Transcript: 9 Hacks to Use Claude Code Better Than 90% of People (In 9 Minutes)

Source video ID: XkSBO-CZDFs

Transcript

• 0:00 — All of my favorite cloud code techniques have two things in common. One, they’re extremely high leverage, and two, they’re dead simple to implement. And today, I’m sharing nine of them in 9 minutes with you. So, let’s get started. Now, tip number one is about MCPs and why you should stop using them. Gone are the days where we need to do MCPs for everything. Instead, you need to move away from MCPs and start looking for CLI. Now, CLI stands for command line interface. And anytime we can move from
• 0:30 — an MCP to a CLI, we want to. Now, why is that? It’s because they’re more efficient. They use less tokens. And oftentimes, we also get more functionality because, think about it, command line interfaces live in the terminal. Where does Cloud Code live? In the terminal. There’s just no overhead required. And because of that, the token costs tend to be way lower. One thing you want to think about though when you’re using a CLI, you are going to want to include some set of skills because the CLI is a tool that Cloud Code uses, but we want to teach Cloud Code how to actually use it. So those
• 1:01 — two things always go in tandem. So the second tip is related to a new feature that only came out a few weeks ago, and that is the buy the way feature or /btw. Now, what this does, it allows me to hold a sidebar conversation with Cloud Code inside of the same session, even if it’s doing something else. So right now you can see it’s doing a web search for me inside of a conversation that is rather long. While I do /btw, I can still talk to claude code and it’s going to have full context like a normal conversation, but it’s not going to stop what it’s saying. So if I say, “Hey, what’s up?” It will reply to me as if,
• 1:34 — you know, it was just a normal combo. The good thing about this is not that I can just talk to it while it’s doing something else. It’s the fact that this conversation does not add to the context window. So it’s almost like free. So BTW becomes a great efficiency hack when you’re dealing with large context windows. So speaking of easy efficiency wins with slash commands, another one is forward/hook and creating a hook that plays some sort of subtle sound whenever cloud code finishes a task. If you’re like me, you have multiple windows of cloud code open at any one time and sometimes you’ll have it do something
• 2:05 — and you’ll forget about it for 20, 30 minutes, an hour, right? However, if you have it played just like a small subtle ding or something, I’m telling you, you are much more likely to come back and see what happened, right? And this ends up saving you like hours over the course of a week. Really easy to do. Just do forward/hook and say, “Play a sound when you complete a task.” Now, you can also choose what you want that sound to be. So, you can point it at any sort of audio file. Now, tip number four is all about the context window, and that is forward slashclear. Now forward/clear is
• 2:37 — something that you need to be doing early and doing often. And what forward/clear does is it resets our entire context window. Now you already know the importance of that right? Clearing the context window because of context rot. Even though I have 1 million tokens available to me when I start a new sessions, the more tokens I use over time, the worse claude code gets. And if you want a great visual of this problem, you can see it right here in the long context retrieval graph. At 256,000 tokens, we sit at about 91 92% efficiency. When we are at 1 million
• 3:07 — tokens, that drops to 78.3 on Opus and 65.1. So, with this in mind, that is telling us that anytime we get to 20% of the context window, 25% we want to clear because if you don’t have a reason to stay in this side of the graph with more tokens, why would you? And that plays directly into tip number five, which is the status line. And that’s what this is down here. You see how it says seven levels, the model, and the context window? Well, that isn’t going to be up there naturally. But because I set this up, I always know what directory I’m in. I always know what model I’m using, and
• 3:38 — I always know where I’m at with my context window, right? I don’t have to do forward/context to see that I’m at 16%. Now, it’s super easy to set this up. All you have to do is forward/status line and then give it a prompt that says whatever you want to be on the status line. So for me, I said I want something that always shows my directory, my model, and my context window usage. And then you’ll always have it up and running. Now tip number six is all about skills. Specifically, the most powerful skill in the game. It is the skill creator skill from Anthropic. And
• 4:09 — despite its name, it does a lot more than create skills. It also modifies skills, improves them, and measures skills performance. And so this means if we ever want to change a skill, well, we can actually then use the skill creator skill to run tests on it, benchmarks on it, and it will give us actual quantifiable numbers that we can compare and see, hey, did our changes actually make sense. So anytime I’m messing with skills, I’m using this guy. Now, tip number seven involves a very potent cloud code feature, and that is agent teams. It’s actually still an
