Pi Agent: мінімалістичний AI-агент для досвідчених користувачів

Pi might be the single most underrated AI tool out there. Now, what even is PI agent? This is a harness that is powering many of the most popular projects out there, including OpenClaw. And Pi is the agent that all of the power users are currently running because it is super customizable. So, in this video, I'll explain why the PI agent is so important, how to actually set it up, and how to make it more powerful than any other agent out there. First off, let's talk about the core idea behind PI Agent. The main idea was to create the most minimal version of an AI agent possible. And it means that it literally only has four built-in tools: read, write, edit, and bash, which means there is zero bloat. The system prompt is super minimal. There's zero extra nonsense, nothing. This lets you customize your PI agent for any use case and to make it truly yours. Now, as I mentioned earlier, OpenClaw, which is the most popular AI agent in the world, is literally built on top of PI. Yet, most people have never even heard of the Pi agent. Now, while the public is completely oblivious to Pi, the real players know what's going on. OpenClaw is for the masses, right? It's for everybody. It's easy to use, easy to get started. But Pi is for those who want to be on the cutting edge. Toby Lutk, the CEO and founder of Shopify, who's also worth $10 billion, is using Pi as his main assistant, and he used it to make Shopify 53% faster overnight and is his main coding agent, his main personal agent. Mark Andre, the inventor of the Mosaic web browser, one of the most important people in terms of all of technology, one of the most successful investors in the VC space, called it a top 10 software breakthrough in history. So, PI agent is a big deal. Seriously, you should block out some time to master P. If you actually sit down for 10 minutes and follow the steps I'm about to show you in this video, you'll be ahead of 99.9% of people in AI. However, you should be careful because Pi always runs in YOLO mode. Meaning there's no permission problems, no checks, no guard rails. It just does whatever it wants to do. As Mario Zechner, the founder of PI agent puts it, if it can write and run code, security prompts are pointless. So none of this like you know prompt defensibility is built into PI. It is just pure power to the agent. And this means that the PI agent is not for beginners. It's for those who want to be on the cutting edge of AI and who really are serious about building their own personal agent that is suited for them and that is best made for whatever work they want to do. Now, radical minimalism is the philosophy behind the Pi agent. For example, the entire system prompt is less than 1,000 tokens. If you don't understand how insane that is, just look at, for example, Hermes agent. It has over like 12,000 tokens default system prompts because of all the different tools and the self-building skills. one case, super minimal, super lean, and this is the beauty of Pi. It only has four tools and this is all the agent needs to do anything you can imagine. That's what makes it customizable and that's what attracts the most powerful builders, entrepreneurs, billionaires, programmers to use PI as their default AI agent. Now, let's talk about the first tool, the read tool, right? It's simple but beautiful. This tool lets the agent open and read any file on your computer or a specific folder if you're running it in a specific folder. And this is how the agent learns. You can read a log, check a config, inspect any text file all through the terminal. So the read tool is yet another tool it can run in the terminal. And without the read tool, the agent would be blind. It wouldn't know what's on your machine. It wouldn't know what's installed or what's broken in your code. It simply would operate blind. The second tool is the write tool. Pretty self-explanatory. Lets the agent create new files on disk. And this is how the agent gets work done. For example, create a rest API with authentication. And it writes every file, root, middleware, config, all of it. Now you might be thinking, "Okay, David, but given how simple PI agent is, can it even do anything?" Well, let me show you some examples of what people are building with it, right? So, this guy wasn't happy with the UI of the desktop apps. So, he used PI agent to just start designing custom UIs for a lot of the different apps he was using. And again, this is an entire custom app powered by Pi with its terminal UI. Now, to better understand why so many people are building on top of Pi, let me show you the back end because Pi basically gives you the toolkit to build on top of AI agents. You can use it as a complete agent for sure and that's how I'm using it myself. But if you want to build applications that are agent first, you can use pi as the toolkit. So first it gives you the terminal UI library which is named pi-tui. Gives you markdown display, multilan editor, loading spinners, flicker-free screen updates, something that clo still doesn't have to this day. And yeah, this is basically the visual layer that makes everything look good in the terminal. Next, the pi coding agent. This is what most people use when they run pi in the terminal, the coding agent. And again, I'm going to show you how to set that up later in the video as well as how it compares to cloth code and open cloth. Then we have the PI agent core. This is the brain allows you to define custom tools. The agent calls the LLM, executes the tools, feeds results back and repeats until this is done. This is just pure agent logic with no