Transcription
If you feel like you've hit a wall with clawed code, it's probably not because you're using it wrong. It's because you have no roadmap for progress. And that's really not your fault. Because this is an extremely confusing space with so many new features and so many new tools being thrown at you every single day. It becomes impossible to know what should you focus on to actually get better at using this insane tool.
But today, I'm going to help you fix that problem. Because after hundreds and hundreds of hours inside of Cloud Code, I'm confident in saying there are really six levels of progression when it comes to this tool. And so in today's video, I'm going to break down the six levels of Claude Code so you not only know where it is you stand, but what you need to do to get to the next level.
Now, before we kick this off, a quick word from our sponsor, me. I just released my Claude Code Masterclass. There is a link to it in the pinned comments and send Chase AI Plus and I take you from zero to AI dev no matter your experience level. So if you're looking for some extra help to get all the way to level six, that's where you can find it. Also in the description is the free Chase AI school. So tons of again free resources if you're just getting started. So make sure to check that out.
So level one of Claude Code is the prompt engineer. Now kind of funny when you think about it cuz a few years ago this was like a legitimate job title. Like that was the upper echelon of AI practitioner was a prompt engineer and now a couple years later it is the lowest of table stakes in terms of how we interact with agentic coders. Now what does it mean to be stuck at this level and why should you care? Well, you should care because the outputs are the very definition of mediocre, right? This is where we get generic AI slop. And the reason you're getting generic AI slop is because your relationship at this stage is very one way. You treat Claude Code as a blunt instrument, as a simple tool. It is not a collaboration and it is not a partner. And because of that, you hop into the terminal and you just start giving it instructions, things like build me my AI agency website. And it might even be more sophisticated with that. You might even have a PRD, an actual list of features, but you're not asking for feedback. You're telling it to do things right off the rip. And because of that, at this level, you get stuck in the trap of the regression to the mean. No matter how good you think your prompt is, if you show up just treating Claude Code like a tool and not a collaborator, your plan is going to have holes in it. And if it has holes in it, Claude's going to fill those holes, but it's going to fill them at just sort of an average output at the mean because that's how AI systems work. And that average, by and large, is AI slop. That's why most websites made by AI coding agents all have the same purple gradient and the same exact font and the same exact icons, right? Because it's the regression to the mean. And this is where most people are stuck.
So, how do we get out of this trap? Well, there's a few skills we need to master. First and foremost, obviously, we need to learn how to write clear, specific prompts. Now, in our example, at the most base level, what are we trying to do with our AI agency website? What is our actual outcome? Right? We don't want to just build a website for the sake of it. What do you want it to do? Are we trying to drive potential customers to a newsletter, to an email list? Are we trying to show them potential services we provide, right? What is the final outcome you are shooting for, right? And then what are the sub features that go along with that? That's the most basic level of prompting that you have to master. Beyond that, you need to start learning how to read and evaluate Claude's output. Right? For us to call something AI slop, we have to actually know what AI slop looks like. And better yet, we need to know what right looks like. If you're trying to build an AI agency website that works, well, do you know what an AI agency website that works actually looks like? Sounds simple, sounds stupid, but you would be surprised at how many people try to build things that they have no idea what the end goal should look like. And lastly, this is where we need to learn some basic terminal literacy. What does it mean when I'm on bypass permissions on and I've started Claude Code in a dangerous state? What does it mean when it's asking me to do specific bash commands? What is a dependency? Right? You don't need to know how to code at any point, but you do at this stage need to slowly start learning the language and the nomenclature of a software engineer. And at the end of it, the true unlock is that you stop commanding Claude Code to do things and instead you start asking it to plan with you. You turn it into a collaborator which moves us onto level two.
