This is how you build applications without knowing how to code using just English. It takes only seven steps and it’s really easy to program or generate applications that can become functional and effective programs. I’ve added all of the links and instructions on my website, at https://sourcetms.com/posts/ai-ides https://sourcetms.com/posts/coding-with-ai #cursor #ai #llm #startup #programming #tech #techtok

This is how you build applications using AI without knowing how to code. It can be done in only seven steps. The good news is that it's all in English. The bad news is it takes some time. Give you an understanding of what's happening right now. No code. So using a graphical user interface or a GUI is dying because what's happened is coding can be done so easily by LLMs and by agents that you no longer need to do this step of making everything visual. People are actually programming using English. The reason that's so effective is because LLMs have been trained on vast amounts of code. They can actually write code better than you can or if they can find novel solutions to what you're trying to build. These are the seven steps. The first is getting an AI IDE. All that is is a development environment and you can get something like Repl.it which is in the browser, Cursor which is on your computer or Windsurf on your computer. The reason you need the IDE is because it's where you do all of your actual programming or chatting with a language model to then actually write files to the computer to build. The next is defining a system prompt. This is essentially rules for AI. If you used an AI playground or an LLM playground, you know what this is. If you're not familiar, this is just a set of predefined instructions for the language model so that every time you interact with it, it adheres to these rules. The reason why we're setting up rules for AI is because within our IDE, we want to make sure that it's constantly following the rules we've set for it. There will be a base set of instructions from the company that you're using the IDE from, but it's best to have your own provided examples for both a list of IDEs as well as rules for AI on my website. We're just going to yap. We want to talk to AI about what we're doing. Use chat GPT and use their voice mode to just talk about the application. You want to do is discuss the main features, concepts, high level points, and pretty much give the AI a good understanding on what you want to build. Once you're done with that and you've yapped away, you want to ask the AI to provide you with a summary. The reason the summary is important is because it's an overview of what you just discussed, what you actually want to build. You're going to take that summary and you're going to give it a new prompt. So you're going to give it a new set of instructions. I've also got these on my website and this is providing it a prompt for a PRD. It is a product requirements document. This is usually something in software companies provided to engineers by a product manager. High level overview of the application architecture, some of the technologies that might be used, perhaps some design elements or wireframes, user stories or user flow. And it gives you a really good idea as an engineer, what exactly the specific required output is. Because you're not maybe a classically trained engineer, you're going to do is take this PRD output and provide it to the AI. Because the language model is doing the coding, it's going to be your engineer. So within your AI IDE, you're going to take your PRD, you're going to send it or put it into the chat environment or save it as a file, as an HTML or a markdown within that IDE. And you're going to give it to the agent and say, generate an application or scaffold out an application based on this PRD. This is the expected output. So now it has a primary goal or target agent or the LLM is going to begin to generate code and actually write out application for you within the AI IDE. We'll create a series of files for you. Next, you're going to enter the most painful part, building and debugging. The good news is that you do all of this in English. The bad news is you need to still interact with your application. You need to test things. You need to build out features. You need to build out new features you and the AI hadn't thought of to make it work. It's a big process. Some of the ways that I find effective to actually move forward if I'm getting stuck or I'm debugging and I just can't figure out how to solve it, is I will ask AI to take a summary of everything we've done so far, and then I'll copy it and paste it into O1 reasoning model, ask it to reflect on the problem and what the potential solution might be. You want to ask AI to teach you as you're going through this. So this could be a part of your system prompt. You want the AI to help you learn the software development process so that step by step you can actually come out of this being a somewhat proficient developer or programmer. Lastly, I want to remind you, your goal here is to create a production ready build. If you have any questions about the development process and you want some help, I've created a Telegram channel where people can talk about AI development technology. Those are the seven steps to getting started.