opinions about what tools you use. And the fourth thing that PI gives you is the pi- which is the unified interface for any LLM whether that's envelopic, openi, google or anything else all through a single API. So these are the four reasons why so many people are building their apps on top of Pi. Let me show you another example. This guy literally built his own full deep research agent under five minutes built on top of Pi. And keep in mind stuff like this is what Perplexity is charging $200 a month for. Okay. The third tool inside of PI agent is the edit tool. And this allows Pi to make changes to existing files. Now without the edit tool, Pi could only create new files. And of course it needs a way to edit them. That's why this tool exists. For example, fix the login bug on line 42 and it touches that line, maybe a couple lines before or after, but it doesn't refactor the whole file. It doesn't need to rewrite hundreds of lines of code. It only edits what it needs to edit. Now, tool number four is the bash tool, and this is the most powerful tool of all. It lets PI run terminal commands, installing packages, running tests, starting servers, checking get status, anything that you would type in the terminal. So while the three other tools handle files, bash handles everything else. And this is why PI agent is so insanely strong with just four tools. Now you might be thinking, okay, David, but why not open claw? Some of you might already have open claw setup. So why would you be trying by agent? Well, first of all is the radical minimalism philosophy. Open claw wraps a massive product layer on top, right? So if you're a complete beginner, maybe that's good. But if you want to customize it and make it really your own or built top off of it, it has a bunch of bloat. Also with Pi, you have full transparency and control. Everything is exposed. Nothing is hidden behind abstraction layers. With Pi, you control the whole package, the whole agent, and there is no hidden stuff, no hidden promps that are hard to change. None of that. Using Pi lets you build your own agent, not someone else's. You know, if you use Open Claw, you're like taking on all of Peter Steinberger's preferences, which might be good, but it not might be good. Real quick, if you want to master AI coding in 3 weeks, then make sure to join the new society. We just dropped a brand new training where in 3 weeks, anybody, literally complete beginner, can go from starting out, which I'm going to show you how that looks like, to mastering AI coding to the point where you're actually able to build anything in a matter of three weeks or less. So, here are the new modules, and let me just show you the outline for week one. This is all the things you learn just in week one. And I really do mean that. This is the simplest way to learn AI coding. So if you want to acquire the skill of being able to build absolutely anything you can imagine and turning that into custom software, make sure to join the New Society. In fact, for the next 6 days, it is for a decreased price of $37 a month. This is the lowest it's ever been in the history of New Society. It's going up to $77 a month 6 days from now. So again, if you want to master coding in three weeks or less, make sure to join the New Society. It's going to be linked below the video. And now let me show you how to set up your very own PI agent. So the official website is pi.deaf and it's very minimal like the entire pi agent and you literally have this one-click command. So go to pi.dev pi.dev and here just copy this command open your terminal paste it in and hit enter. This will install pi globally on your entire machine. To check if you successfully installed pi just type in pi-v and you should see a specific version. However, there's an issue. If you type in pi, you can see that it starts the agent for me, but for you, it will not start the agent because you haven't set the provider and you haven't set the API key. And these are the two hardest things about pi. You have to realize this is a trade-off, right? Open claw makes this simpler, but it also comes with a lot of bloat. So, anytime you have like a ease of use, it also includes a lot of unnecessary tokens. That's why if you use open claw, every single time you message it, you're sending like 12 to 16,000 tokens and you're paying for that. If you're using opus 4.6, that's not cheap. So you're paying Sam Altman or Dario Ammoday or you know these big companies every time just because Open Claw is bloated. With Pi that's not the case. It's very minimal, very optimized. However, the setup is a bit more complex. So I'm going to show you how to do that. So if you can stick with me for the next 60 seconds, you're going to be fine. So when you when you just installed Pi, right, you check your version. Okay, it's showing good. We need to do two things. We need to set the default provider so that you don't have to run it because otherwise you need to do like something like this, right? every single time set custom parameters anytime you want to run pi agent which is much less convenient than just typing two letters pi. So I want to make it as simple for you as possible where you have my setup right my setup is beautiful I type in pi it loads all the skills and I can just say hey and it's instantly loaded up and knows what's happening and I have open 4.6 selected and it's running with open router. I can easily switch models and I don't have to export my API key and every single terminal session. So this is a beautiful setup, beautifully convenient, super easy, amazing. I'm going to show you how to get that. So when you install Pi out of the