So level two of Claude Code is the planner and this is where our relationship starts to become a back and forth with this tool. It is no longer one way and the simplest way to get to this level is to start using plan mode. Now why do I say that? It seems like such a simplistic thing. Well because when we are in plan mode it's going to force us to collaborate with it with questions. So in plan mode when I say hey let's build my AI agency website for Chase AI Plus it is going to come back with a series of questions to figure out what it is I am actually shooting for and it's also going to have some recommended steps as well. Now, in tandem with plan mode is the idea that we are not just shifting our mindset, but we're shifting how we talk to Claude Code. Because like I've said before, one of the biggest issues with Claude Code, especially if you come from a non-technical background, is that you don't know what you don't know. And that is dangerous really, especially when you start dealing with things like other people's data, right? If you don't come from a technical background, you have no idea what you're supposed to do. And not only that, you probably don't even know what questions to ask. And as good as Claude Code is, right? Even as good it is to ask me some of these questions like hey how do we want to do this what do you want to do with tech stack what is our goal in design there are going to be times where you need to nudge it in the right direction and this is the level where we begin to cultivate that skill of asking Claude Code the right things and truth be told asking the right thing often times is something as broad as saying hey what am I missing here? Right? Something I always ask Claude Code is like what are the unintended consequences? What have I not thought about? Or another good one is like hey what would someone who is an expert, you know, web designer, someone who's an expert in database construction, what would they be thinking about in this place? Now, Claude Code, Opus 4.6 especially, is getting really, really good at this at like being able to intuit it what path you should go down. But let's be honest, you are still in the driver's seat here. So, when you tell Claude Code to do certain things, it's going to go down a certain path. But if you ask it these leading questions about, hey, what path is better? What do you recommend? Right? It will kind of realign into the right direction which is very important.
So to recap, what skills do you need to master at this stage to move on to level three? Well, you need to master the use of plan mode, which honestly is very simple. Big thing, collaborative questioning, right? What am I not thinking about? And then letting Claude Code push back on your ideas. Again, this goes to the questioning side, like tell Claude Code to be adversarial, right? You can even get kind of complicated this with this. And when you have a build code, say, "Hey, I want you to spawn essentially like a, you know, devil's advocate sub agent that's taking a look at everything we're building and telling us what's wrong with it or if there's a better path." Right? That's a great thing to do. But the trap at this stage, and this goes into the unlock and going into level three, is at a certain point, it's not enough to just give it words, right? It's not enough just to ask questions. We need to start introducing specific context often in the terms of files or screenshots or examples. So to go from level two the planner to level three the context engineer we need to start thinking as we interact with Claude Code is what outside information does it need to give me the outcome I want.
Now level three and level four is where I believe most of you probably reside. Level one and two are definitely the beginner stages. Maybe you just started Claude Code, but these levels are where a lot of us get stuck. And I think that's partly because this is the stage where you become aware of everything that you can add on to Claude Code and all the things you can bring into this ecosystem. And it becomes very confusing and hard to determine what you're supposed to keep and what you're supposed to cut and what's the right tool for the job.
The number one skill to master here out of all of them is the idea of context window management. This will determine how effective Opus 4.6 actually works for you and what sort of outputs you get because of that. And that's because we always, always, always are in a fight against context rot. Now, we've talked about context rot a ton on this channel, so you know what it is. It's the idea that the more the context window fills up, the worse the AI model performs. That's exactly what this graph shows. Sure, this is on older models like Sonnet 4 and GPT4.1, but this remains the case today. This has not been solved. Now, a very quick recap on the context window and how we deal with this situation. Remember, we can always do forward/context to see what our context window is at any given moment inside of Claude Code. Tokens are the currency of all AI system interactions, right? And Claude Code has a budget of 200,000 tokens at any one time. Now, context rot, the idea that our efficiency goes down over time. It is a pretty sudden fall-off and it's somewhere between like the 50 to 60% range for Claude Code. So, 100,000, 120,000 tokens. This number does not change if the context window gets bigger. Right? Opus 4.6. Oh, they have a 1 million context window now. Does that mean is it 500,000 tokens in that scenario? No. No context window. This whole context rot thing still applies at like 100,000 tokens whether you have a million token window or a 4 million token window or 150,000 token window. Right? So just increasing the window doesn't solve this problem and like I said everything we do adds to the context the system prompt the system tools skills MCP tools memory files messages etc etc etc. So obviously what is important to do here? Well, it's important to keep tabs on this window so when we get near the dead zone, right? That last 50, that last 40% event, we reset it. And to reset the window is relatively simple. We're just going to do forward slash clear. Now, you have other options like compact, right? Compact will automatically essentially summarize your conversation, restart the next window, and it's like you gave the next session of Claude Code a prompt saying, "Hey, here's what I was talking about." I don't suggest using compact in nine out of 10 cases or 99 out of 100. Well, why is that? Well, because you're expecting Claude Code to pick the right things to bring into the next context window that it should. That's not always the case. Secondly, by virtue of how Claude Code works, it's going to pick the right context, right? It has access to your entire codebase. It's smart enough to realize, hey, I started a new session. It's asking me to do A, B, or C. Well, let me go look up the correct files. Also, you can manually nudge it in the right direction if that's the case. And and if you are having such an important conversation that you feel like it must be brought into the next window, well just have Claude Code summarize it in the way you want and then manually paste it over. I think that's the best practice. Secondly, it is obviously rather tedious to always do forward slash context and see what's going on here. Much better is to create some sort of persistent status bar like you see down here below, right? Where it shows me the folder I'm working in, the model, and then my context window, which in this case is 19%. There's very easy ways to do this, and the simplest way is to just prompt Claude Code to create a persistent status bar like this for you. You could literally take a screenshot of this page, drop it in Claude Code, and say, "Hey, build that status bar for me. I want it to be global and I want always to be on." It's that easy. But this will give you way more insight so you know when you need to start being on your toes and ready to forward slash clear. Simply having that mastery of the context window will probably put you in the top half of Claude Code practitioners because most people don't do that. They just code code code code till they hit auto compact and they wonder why half of their session is spent getting poor outputs from Claude.
