I'm simply going to take a screenshot or just copy, in fact, the error message, which is this one, and we'll go back to our interface and I'll tell it to fix this error. And there I'm going to simply put this error message here, as you can see here, there I have. But actually, the information, I can also add a screenshot. It will be very interesting to insert a screenshot here. I'm going to click here to actually add the image of my screenshot. I just took it from there. There you go. And I'm simply going to ask you to fix this error, and we'll see then Lou, in fact, it has the ability to reason to actually understand where the error comes from, to understand exactly what is needed to fix it, and then it will propose solutions. As a result, we're going to let it update the code. As you can see, it's currently updating the code. We'll wait for it to finish that to re execute and retest the execution, not this program to see if it managed to fix it or not. 25. Updating My Claude API in My SaaS: So he has finished fixing the error. Now I can simply see my application. We are going to test it. There, I left the PDF there and the job description. I am going to click. And you will see that I have an error message. There it tells me that it cannot use the Cloud API, and that's completely normal because I need to provide it with the API. That's a good sign. It means that it has actually moved on to reading the PDF and reading the description. At the moment of requesting the link for analysis, it cannot establish the connection with the other. As a result, it's simple. I will come back here. Look, I will simply paste my API that we created. On the Cloud platform, this is a test app for the training. So I will put it here. I just need to press Enter and you will see. I don't actually need to follow those steps to create or update the file and the stack, I will let it do it by itself. Why? Because if it works, you will see that it will create the necessary file. When you see it there here, it will do that for me, and therefore, my system can be ready. Now we just need to test to see if the stack responds correctly to its request to see if the error has been fixed or not, 26. Generating a Professional CV with AI: Well, there he finished. What am I going to do? I'm going to go back to a page. I need to refresh the pass, very important. Redo the PDF applauding. We are going to copy our job description, ok show a second time and we will see here. Now, what message will it display for me or will it succeed in making the PDF or the new resume? The, I fresh. Then as a result, it tells me there's a problem. In fact, when it works here with the cloud, API, the same thing. There I'm simply going to do a print camp. We will copy that message, and I go back to my code, the same thing. I will paste it. For me, I completely leave the way I develop because it's a real way. When we have a developer, it's the same thing. They have an error message. They will try to find out. With AI, it's the same thing. But the power of AI is it capable of detecting its problem and solving it in a short time. As soon as it's done, I try to run the test a second time to see if it succeeded or not. There it was finished then. I have to go back to the application and simply test a second time to be sure if the problem has been definitively resolved or not. How to do it? I will come back here. We will refresh. Same procedure. I have to actually add my PDF here and there you go. I just added here. The description, we will rather paste the job description, which is this one. Then actually we will click here to start the optimization of my project. Now it's loading, that's a good sign already. It took a bit more time, so it's not like earlier, it sent the message directly. Now it's showing me that it's working. We will see the result if it is able to launch the PDF or the final version. If I can download it, if I can see it, read it. That's very, very important. That's already good. First good sign, it gave me a green message saying, Listen, the CV is ready, it gives me the option to download it. Now we will see if the download is done correctly or not. I click and Va a new file opens up and it's great. Now I have a CV that is optimized for the job offer I provided. Well, it's true that here design wise, it's not very beautiful. But still content wise, we can see that it has done everything necessary to make it suitable. Of course, this information here is the same as what's in the CV. We just provided a simple open source thing so as not to use a real CV just for the test. But of course for you for your application, it was absolutely necessary to test with a real CV to know if I can download it or not. Here, I can obviously view it, download it, print it as a PDF, and assist with it. I can further improve this work if I want to return to the application and ask it to work a bit on the design to have a slightly more advanced or more attractive design. 27. Using a Prompt to Improve My Design: All right let's get started. Here we're simply going to input the new prompt that will work on the design. I'm going to ask you to create a modern design that is beautiful and professional for a high quality downloadable resume. I'm launching this prompt. Of course, it will try to understand the mechanism and the objective of this PDF that I want to download and then it will start creating new files and connecting with new libraries to enhance the design. For my PDF, I'm going to try deploying this code and then we'll test it together. 