box, it makes you it forces you to do this every single time in every single terminal session. You have to export your API key and then you have to do two parameters. The provider set it to open router and then the model set it to model because it doesn't remember that. So I'm going to show you how to make pi remember these preferences. And in fact, I turned all of this into a single prompt which I'm going to link below the video as well. And this is a GitHub gist that you can just copy and give it to your cloud code or any other agent that you already have installed. So it does these steps for you. Or you can just follow along and do these steps, right? So first you need to create the pi config directory. And here's a copy paste terminal command. Then you need to store your open router API key in the off.json file. So let me show you how to get your open router API key. First go to open router.ai and I'm going to link that below the video. Then top right, you need to either log in or create an account. Super easy. It's completely free. Takes like 20 seconds. Next, go to settings and on the left, you can see on the left, click on credits right here. And you need to charge up some credits. Now, you don't need to do $167 like me. I'm spending thousands of dollars a month on open router because this is what's powering all of my agents. Just charge up like five bucks. Charge up 10 bucks. And the beauty of API pricing is that you only pay for the tokens you use. So, I would recommend you charge up a couple of dollars here. But then once you do that, go to the top left and click on API keys. Here is where we're going to get our key. So, on the top right, you can see this big blue button, create. Click on that and type in PI agent YouTube subscribe. If you're watching this and if you want to see more videos on advanced AI tools like the PI agent, make sure to subscribe. Me and my team spend a lot of time researching, scrolling Twitter, testing these different tools to figure out what is worth using and what isn't, right? So for every video you see, there's probably like five to 10 different AI tools that we decided to scrap because they were actually trash. So if you want to see more about this type of stuff where most people only talk about Open Claw, about the most mainstream AI agents, they don't tell you about the the real cutting edge of AI. So if you want more of this stuff in your YouTube feed so you become smarter yourself. And if you want to support me in a completely free way, make sure to subscribe. Again, it's a completely free and it's below the video. And I think right now only 25% of people are subscribed. So it would actually mean a lot if we can get that at least to 30%. Thank you. Now the next step is optional, but I like to set a credit limit on my API keys. So just put it like 10 bucks, 20 bucks in case it gets leaked. It's good to have a credit limit there and then copy your API key. Boom. And back to the gist. And again, this is linked below the video so you can easily follow it. The best thing, honestly, like the best thing is using AI agents to set up other AI agents. I'm not even joking. I used the if you watch my very first video on cloudbot because that's what it was named back then. I used open code to debug it in the VPS right because it was so buggy and there was so many things going wrong. And this isn't just a meme. This is actually the way. So what I would highly recommend you to do is like clangerously skip permissions. Basically running cloth code in fully yellow mode and then having cloud code help us set up the pi agent. Right. And again, you can do the same with Codex or any other agent you already have. Or if you don't have any agent, just follow these steps right here. That's why I laid it out for you in a single spot. So you don't have to spend many many minutes figuring it out because that's what I had to do when I was first using PI agent. I was like, why do I have to keep setting the provider every time? Why do I have to keep exporting the API key every time? Is there a better way? Yes, there is. Just follow these steps. So anyways, if you are using uh clot code, you can just click on raw here, copy the entire thing, switch to clot code and say like I want you to execute these steps or if you want to do a safer way because we're running in dangerously skip permission say like do not make any changes yet just verify if the following steps have been implemented for my PI agent setup steps. And then I'm going to paste in the big gist. And then I'm going to do again do not make any changes yet. Just investigate this and give me a super concise answer. There we go. So I'm telling you guys this is the future. You're going to have AI agents. And just like that with the fast mode I couldn't even finish a single sentence and we have cloth code already reporting it. All steps have been implemented. Okay. So as you can see I've done them. That's why how I learned what to do. But if you're doing this for the first time it it will not be implemented. And I would highly recommend you use cloth code, open code, code codex to help you set this up. Now, one thing that you should change is the default model, right? So here it defaults to set, but obviously we want to use something better. Ideally, Opus 4.6 or if you want to save some money, you can go with like um you know Kik 2.5 or you can do the Miniax M2.7 is also great. The Chinese models are insane at the price efficiency. If you want a really good power, but you want to use American, you know, hosted models, GBT 5.4 mini is great. Honestly, surprising and yeah, nearly 100 tokens per second. Or if you want