10 Ways to 10x Your Claude Usage

Everyone who's used AI has ran into this problem of the usage being limited. And many people think that it's limited by the amount of messages per day. Like you get 45 messages and then boom, you're cut off. But that's wrong. It's not how it works at all. And once you understand what's going on under the hood, you can get way more out of the exact same plan that you're paying for. And that also goes for the free plans and applies to all AI models, not just Claude. With without spending an extra dollar. And here's the best part. Almost everything I'm about to show you comes straight from Anthropic's own documentation on how to get more out of their services. People just don't talk about this stuff. So we're not guessing here. This is from people who actually built the thing. So here's the most important thing upfront. If you take one thing from this video, make it this. Claude does not count messages. It counts tokens. What are tokens? Think of tokens like this. They're units of effort. Every word Claude reads and every word Claude writes costs you some tokens. So very high level, a tiny little, hey, fix this typo message costs you very little. But a giant 20-page document with web search turned on running the heaviest model, that one message can cost you as much as 30, 50, maybe even 100 small messages. So to explain this further, we made a little custom site that shows this, visualizes this so well. We'll put a link in the description below. But here you can really see what a token is. Because it's not a word. In this sentence, the cat sat on the mat, dot. You can see over here, this might be 6 words, but it's only 7 tokens. The is a token, cat is a second token, and so on. Now let's go to the other extreme, which is a really complicated and long word. How about supercalifragilisticexpialidocious? That's one word, but it's 10 tokens more than the sentence with the cat. So just using this word is more expensive and will drain your token budget, your usage limitation with any AI more than this sentence. And that's what matters. As I told you, every AI service runs on tokens. It's really the gas cost for you driving this amazing vehicle that AI is. Okay, trust me, this knowledge is going to be essential when we get to these tips, because you do have to understand tokens. And there's one more thing to understand, which is the context window. You might already have an idea of this, but it's really good to re-explore that idea. And we we have a separate tab in this little app here to show you, because the context window is basically the budget of token usage you get within one conversation. Now, let's say you're starting a brand new chat. There's no context in your account, no memories, no extra stuff, no system prompt, nothing. Let's just set this to zero. And as you send more and more messages, well, this context window fills up over time. And once you send too many tokens, the context window will fill up. And then look here on the left part, you will see that it starts cutting out the first messages that you sent in order to stay within the context window because it's limited. This is what happens when people say, oh my God, my AI forgot. Yeah, you send too many tokens and every AI has a limited window. And for this video, that matters because it doesn't just limit the long-term memory, as I showed you of this, where it starts forgetting the first few messages you sent at a certain point, but it also matters to how quickly your account fills up. So let's look at how to be more efficient with the currency of tokens with the following tips. And with that, every upcoming tip will become really simple to understand because it always comes back to how do I stop wasting tokens. First up, here's the easiest one. You might notice already, but use the right model for the job. So again, we're doing this video on Claude, but this really applies to all AIs. In Claude, you can select various models. Here. You can use Opus, which is, you know, the smartest model. But as you can see right next to it, it consumes usage limits faster than other models. Opus is more token hungry, and the tokens are more expensive, but we don't have to focus on that right now. The mistake that many people make is basically for anything that they do within Claude, they leave Claude cranked up on the most powerful model all the time. And using a big model like Opus to rewrite this email so it sounds friendlier is like taking a Ferrari to go pick up groceries. Sure, it works, but you're burning way more fuel than you need to. And my recommendation is that for simpler, lighter tasks, use Sonnet. Quick intermediate side note, this effort selection here also affects how much usage and therefore tokens it's going to use up. The higher the usage, the more tokens you're going to spend, but the smarter of a model you get. And for many day-to-day tasks, you don't need a lot of intelligence at all. This one change alone can stretch your limit dramatically. One quick note on these models, they keep updating this, but what you're learning in this video is timeless. So in the future, the model names will change. But this idea of having one frontier model that is state-of-the-art and super smart, and also takes up a lot of tokens, and then other models that are more efficient will always stay the same. So this knowledge will apply to wherever AI goes in the future. And in this video, I want you to pay attention to the techniques not the names of the models. That's what's important and transferable. On to the next tip. And we kind of touched on this already. And that's turn off thinking when you don't need it. And that really has two parts. The first one we covered already, you can pick the effort from low, medium, high, or maximum. And then there's also this thinking option. Now, I understand that this can be a little confusing. There's different models, I get that. But why is there options within the models? You can essentially think of it like this. If me and you do some work together, and I give you a task, Hey, go ahead and draft me this email. There's a way of doing that task where you kind of just sit down, you do a quick draft, and you hand it back to me. And then there's a second way of doing that task