Now the rest of the skills to master this level all sort of play off the idea of context window management. Right? We want to use few-shot examples over extremely long instructions. We want to learn how to actually use Claude.md the proper way. Now, the Claude.md file was very much in vogue during the early months of Claude Code and you still see people talking about it. This is straight from Claude Code docs. It is a markdown file to add to your project root that Claude Code reads at the start of every single session. This sets the standards and the conventions for everything you're trying to do, right? So, if you want to code in a certain way and there's certain coding conventions and decisions you always want it to follow, well, you put it inside of Claude.md. And we've always had the ability to do forward slash init, right? If you were starting a project or even like a brownfield project, if you did forward slash init inside of Claude Code, it will essentially take a look at how you've written all the code, create the Claude.md file and say like, hey, here are the conventions based on what I've seen. However, we are now seeing studies done like this one last month where things like claw.md and agents.md are actually harmful to these coding agents. In fact, across multiple coding agents, they found that context files like that tend to reduce task success rates compared to providing no repository context while also increasing inference cost by over 20%. So what does that mean for us? It means the skill of knowing what context Claude can benefit from and what will only hurt it is one we need to master. And oftentimes the answer is Claude Code needs some context, not too much context. And oftentimes less is more. And it's at this point you will naturally begin to progress to level four where you then become exposed and begin seeing all the outside capabilities we can add to Claude Code. Things like MCP servers, things like frameworks, things like CLI tools, and we essentially become a kid in a candy store with all the tools at our disposal.
Now, I think the kid in the candy store is an apt analogy for level four because when you first get introduced to all these different MCP servers and all these plugins and all these frameworks like GSC and BMAD, you become enamored with the additional capabilities you are provided with. However, we need to be very clear about something and that's that capability does not equal performance. And that distinction is something many people never get to, right? And so they get stuck from hopping from framework to framework and they're constantly wondering what the best MCP server is. So they just install every single one they can think of and they spend all their time searching for plugins. And you'll see this repeated in level five when we look into skills, but it's this idea that I just need one more MCP, bro. Just one more framework, bro. And it's finally going to work out. Now this is totally natural because these MCPs and these tools we can add and these frameworks these really can boost the capabilities and the performance in many cases of Claude Code but we need to be surgical about it because the answer isn't all of them all the time it's some of them in specific use cases and the truth is especially when it comes to things like frameworks like GSD for example I love GSD I've done multiple videos on the GSD framework many of these frameworks I think are becoming less less and less needed over time. We're seeing many of the capabilities that made these frameworks so strong be repeated inside of Claude Code itself natively and nine times out of 10 the native implementation tends to have an advantage over taking some sort of third-party tool and layering it on top of Claude Code.