some balance, I would recommend Sonet 4.6 from Enthropic. You know, this is a really solid model that isn't as expensive as Opus, but it's still very strong. But, uh, you know, I would say like, okay, Opus, boom. And then we have the terminal say like is the model boom. You just put the exact slack. Let me check what the actual model ID is. No. Say like is our model in the config this. If clothco is going in the wrong direction, you can just press escape to interrupt it. But yes, our model is set to cloth opus 4.6. So you know do not use sonet 4 as it's here. Just use at least set 4.6. Anyways, let's continue with the pi agent setup. If you did this, you know, I can do ctrl c to kill clo code. You should be able to just type in PI and PI agent comes up. It's like, hey, who are you? Hey David, what's up? I'm PI agent. Powerful extensible AI coding agent. I'll help you by reading files, executing commands, executing code, running new files, and running specialized skills. What do you need? So, we're talking to PI agent right now. Now, you might notice here it says based on the agents.md file. But by default, PI doesn't come with agents. MD file. And this is actually the very first thing you should set up. This is the global system prompt that will influence every single session you do with the PI agent. And I'm going to show you how to set it up. So I'm going to do Ctrl + C to kill the terminal to kill that session of Pi. I'm going to do up arrow, which allows you to cycle between recently ran commands up and down arrows in the terminal. Pro tip. I'm going to again relaunch cloth code with dangerously skip permissions. And I'm going to have cloud code debug this. So here is the say check if we have the agents.mmd file for pi agent set up. Check. Boom. Here's the steps. Check. Do not make any changes. Just answer in short. This is literally how you do it, guys. There is no secrets. You use an AI agent to help you set up another agent. Yeah, you have it set up. Okay, so I have it set up because I I did it. But this is the location, right? global at.py/ aent/agents.mmd exists. So as you can see uh it has some info about me and also I have per project instructions five different repos have agents.mmd files. So if you don't have agents.mmd file this is the location where it needs to be created.py/ py/ aent/agentd. And again, what you could literally do is like screenshot something like this. You know, literally take a screenshot of the video right now, go into your close code, paste it in with controlV, not command V, and say, make sure to create this file and put in basic instructions and then obviously you can customize it, right? The whole point of the agents. MV file is that these instructions are custom to what you want. Like use TypeScript unless told otherwise. Don't ask for permissions, just do it. read well with pi that's ne not necessary but you get the point right you basically put in your own preferences your own things like okay maybe you prefer responses in German instead of English maybe you prefer longer sentences maybe you prefer all the lower caps or someone who's more emotional and supportive or like more cold more robotic answers everybody has different preferences and that's exactly the whole point of pi so anyways once you do that we can finally start testing pi agent and we can finally start having it self-change and self-improve because it can do this to the point where you would not believe like cloth code cannot change its own UI codeex cannot change its own UI pi agent absolutely can okay so I prepared a prompt which feel free to copy that basically create a matrix style theme for pi and this will showcase the self updating properties of pi where it can uh update itself and redesign the user interface so look at this it's reading about itself uh it's figuring out what to do and then it's uh going to replace some of the config and user interface and again I want to stress this in cloud code you could absolutely not do this cloud code doesn't let you self update the UI right anthropic decides what the UI looks like same with codex openi decides what the UI looks like but with uh pi agent the whole point is to build your own agent so if you want to change how the terminal user interface looks like It absolutely is possible. And once those changes are applied, we're going to do one command / reload, which I love this command. This should be in every other agent where like instead of having to, you know, start a new terminal or start a new session, you just do slash reload and it u reloads the agent and all the settings are applied. So let's see how this matrix style theme will turn out. And by the way, below the chat input field, you can see a lot of useful information such as the percentage of the context window, the current cost of the session, and obviously the model you're using, and the reasoning effort in a very clean way that again, I wish every agent adopted this just observability on what the agent is doing, how expensive the current session is. And again, we're using Opus 4.6 like this is like the most expensive model out there. Um, I would recommend you probably start with Sonet 4.6 X. If you want something something like balance and if you want something cheaper, go with GPT 5.4 mini or one of the newer Chinese models. Or actually, you can just probably do the Gemma 4 if you want. This is a small model though. GM 5.1 or Gro 420 is actually also good. We could like surprisingly good to be fair. Uh that's why Elon retweeted me. But let's see. So, okay, it's done. Let's do reload. Let's see how our new theme looks like. Boom. Look at this. We just completely