where you go ahead, you draft the email, but then you sit down and you're like, hmm, I wonder if it could be any better. And you scrap the first draft and you make some improvements, you do a second version, and you're like, I wonder if I really understood Igor's assignment well. And then you scrap the second version and you redo it again. And then you're like, could I make it any better? And you do that again. That right there would be you thinking and putting in more effort. Obviously, that took more energy and time on your part. And with AI, with thinking on, it kind of does this loop of prompting itself to give you the best possible result. The cost of it is more tokens. And that's amazing for tricky problems that need it. But if you're just asking what's a good subject line for this email, you do not need Claude sitting there and pondering the meaning of life first. Flip the toggle off for easy stuff and you'll increase your mileage with Claude. Cloud usage. So here's the next one. And this is huge. It's actually so big that I've made it the central piece of my AI teachings everywhere. If you follow this channel for a long time, you've heard this. If you've joined the AI Advantage Club and took the courses I have in there where I teach how to use Claude and how to manage context, that's the central piece. Live inside of projects. This is available both in Chat and Claude. And if you go inside of a Claude project, you will see that here on this side of the screen, it shows you everything that it attaches to its context and therefore also everything that you pay for in tokens with every message you send. But here's the magical thing about projects. Anthropic does some technical magic that we don't need to talk about right here that makes the context you put over here way, way more efficient than anything you put into a chat, especially the attached documents. So here's what this behaves like in practice. Let's say you start a new chat and you've got a document that you keep coming back to. Most people paste that same document into brand new chats every single time. Instead, go into projects, make a new one and attach it into the files here. And what it does in practice is every time you work inside of a new chat here, the documents that you attached here are not entirely free, but at a massive discount when it comes to your usage. So if you work with the same file a lot, this is the single best habit you can build. Upload once into a project and use it forever. Seriously, if you watch your usage, which we'll talk about shortly, night and day. And hey, real quick, if you're getting value out of this video, do me a favor and hit the hit the subscribe button. We make videos like this every single week, and it actually really helps. All right, back to the next tip, which especially for people who just get started with AI is often a little counterintuitive. Start a new chat more often. Now I know there's something beautiful and simple about just starting one giant conversation that goes on forever. It's almost like one neat folder for your entire life or having one assistant that just helps with everything. But remember, tokens, Every time you send a message in any chat ever, actually, Claude rereads the entire conversation up to that point. Yes, you heard me right. If you send 50 messages in a chat, on the 51st message, it reads all of your 50 messages and everything it responded to before it gives you an answer. You know how many tokens that costs? A lot. That's how much. And if you're starting a totally new topic, just open a fresh chat and you're not going to be paying for the 10 other tasks that you did above that are completely unrelated to what you're doing now. Doing everything in one chat is the easiest way to burn through your usage. So don't do that. And it's not just usage. Even the responses are gonna come in faster because Claude doesn't have to reread the 3-hour conversation about something completely unrelated. Next tip, switch off tools you're not using. This one can be quite impactful, especially for people experimenting a lot with Claude, because you start looking into various features. You see you can connect all of these apps like Canva, Gmail, Slack, all the good stuff. And those can be super useful, but here's what most people don't realize. Every tool you connect comes with its own little instruction manual. Sometimes it's not even that little that tells Claude how to use the tool. And if those manuals are all loaded up at the start of every chat, they're taking up room in the background, whether you actually use them in that conversation or not. Now, the good news is there's two easy ways to handle this. First, the obvious one. If there's a connector you genuinely never use, just switch it off right here. I go to plus and here on the connectors, I can just turn them off. Boom, the invisible instruction manual that I would now be paying for in tokens on every message is gone, clean and simple. But now you might want to keep these on, and there's a smart way of going about that. And almost nobody knows about this feature. So instead of completely disconnecting your tools, you can keep things lean. Look down here under tool access, you've got two options. There's tools already loaded at the bottom, which keeps all the instruction manuals open all the time. And that fills up your chat faster, costs a lot of tokens. And then there's load tools when needed, which only pulls a tool in the moment you actually need it. So make sure this is set to load tools when needed. And now your tools are still right there, ready to go if you ask for them, but they're not sitting in the background taking up space the whole time. Your conversations stay lean, they run longer before Claude has to start trimming things down again, and you get more out of every chat. Set it once and you're good. Okay, next up, and this is something that you will naturally do as you use AI more and more, but batching your questions helps. It's an easy win. Instead of saying, summarize this, and then giving it a text, and then saying, now list the main points, and now write me a title for a blog. If you put it into one message and say, summarize this with the text, list the main points, write me a headline in