28. Testing and Validating the New Design: And there you have it. He executed it. I see right here that he has indeed previewed the application here, but now he just finished. The realization involves the design, we asked for a simple and modern design. Now let's applaud. In fact, the file is just to test together. In fact, what it will look like, so we will give it time to change. So now he is analyzing the PDF, the text, and we are waiting here for the launch of our final design to see if the visual quality has been improved or not. There you have it, we will click to download the PDF. Indeed, now I clearly see the intervention in terms of design in terms of the addition of colors and titles, and there you have it. Afterwards, of course, I can ask it to do a bit more on the design side, but it remains as it is for now. A part that is very interesting, very important. I can even attach screens related to CV templates and I can ask it to format the CV in that way. That's possible too. 29. Creating and Connecting a Database: So there is also the possibility to connect to a database with our application. Here, I can choose, for example, Google's firebase model. I can create a free database there where I can connect it. I can even create user accounts so that users can simply log in with my application, create their account, and there you go. They can, for example, create multiple CV templates and actually use them every time they have a job offer, they can copy it and keep the CV directly in storage. For that, I will type failways. I enter here just to retrieve the access, and I will give it to the AI so that it does what's necessary to make the connection. Here, I execute to enter this database. Here we can simply create a new project. Here you click to create a project. You give a name to the project, we will simply call it. We did it for the CV or we will simply call it PDF. PDF, small here, I can of course, join the developer program when I can choose not to join it. I click on Continue. Here, if I want to enable the option, Google Analytics, that's for tracking my application to see visitors and information. If I want, I can leave it enabled, otherwise, I can disable it and I click on Create Project. It takes a few seconds and you will see that your project will simply be open and I can of course retrieve the codes from file buys and put them on Kline, so they can make the connection and create a database for my application. Very well, it's done. I click on Continue. Now I'm inside this project. Here it asks me whether my application is for IOS, Android or a web application. In our case, it's a web application. I give a name to this application, I can use the same name. In fact. So that we assigned to the project, it's not important. Here, I click on Save and you will see that it will give me a code which I can obviously give to Klein and Klein will take care of it. As a result, I just need to copy paste this code, this part where I have my firebase configuration information and I will put it up there. Thanks to this code, the AI can simply make the connection. I copy this code. I can simply take, maybe there's no need to include the function. I copy this information. There, of course, I click on Access to complete this configuration. I go back to my client and I will tell him actually here, we will simply launch a new prompt for him. There I will tell him connect with my data Bass there I give it the name Ferbes and then I go back to the line and I simply give him, in fact, the information. He will understand everything. As a result, I need to give him this information. I then launch what will happen. He will do everything necessary to connect my application with the database. 30. Setting Up a Secure Client Dashboard: So we also have the possibility to do what we call authentication in our application. Authentication here simply allows people to log in with an account, create a login, create a password. As a result, when I go into authentication, well, the authentication part, I can access it here just as I can obviously access it from the tab, create, you will see there, create there is authentication. It's the same thing. There, I click on Start, you will see that it will ask me to select, what is the authentication provider? That is to say, do I want people to log in with a username and password or with an SMS they will receive on their phone or with Gmail, Microsoft, or even linked to Facebook? So as a result, I can choose the provider here. I I activate, for example, the email address, if I click here, you will see that I simply need to activate this option, click on Save. I can, of course, add the possibility to log in with other providers by clicking there. I need to enable this option. Of course, you just need to click here and activate this option. That's left to do is simply go to my artificial intelligence and there I can tell it to simply create for me a user account to access the system. At that moment, when we tell it to create a user account, it will check if in my database I have activated this option automatically. In case it's activated, of course, it will create for me in my interface, the space simply to be able to give control. Here, for example, at the top right or left, I can give it the possibility to add a space here for users to log in before accessing this application.