changed the user interface of PI agent. Like you could never do this in cloth code. You could never do this in CEX. But with Pi, this was possible with just a single prompt. And again, obviously this is more of like an animation that's like cool to have. But if you have even the slightest bit of imagination, you could see how this could be used. Like you could update Pi to your own preferences. change user interface, change the tools, give it new skills, change the system problem. It it doesn't come with any like bias, any built-in opinions. Right? So, this is just the very first thing. Let's take it a step further. So, the second thing I'm say, let's continue improving this, build an extension that shows hacker style system. So, we're going to basically have it showcase stuff about my current system, CPU, RAM, uh, GPU load, and anything else that might be useful while we're using an AI agent. And we're going to have PI self update itself. And then once that's done, just slash reload to apply the changes. Okay, it's trying to selfreload. I don't know if that's possible. I'll do the reload myself. Reload. Boom. And there we go. Look at this. So we have the system a monitor with information about my system right now. So you can you know if you want to you can see your RAM your load temperature disk as it's running right. So this is like real data about my MacBook as you're using your AI agent. And again obviously this use cases aren't probably relevant for all of you and obviously like the animation is super silly and not really useful but the main point stands I just did this in two prompts. Like what other agent allows you to do this? Pi comes super minimal, just four tools, 1,000 tokens as the prompt, letting you build itself into whatever you want. And even if that's, you know, not a silly hacker theme, you can see that it's very easy to update Pi in a way that you want and to make the UI feel and custom to anything, any use case you're doing. And I might make a video like something like eight use cases for PI agent to really go through what this could be used for. But for now, use your imagination, right? And by the way, if you want to just undo this, you know, if you don't want to have these themes, all of this is stored in the slash extensions and themes folders, right? So if you want to delete these, just send this to Pi and it can revert these changes. So anytime you create a theme or an extension that you don't like, you can just deactivate it by having PI agent remove it. And now if you do reload, we should have the default clean interface. There we go. So don't be scared that you're going to mess something up. If you created an extension or a theme that you don't like, you can always revert it. The next feature inside of each and I want to show you is the slash tree. This allows you to create a new session tree. So for example, if you have a feature and something breaks mid during that feature, right? maybe some tool or you want to create change the theme or add a new skill. You would start a tree and again imagine a tree where it's growing on a single trunk and then you branch off. Instead of like polluting the main branch, you can solve that problem in a separate branch. So when you type / tree, you can see that it opens this interface where you can select from which message you want to branch off of. So either the light latest one or maybe this one. And then you hit enter and you created the tree. It says you're already at this point. So obviously you need to go uh back let's say here and every step is not only your messages but also the assistant. So that's you know PI agent is the assistant user assistant is kind of the standard set by open AI for how these models are trained in RHF. So user that's always going to be you assistant that's the AI agent. So let's say I want to go here and either summarize or no summary or summarize of custom prompts. I'm going to do no summary and boom, we are in a tree now and we can continue here and work on something else without affecting the main branch. So here's the easy way to understand the slash tree command. It's like undo and branching for AI conversations. So most AI chatbots and agents are a straight line where like you can go back and forth but it's all on the same straight line. So if you want to, you know, make changes, you either polluting the context window or you have to tell the agent something that it doesn't need to know. With in pi it's a tree. So for example again you're building a feature and something else breaks instead of derailing the main conversation talking about debugging that other bug you tree / tree back branch off fix the bug and then jump back to the main branch. Pi just summarizes what happened in the side branch super context summary. So the context is in the main branch of the tree. The next essential command inside of pi agent is the slashfork. So if we jump back in can do slashfork and this creates a new fork from the previous message. So again here obviously this is a super short conversation but you can fork off into a completely different session right so within a single session everything is a tree but you can have different sessions. So if you want to fork off at any point and this is the same as forking a GitHub repo you know you have a repo that that is good and you're like I don't like the direction where this is going this project is going but I really like the foundation. I really like what's built in here. You fork it, right? This is what Rue code did with client. So if you're not aware of this, client is like a popular open source coding agent. And then came Rode which cloned the