a casual style, it's gonna take up like one-fifth of the usage. Because remember, every back and forth with Claude or any other AI rereads the whole conversation. And here's a little bonus, and this comes from experience. The answers you get from sending things at once are usually better this way too, because Claude can see the whole picture at once and understands the destination that it's working towards, and it's just more efficient, and it just works better than redirecting it multiple times in different directions. But feel free to make your own experiences. Okay, here's the next tip, and it pairs perfectly with the last one. Be specific with your messages the first time around. When you give an AI vague requests, you might have heard this, what happens? It gives you vague responses. Worst case, it might come back with something like, sure, happy to help, but can you tell me a little bit more about XYZ? And every round of this is costing you. So front load it by telling Claude exactly what you want, who it's for, and any details it might need all in the first message. Clear in, clear out. And you skip all those clarifying questions entirely. And get right to the output you need, which obviously is more token efficient. All right, here's the next tip. Let Claude do the remembering. If you're on a paid plan, Claude can actually look back through your past conversations and pull whatever is relevant to you. They call this memories, and it means that you can stop copy-pasting your entire life story into every new chat when it's needed. You don't have to re-explain your business, your project, your preferences, your goals every single time. Claude can just remember them. That's tokens you're getting at a massive discount instead of retyping it and paying paying for it all over again. Now, here's the thing. Memories work well. I do recommend using them. But what works actually way better is actually taking control of your context and therefore the tokens by manually curating this. Now, there's probably a thousand ways to do it. And not all of them are created equal. Over the years, I created a step-by-step methodology and an exact form. And there's a few tips around that, that go way beyond the scope of this video. But that's why we created something we call the AI Advantage Club. It's a community. And on this context management piece that is so essential, particularly I created a course that in the club we call the Level 2 Navigator course. We literally tied it to the most essential level of our free level system, Explorer, Navigator, Architect, because it is so important. So if you want to learn how to set up a project with the right context and a bunch of tips around it from me, well, check out the Level 2 Navigator course. You can join the club and gain access to everything for just $1 on our trial. You can find a link in the description below if you're interested in that. But now let's look at the next tip, how to increase your mileage with AI. And this one is just about being aware of your token usage. And what better way is there to be aware than to measure it? Most people don't even know this exists. But if you're on a paid plan, you can go to Settings, then Usage. And these free progress bars show you exactly how much of your sessions you have used. This is basically your token budget across Claude AI in the web, Claude the application, Claude Code that you might be using with this account. All of it is tied to your plan and you can track the usage in here. And this really matters because if you know this exists, you're not flying blind. You can see when you're getting close to the edge and you can see exactly when your limit comes back. And I found that knowing my reset timings completely changes the way I work with my plans because oftentimes you don't need a bigger plan. You just need to be aware of these reset windows and do a little bit of planning on how you're going to use AI and save a lot of money, frustration, and also time in the process. And that gets to my last tip, which is a sneaky little pro move that most people don't talk about, but it's timing your heavy work. And there's a sneaky nuance that Anthropic won't even admit to this. But at the base level, it's just, hey, if you have a lot of work, time it with your reset windows here for this usage tab. But beyond that, there's something that is referred to as peak times. Now roughly that's somewhere around 8 AM to 2 PM Eastern time. In Europe, that would be 2 PM Central European to 8 PM Central European. Those are the peak times. That's when most people use their models. And again, they never officially admitted this, but many users noticed that during these peak hours, you burn through your 5-hour window faster than usual. Your weekly total doesn't change, but inside of that window, a session gets eaten up quicker. So if you've got a big token-hungry task, like a giant document or a long research session, try to do it outside those peak hours. If you're in the US, afternoons and evenings or weekends are a good rule of thumb to get the most bang for your buck. If you're in Europe, it's the first half of the day or the weekends. You'll get noticeably more done in a single sitting. And then a final bonus tip before we wrap up here is if you're doing something important and you're in a situation where you literally cannot stop right now, well, if you ever hit those limits, all the paid plans have an escape hatch called usage credits. You can basically pay as you go for extra usage on top of your subscription instead of getting blocked out entirely. And you can turn that off anytime you want. So if you did everything in this video and you still ran out of tokens and you just need to get things done, That's your safety net, but it'll cost you. So there you go. Those are the main ways to get more out of Claude without spending a single extra cent. And again, we used Claude as the example, but this really does apply to all services. They all use tokens and independent of what model is out, these principles will make your money go further. If you enjoyed this, make sure you subscribe to the channel. We upload new videos every single week. And with that being said, my name is Igor Pogany and I hope you have a wonderful week.