So if the answer to progress through level four is like okay I need surgical MCP selection depending on my project then the obvious question becomes okay how do I know the answer to that? Right? How do I actually know how to be surgical? How to go from all these plugins and all these MTPs to the right one? Well, there's two ways we're going to do that. The first is by understanding the building blocks of what it is we're actually building. Right? And the second one is asking Claude Code to explain how it did these things. And these two things are related, right? This is really and this is why a lot of people get stuck in this phase is like if you want to move away from vibe coder into like AI dev. These two things understanding the building blocks and asking Claude Code to explain how it did things are the path forward to doing that because most people when working with Claude Code do neither of those things, right? They simply even if they do something like plan mode, it's just next next. Sure, I'll take the recommended thing. Next, next. Like the only the most used button on their keyboard is the enter button. And that's a problem and it's actually kind of troubling because Opus 4.6 is so good. Claude Code is so strong. If you are just some monkey on the keyboard who just hits next, next enter enter. Yes, yes, yes. You will still be able to build a lot of stuff. You truly will because Opus is that good. But the problem occurs when we eventually need to do something that's more complex. And this goes back to my idea of like you don't know what you don't know. If you start dealing with complex projects, you are going to need to start understand at a fundamental level what the building blocks are. What is a front end? What is a backend? What is authentication? What are databases? How does security work? How do I deploy this? You don't need to know how to write any of those things. But you do need to know how all the building blocks come together. And again, you don't need a CS degree to do that. You literally just need to ask Claude Code why and how. Right? If Claude Code does something that you don't understand, whether it was a fix for something or a decision, you need to ask these questions. And one of the, frankly, the best use case for AI is the infinitely patient tutor, because I can ask it to explain something to me in 3,000 different ways, 3,000 different times until I understand that. But to do that does require a mentality shift. And that's something we've talked about throughout all the levels, right? You need to take an active approach and an active role in how you talk to Claude Code, right? If you are passive and you just let it do things and you don't care, well, you know what actually is your moat to be honest? Like why can't I replace you with anybody else and just throw them in your chair and have them do the same stuff? Question you need to ask. And these questions and understanding go beyond just like the philosophical idea of like you need to actually like take an active role and be kind of smart about it. You will screw stuff up if you do not understand this. Especially when you start installing MCPs like Superbase, right? If we're using something like Superbase and we're using databases and we're dealing with authentication users, what's going to happen if you leak people's data? What happens if you're trying to take on a client project that has like anything at all to do with privacy? Your answer to those questions if something goes wrong can't be Claude Code told me to do that and then have no further explanation beyond that. So it is imperative especially if you want to work in AI where money is involved and data is involved that you begin understanding how this all works and again that understanding has nothing to do with the code or understanding the code. It has everything to do with the theoretical conceptual view of how these things work and come together. That's it. And so I'll get off my soap box on that part and talk about okay cool get it tools outside complexity. This is the level where we equip Claude Code and I know how to go out into the internet and into space and find additional add-ons and things that work for the particular job. What's next? How do I move on? Well, this is where we move on to level five. And this is where we start working talking about like workflows, right? And the idea that at this point you're getting pretty sophisticated with how you use Claude Code. Well, you're probably at the space where you're doing the same thing over again over and over and over. Well, how do we like sort of productize that? Well, not even productize that. How do we put it into a nice neat little box so I don't have to repeat myself over and over and over again? And this is where we're going to go into the skills and how we can turn these different actions inside of Claude Code into workflows.
So, level five is all about skills, how to use them, how to personalize them, how to create our own so we can maximize our output. Now, just like before in level four, these skills, we can fall into the same trap when it came to things like MCPS and plugins and frameworks, right? We're a kid in the candy store. There's 10 million skills out there, and you are going to want to go nuts and install all of them. Now, luckily for us, skills are pretty lightweight because remember what exactly is a skill. Skills are just text prompts. That's it. It's just a way we tell Claude to do a specific thing in a specific manner. And what we're looking at right here is probably the most important skill of all, and that is the skill creator skill. This actually just got updated from Anthropic just a couple days ago. I did a full video on it. I'll put a link above. But if we want to supercharge skills, the best way we can do it is create our own. And the skill creator skill not only allows us to create our own skills, but it allows us to optimize skills we already have. We can run evals, we can do benchmarks, we can do all sorts of stuff. And I won't belabor the point in this video. Definitely check out the one above. But this is a massive, massive value add that should really change how we approach skills because with this new skill creator tool, skills, although they weren't entirely a black box before, again, they were just text. We could look at them. It was kind of a pain in the butt to determine is the skill I'm using optimized? Is this the best way to do it? And using the skill creator tool, we can now run actual tests. We can get evaluations. We can get hard data about is the way we're doing things the right way? Is it actually helping at all? Now, to add the skill creator tool, you can just do /plugin, and that actually brings you to uh the plugin marketplace for Claude Code. This is like a built-in thing. So, there's all sorts of ones again like Context 7, superpowers, GitHub. For the skill creator one, you would just search for skill creator and you can see right here it's installed.
