Anthropic’s Secret Guide to Claude Usage Limits (7 Tips)

0:00 Anthropic has a secret guide that most people don't know about on how to stop hitting your usage limits in Claude. And I know just how painful it is hitting your usage limit, so I went through the

0:09 whole thing and pulled out the seven tips that actually matter that you can apply today. What's eating up your limit, what to change, and a real example for each one. So, by the end of

0:18 this video, you'll be getting a lot more out of the same Claude plan without having to pay more for it. So, let's dive right in. Tip number one is to stop using one giant chat for everything, and

0:28 this may seem obvious, but most of us don't listen to this advice. Now, on this page, Anthropic lists out everything that counts against your limit, and one of those things on that

0:37 list is the length of your current conversation. And on their other page about how usage limits work, they say that really long conversations, the ones where Claude has to start summarizing

0:47 your older messages, use up more of your usage limit, and if you think about it, that makes obvious sense. And they say if you're getting close to your limit in a long chat, try starting a new one. So,

0:57 this matters because I know a lot of you have one chat you've been using for like 3 weeks because I do the same thing. I'll have one long conversation for a previous task, and I'll just go back to

1:07 that same one and continue using it even if it's for something completely different. That single chat thread might have your emails in there, your meal plan, the spreadsheet you were trying to

1:15 fix. So, let's say you go into that chat and ask a question about a recipe. That quick question is now sitting on top of 3 weeks of conversation, and that will work against you. There's actually a way

1:26 to tell when this is happening. Anthropic says in long chats, you might see Claude say it's organizing its thoughts. And if you see that, that's the kind of chat that is using up more

1:34 of your limit. So, the rule is very simple. New topic, go to a new chat. But then you're probably thinking, if I keep starting new chats, do I have to upload all of my files again every single time?

1:44 And that takes us on to tip number two, which is to put anything you use more than once inside of a project. Anthropic says that when you upload documents to a project, they get cached. And if you

1:54 don't know what caching means, think of it like a bookmark. Claude has already been through that specific document, so it doesn't have to start from page one every single time you interact with it.

2:04 And in their words, the cached portions count less against your limits than the new content. Let me quickly give you an example. Let's say you've got a brand guide and a price sheet, and you attach

2:14 them to a new chat every time you need to write a proposal. Every time you do that, you're using up your limit on those two files all over again. If those projects are sitting inside of a project

2:24 in topic says you can ask questions about them while using fewer messages than if you were to upload them each time. But there is one catch here that's very easy to miss. The cache expires

2:34 after a period of inactivity. So if you come back to your project after a long break, your first message counts all of that content in full again. So what I do is try to knock out your project work in

2:45 one sitting if possible. To recap, the rule is if you're going to use a file more than once, it goes in a project. Now that covers your files, but the next tip is a setting that a lot of people

2:55 never touch. Tip number three is to stop using the biggest model on the highest setting for everything. So going back to that list that I showed earlier about what counts against your limit,

3:05 Anthropic has model choice on there. They have effort levels on there, and they have tool usage with things like research and web search. And on the usage page, there's even a separate

3:14 weekly limit just for Fable. And their advice here is pretty direct. They say to choose a lower effort level for routine tasks that don't need Claude's most thorough responses because higher

3:24 effort levels use more tokens. And I know this does not sound like rocket science and it's pretty self-explanatory, but I think more of us, including myself, need to pay

3:32 attention to this more. They also said to turn off the apps you've connected when a conversation doesn't need them, and to ask Claude not to search the web when you don't need current information.

3:42 So for example, if you're asking Claude to clean up an email and you've got the top model selected on max effort with five different connectors turned on, you're spending a lot of your limit on a

3:51 task that didn't need any of that. And I actually tested this in a previous video when I built the same website with every model, and the highest effort level wasn't even my favorite result. So, the

4:01 rule is to start low. Use a lower effort level for your everyday tasks, and only turn it up if the answers come back and it's clearly missing something. So, those are the three big ones. The rest

4:10 of these are pretty quick, so let's move on. Tip number four is to stop repeating yourself. Anthropic says to refer back to things you already said instead of typing them out again. And they even

4:20 give you the phrase to use, which is as I mentioned earlier. And if you're on a paid plan, Claude can search through your past conversations, so you don't have to give it the same information

4:29 again in a new chat. One of the examples they give is literally just typing, "Let's continue where we left off." and then the name of your project. So, this is what makes tip number one actually

4:39 work because you can start a new chat and still pick up right where you left off. To recap this tip, before you paste something into Claude again, just ask it to go find it or refer to a specific

4:49 project that you worked on. And this in return will save you on token usage. Now, tip number five is to send one message instead of five, which you're probably thinking, "What does that even

4:57 mean?" So, let me explain that right now. Anthropic says if you have multiple related tasks or questions, group them into a single message. Now, this part's funny cuz if you ever text a friend and

5:07 you're basically just trying to text them a paragraph and you send them like five different texts back and forth in order to come to the same conclusion, a lot of that is unnecessary. You could

5:16 just do that in one paragraph in one text. And their example show is math problems, where instead of sending a separate message for each one, you send them all at once. So, in real life,

5:26 let's say you send an email to a client, a subject line for it, and a follow-up for next week. A lot of people send that as three messages back and forth. Now, what we should do is put all three into

5:35 one message, and you get everything back in one shot instead of multiple different messages. And the same goes for editing because they say to send the entire text in one message rather than

5:44 breaking it up. So, the rule is if the tasks are related, they go in the same message if you really want to optimize your token usage. Now, tip number six might sound pretty stupid, but this

5:53 simple trick will save you on token usage. And that is to take a few seconds before you hit send. Anthropic says to review your message for clarity and completeness to reduce the need for

6:03 follow-up messages. And they say to avoid vague questions that might need clarification. And this is particularly important because over time, like myself, I get more and more lazy when it

6:14 comes to communicating with the AI. So to give you an example, a prompt like write a post about our sale is going to get you a generic answer, and then you end up sending more messages to fix it.

6:23 And then if we compare that to a message like this, write a short Instagram caption about our spring sale, it's for busy moms, and mention that it ends Friday. You can just send Claude that

6:33 one message and you're done. And for writing, they say to give it the requirements, the target audience, and the key points up front. So the rule is, before you hit send, check that it says

6:43 who it's for, what you want back, and what it has to include. And this brings us to the last tip, which is how you actually know if any of this is working. Tip number seven is to actually look at

6:53 your usage. I know so many people probably don't do this, and I personally don't do this enough, but this is one of the biggest things you can do to save on your usage. So if you're on a paid plan,

7:03 you can go to settings, then usage, and Anthropic shows you progress bars for how much you've used. And there are two specific things to look at here. The first one is your current session, which

7:13 is a 5-hour limit and shows you how much time is left inside of it. The second one is your weekly limits, and it tells you when they reset for all models. So what I would do is check this before you

7:23 start on a big task. If your session is almost used up and it resets in 40 minutes, just wait the 40 minutes. Now, those are seven pretty practical tips you can apply right now to save on token

7:33 usage. And this is all information that Anthropic gave us to help us save. If you guys enjoyed this video, make sure to subscribe to the channel for more content like this. And without further

7:41 ado, I'll see you in the next video.