client repo. And after that came kilo code which cloned the R code repo to continue improving upon upon that, right? And it's the same within the context of your PI conversations. Another useful command is the - C parameter. So if you want to continue in the same session in the last session that you started you just do C. So again if you if I kill pi it's like boom. If you do pi obviously it's a new session right and there's nothing here. You can see in the bottom left corner 0%. However if you want to continue in the last session you have to do pi- c and you can see it opens the last session that we actually talked about. So this is the most most used shortcut or command I use is literally pi-t it's it's a single parameter two characters and it means continue in the last conversation. So you know sometimes you want to start a fresh session with zero context in there which is completely fine. Sometimes you want to continue where you left off so you just do C. And by the way all of the sessions are autosaved to the py directory/ aent/ sessions. And if you want to pick a specific session manually, just type in pi-r or d-res and you can browse all of the past sessions and pick one of them. Also, if you go to pi.dev/packages, this is where you can find a single place of all of the most popular packages for PI agent. And this could be extensions, skills, themes, and anything else from the open source community. So that by default, it sort sorted by most downloads. And you can see that pi pom pom is the most popular somehow. Then pi sub aents extension for delegating task to sub aents with chains parallel execution and ti clarification. So yeah, some of these things are built into open claw and hermes agent by default. But if you don't need them, they're just bloating your system and your and your prompts. Then you have a latch key tool that injects API credentials into curl requests. Then you have task plane agent orchestration for pi parallel task execution with checkpoint discipline. Let's check this one out. So this is like a paperclip competitor basically to like designed for PI agent. Look at this orchestrator dashboard lanes and tasks looks very interesting. So yeah again pi.dev/p packages if you want to see some of the most popular packages for the pi agent. This one's also interesting. Pi studio extension. Okay screenshot to paint studio for pi agent. Look at this guys. I might need to make a full video on some of these most interesting packages. Again, if you want to see that below, if you want me to make more content on the Pi agent, comment below and make sure to subscribe because if I see a lot of people subscribing from this video, that will be a very strong signal that hey, you guys clearly want to see more videos on Pi. It is another Pi extension by design deck, a tool for visual design decision-m. So, this is a opens a multi-slide deck in your browser with live previews of each option. Compare, pick and generate more options. So when when like designing and and building it shows you different user interfaces, different color. That's crazy. Okay, let's look at the video. Imagine demonstr UI and style choices for the web app. All right, so first you select what color you like or you can generate another option if you don't like any of these. Generate a pink one. Okay, there it is. You can select the pink one and then typography. Damn. Okay, navigation pattern. This is the future of building, guys. Instead of having to like generate everything from scratch from your brain, you select the options and like your taste is going to be refined. And if none of these options are good, then uh wow, this is actually OP. Okay, now it's doing a design deck. My bad. Security model, dashboard, aesthetic, data visualization style, and you just select the options that you want and the app is built based on that. Really cool. Really, really cool. Oh, and one more thing. Don't be surprised if you see the domain shitty coding agent.ai. This is a joke from the founders of pi.dev. At first, I thought this is a fishing domain, but it turns out to be a legit domain, right? So, still, I would recommend you go to pi.dev, if you want to install pi agent, but uh if you see the shitty coding agent.AI domain, it's not a fishing link, so don't be scared. That being said, if you are serious about AI and you want to master AI coding, then make sure to join the new society. In the first three weeks, you will learn how to build anything with AI, even if you're a complete beginner. And even if you've never built anything before and haven't used a single coding agent, look, you can look at this module by module outline. Every single module, you learn something new and we build on top of the last one. And it's like one, two, three minute modules. Anybody can follow this with the resources telling you how to set this up, you know, telling you exactly what to do, how to use these agents and uh how to deploy your app, right? So that's week one. Week two, we go even further. Here's a plan for week two. How to improve um cloth code, breaking some of the limiting beliefs, basics of a terminal, everything you need to build custom software just by typing plain English. So if you want to master a coding in three weeks, make sure to join the society. And again, the price is the best that it's ever been. We're revamping and releasing a lot of new content and and we're getting close to releasing that. In fact, in 6 days from now, the price is going back up to $77 a month. So, this is literally the best time to join. So, if you want to master the coding in three weeks, click the link below and join the new society right