So here at level five with skills, this is where we really turn Claude Code into our own. We can use the exact same MCPS and the exact same plugins in someone else. But with skills, we can make sure Claude always uses them in a specific way that aligns with our tastes or our branding or how our company does something. Now, the trap with skills is kind of what we talked about before, and it's that we just create too many. Remember, we don't load all of these into the system message at the beginning so it's not clogging up our context, but we do preload a list of all the skill names as well as a short description. But the more options available, the less likely Claude Code will actually trigger the correct skill when I want it to. Right? If it has a hundred options for skills and I kind of vaguely tell it to do something that should invoke the skill, what are the chances it uses it versus if I have five? Now, there are some ways to deal with this. First of all is you always have the ability to do forward slash, right? So, if I do forward slash and then the name of the skill, whether it's custom or something, then it will 100 invoke 100% invoke the skill every time. The second way to do it is when I create a skill, right? If I use skill creator, right, and I'm trying to create a specific skill, you know, hey, specific skill does X. Well, I don't always have to install my skills at the global level, right? Chances are I'm only going to use specific skills to do specific things in specific projects. Your skills can be project-based. Again, very simple thing to do, but it allows you to kind of go nuts with skills, but since they're all localized, we don't deal with the issues of, "Oh god, Claude Code has access to 10 million skills. How's he going to choose the correct one?" Right? So, again, we always want to use a scalpel. And also remember that skills can reference other skills. You can have like higher-order skills. So, I can create a skill creator that says, you know, use skill A and skill B in this order of A, B or C. Right? So skills can reference skills which can reference skills. It's skills all the way down, right? So you can kind of go crazy with this. The point is you can customize this as much as your heart desires. And with the skill creator skill, we can now actually have some confirmation if they even work or if they're optimized. And so the skills to master are skill authoring, right? Is my workflow can I actually articulate my workflow in a way that Claude Code can make sense of it? Chances are yes. Slash commands, workflow skills, like we talked about, skills of skills, and then skill creator really is the unlock at this level. And luckily for all us, this just came out a few days ago. And so, we're able to move on to level six.
Now, level six is going to be all about scaling Claude Code because everything we've talked about from level one through level five, this is all, you know, under the assumption that we're kind of just dealing with one instance of Claude Code. Well, what happens when I want to use a bunch of Claude Codes all at once, multiple sessions, doing multiple things, multiple work trees, sub agents, agent teams, right? That is where we move into really becoming a manager of Claude Code.
Level six is where you become an orchestrator of multiple Claude Code instances at the same time, like you are a manager. And there's really four levels to this and we'll go through each one in detail, but as an overview, level one is where we just have multiple sessions open at the same time in the same directory working on the same project. Level two is where we up the complexity and we start looking at Git work trees. Level three is where we automate that process a little bit and we have separate instances of Claude Code working, but it's via sub agents and they also are working via Git work trees. And then level four is what is still an experimental feature of Claude Code and that is agent teams.
Now that first version is pretty simplistic, right? I'm inside of Claude Code right now inside of the six levels project. If I open up another terminal and I run Claude Code again, I now have two instances of Claude Code running on the same project. Right? I can in theory get twice as much done and manually coordinate them to do different things. But we can up the ante here. What we have available to us is Git work trees. Now this is a little bit more technical. This is all diagrammed here. But think of it like this. Normally like in our last example, if we have multiple instances open inside of the same project and they're on the same work tree branch, aka exactly what we did here. Think of it think of it as having multiple Claude Codes, multiple human beings working on the same project at the same desk. And there may be times where they both need to edit things on the same piece of paper. This can cause conflicts potentially. Versus with work trees, imagine instead of all of our workers at the same desk, we instead separate them and they all get their own desk. They all get their copy of the same work. And at the end of the day, we merge their work. That's the theoretical idea of work trees. Now, to start up work trees, it's pretty simple. When we start Claude Code, we're just going to use the d-work tree command. So I'd open up my terminal and I would do something like Claude work tree and then it would be like front end, right? And then I would open another terminal and do Claude work tree backend or Claude work tree authentication, right? I could essentially divvy up all the tasks, give all these workers their own desk and have them work separately and not have to worry about conflicts. However, there is a level of complexity when it comes to merging all of that work together and making sure it actually fits. Now Claude Code can help you with that, right? You can tell it, hey, I finished up work on all the work trees. I want you to merge all this together. So, understand there there are workarounds. Claude Code can handle a lot of the hefty lifting here, hefty lifting here, but just get this is a little bit more of a technical path to go down.