• 4:39 — experimental feature that you need to enable to have access to, but it allows us to go well beyond simple sub aents. That’s because normal sub aents operate in a vacuum, right? I can have Claude code talking to three different sub aents doing three different tasks, but they’re just doing their own thing. The only people they talk to is Claude Code versus agent teams, I can have a bunch of sub agents doing things, but if it’s in a team context, these sub agents talk to one another. They coordinate their efforts, which allows me to get better outputs. Now, it comes at the cost of slightly higher token usage, but that’s
• 5:09 — just something you have to balance out. And I found that the outputs are totally worth it. And like I said at the start, you have to enable agent teams in the settings. Easiest way to do this is just go to this page, search up agent teams cloud code documentation, copy this page, give it to cloud code, tell it, hey, I want you to enable teams. And then once you do that, understand to then invoke agent teams, you need to explicitly say create an agent team. So it won’t do it on its own. You have to a enable it and then b explicitly say it. Create an agent team to do a b and c. Now tip number eight might seem like the
• 5:40 — simplest, most obvious out of the bunch, but I would argue it is the most important and the most powerful. And that is ask open-ended questions in plan mode. One of the biggest problems if you’re coming into cloud code from a nontechnical background is you do not know what you don’t know. Now, if you’re inside of plan mode and you ask it to do something, it’s going to ask some very surface level questions that you probably didn’t think about to better guide its outputs. But the best way to actually force it to ask you deeper level stuff is to ask it questions like, “What am I not thinking about? What are
• 6:10 — the unintended consequences? What would an expert in whatever you’re doing be thinking about in this moment?” If you take that sort of mentality to claude code when you’re coming up with your plans, again, you are going to get infinitely better outputs and you’re going to educate yourself in the process. And lastly, for tip number nine, you need to start using Claude Code as a second brain because Claude Code is brilliant as a personal assistant, especially if you use a tool like this. And this tool is Obsidian. It’s completely free and it integrates so nicely with Claude Code. You simply
• 6:40 — download Obsidian. You create a vault, aka just some sort of folder on your computer that is designated the vault, and then you open up Cloud Code inside it. From there, you can have Cloud Code create all of your markdown files for you in an Obsidian type format, which makes it very easy for you to navigate through this complete morass of documents, but it also helps Cloud Code itself stay a bit more organized. So, you have sort of the symbiotic relationship between Claude Code, Obsidian, and yourself. And again, it doesn’t really cost you anything to do. And in terms of how to actually set this
• 7:10 — up using the terminal, it’s extremely trivial. All you need to do is navigate to whatever folder you have designated the vault. In my case, it’s literally called the vault. And then you open up Cloud Code. And there you go. You have now successfully combined Cloud Code and Obsidian. Just like that. And what you saw here with like the graph and all that, that’s the actual Obsidian app. So everything Cloud Code builds inside of the terminal, you set it as a markdown file and then you can view it through Obsidian. And if you’ve never played
• 7:40 — around with Obsidian before, again, it’s it’s a pretty great tool. Like it’s very easy to see how all these things connect. And if I actually click on something, it will then show me all the related documentation. So to you, the human being, everything’s like very transparent. It’s like how all these connections are being made. And again, the perfect use case here is some sort of personal assistant. Obsidian is great, but not really in a traditional coding type project. It’s great in any project where there is just a massive and growing amount of markdown files or
• 8:11 — text documents. That’s where the play is. So, those are nine of my favorite little Claude Code tips, tricks, and hacks that I think if you put into effect right away will sort of vault you ahead of most people who are using Claude Code. So, as always, let me know what you thought. Make sure to check out Chase AI Plus. There’s a link to that in the comments if you want to see my Cloud Code Masterass, which is the easiest way to go from zero to AI dev, especially if you don’t come from a technical background. I also have the free Chase AI community. There’s a link to that in the description. And as always, I’ll see you