Now, the third option is to just have Claude Code create those work trees for us as sub agents. So instead of agent A, B, and C being literal terminals you opened, we could have just that main Claude Code terminal that we're talking with and say, hey, I want you to spawn three sub agents in three different work trees and you know, I want one to work on authentication, one to do front end, and one to do tests. So all of this could kind of be abstracted away, and Claude Code can take command of the whole situation. The issue with this is it's not as obvious to you as to what's happening, right? Because the main Claude Code terminal is kind of doing everything for you. And so then that leaves us with number four, which is agent teams.
Now, as I said, this is an experimental thing. So, we need to actually change some settings in our settings.json folder to get this work. I'll put a link above or will be in the description as well to my agent teams video which goes into this in depth. But to get it working is pretty simple. Um, you literally can just copy this page or give the URL to this page. Just search up agent teams Claude Code documentation. Give it to Claude Code and it will edit the settings itself. So let me explain how agent teams work. This Claude Code thing up here, right? This is your imagine this is acting as your primary terminal. This is what you're interacting with. At this point you've enabled agent teams. You now want Claude Code to use agent teams to do three things for you. One, you want it to work on authentication on your website. Two, you want it to work on front end. And then three, we want it to work on payments, right? We need to set up Stripe or something. Okay, so there's three things we want it to do. And we want Claude Code to spawn sub agents for it. Now, if we say we wanted to use agent teams, what it's going to do is it's going to spawn these three sub agents, but these three sub agents can all talk to one another. Normal sub agents do not speak to one another. They kind of work in a vacuum, but in the team, you know, hence the team name, they all talk to one another, right? That way they make sure they aren't stepping on each other's toes. And we have conflicts like in our analogy they're all at the same desk. However, it also spawns this guy who's like a supervisor, right? To make sure everyone's doing the right thing. So, he also talks to everybody, right? He's making authentications, doing his thing, he's telling payments what needs to be done. So, he's kind of the ring leader and he reports to you, right? So, they essentially spawn this agent team spawns this middle manager who deals with all our little worker bees who are all talking one another to get you your output.
Now, as for when you should use agent teams, Claude Code spells it out exactly for us, right? Research and review, new modules or features, debugging, cross-layer coordination, really anything where parallel exploration adds real value. Now, the one downside with agent teams is because of that extra coordination overhead, like responding a session of Claude Code to literally act again as that middle manager, it does use significantly more tokens.
Now, to actually use agent teams, you have to use specific verbiage. And there's an example given here in the documentation where it says I literally want to create an agent team. Your nomenclature, whatever words you use, has to say something to that effect or else it won't do it. Now, when you're actually using agent teams, it will look something like this. So, I asked it to create an agent team for me. One for front-end authentication, another for a blog, right? And so, you'll see it say, "Hey, three agents have been launched." It will break down exactly what all what their actual tasks are for each teammate. And then here down at the bottom of the status page, you'll see all of those agents. So at main is like the actual terminal you're talking to. And then we have the three sub agents and they'll be color-coded. So O, blog, and front end. So I can hit the down arrow and I can kind of switch between them. If I hit enter, I can actually message the individual sub agent as well. So right now I'm talking to blog and I can say, "Hey, actually change a direction. I want to do A, B, or C." So you can either talk with your normal Claude Code and have it send the message downstream or you can go to them individually and actually get your hands dirty. But those are the four ways you can begin scaling up and exponentially increasing your outputs with Claude to Code. Right? We don't have to be in one session. We can have as many sessions going as you want. But this is definitely a practiced skill and being able to context switch even within the you know ecosystem of a singular project actually can become kind of hard. So I would say from personal experience that there is diminishing returns right? You'll see videos of people with like eight terminals up and I have no idea how they do it or if what they're doing actually increases their productivity. But having two or three up oftentimes can be very useful.
So, I hope having these six levels of Claude Code broken down to you in a systematic way that includes the skills you need to master, the traps you need to avoid, and what you need to look for to progress to the next level has helped, right? Because it's very easy to get stuck and overwhelmed because there's just so much information. So, as always, let me know what you thought. Make sure you check out the Chase AI Plus community. There is a pinned comment there. I have my brand new Claude Code Masterclass, which is perfect if you're someone who's trying to go essentially from zero to level six, even if you don't come from a technical background. And also, I have the free Chase AI community, which is in the description, so make sure to check that out if you're looking for some free resources. Other than that, I'll see you guys.