In this video, you’ll learn:
Why long conversations cost more than short ones (and what to do about it)
The single highest-impact habit change to extend your usage
How Claude Projects use RAG (retrieval augmented generation) to protect your context window
A step-by-step walkthrough of the Settings → Capabilities toggle that paid users should turn on
5 practical habits that change how fast you hit your limits — whether you’re on a free or paid plan
00:00 The Dumbledore Problem (Why Limits Exist)
00:22 What This Video Covers
01:43 How Claude Usage Limits Actually Work
03:40 Why Long Conversations Cost More
04:22 The Context Window Explained (200K Tokens)
05:11 How to Turn On Auto-Compression (Paid Plans)
05:57 5 Habits That Change How Fast You Hit Limits – Habit #1: Start a New Chat When You Switch Topics
07:10 Habit #2: Front-Load Your Context (Don’t Build It Message by Message)
09:36 Live Demo: Writing a Front-Loaded Prompt
12:54 Habit #3: Use Claude Projects for Ongoing Client Work
13:36 How RAG Works in Projects (And Why It Helps)
15:36 Habit #4: Don’t Re-Upload Files Already in the Chat
16:13 Habit #5: Turn Off Tools You’re Not Using
17:02 Quick Recap: The 5 Habits
17:38 That’s a Wrap — Subscribe for Moore
The Dumbledore Problem (Why Limits Exist)
[00:00:00] Think of Professor Dumbledore, and he needed the pensive to take out memories and put them in there to store because his brain got too full. That’s what we’re dealing with.
[00:00:14] How can you get more use out of your AI tool, such as Claude, without hitting a wall or being throttled by usage limits?
What This Video Covers
[00:00:22] This video is going to explain how to maximize and best practices you can use, to make sure you’re not gonna get throttled. If you’re using a free or even a paid account, you’re able to get the most you can out of that platform for the uses that you need. So, I hang out on Reddit. I like visiting some discussions, and someone was talking about how I entered two prompts and already used up everything. And I’m like, how in the world did you do that?
[00:00:47] I literally have daily chats I’m going with in multiple projects in my Claude Pro account, which is 20 bucks a month, and I have yet to run into that. Now, I have seen signs where Claude is telling me it’s compressing in conversation so we can keep talking, which, actually, this is very timely. This is actually gonna help me with that as well, so hopefully this will be helpful for you to learn.
[00:01:12] If you’re using Claude regularly. You may have also seen the same thing I did, where it’s saying, Ooh, so we can keep chatting, I’m gonna compress this, or, Oh, you’ve hit your limit and you need to wait in order to access more Claude. If you’ve wondered why it seems to happen faster on some days than others, or why a friend says they’ve never hit it, when you’re hitting constantly like my person on Reddit, there is an actual explanation for that, and it’s fixable.
[00:01:37] So I’m gonna walk you through that today. This is very practical. We’ll get into it, and I’m gonna share my screen as well.
How Claude Usage Limits Actually Work
[00:01:43] First things, that most people don’t know – let’s talk about usage limits, why they exist, and how they actually work. People may not know this, but usage limits in Claude, and probably a lot of other AI, but we’re speaking specifically about Claude. They aren’t just about how many messages you send; they’re about how much Claude has to process to respond to each one. Okay?
[00:02:07] So here’s what’s happening under the hood. Every time you send a message, Claude rereads your entire conversation history to generate a response. Every single previous message in that chat counts. This made a lot of sense to me when I researched this and found this out. I used to use ChatGPT, and I just had a running chat.
[00:02:26] I thought, good, all the context will be in that chat. And eventually it was like, I’m tired, boss. I can’t do any more. But basically every single previous message, if you’re talking a lengthy chat, that’s gonna add to what it has to process.
[00:02:41] So like, a 50-message conversation costs dramatically more per reply than a five-message conversation, even if your actual question is the same length. It’s also why large file uploads eat into your limits faster. Claude’s processing that full document content every time you ask a follow-up question in the same chat. I do this a lot. I’ve had historically, and I won’t do this anymore. I’m gonna take my own advice here.
[00:03:07] I’ve had long chats going where every week I just go into the same chat and ask it to help me with a video script. And I’ll even go in afterward after I’ve edited a video. I’ll upload subtitles of that video. Say, can you help me with a good YouTube description that sounds like me?
[00:03:24] The more I pile onto that chat, it’s groaning under the weight. Okay? I could easily go to a new chat, and I’ll explain more about why this particularly works in Claude Pro with the use of projects, but I’ll also give you some options for while I’m not paying, how do I do this? To go from one chat to another.
Why Long Conversations Cost More
[00:03:40] So long conversations. The crux of this is long conversations in your AI tool are expensive. Fresh conversations, cheaper. Okay.
[00:03:49] Limits reset on a rolling basis, but not necessarily at midnight. So if you’ve hit a wall, you’re probably just waiting a few hours, but not necessarily until the stroke of 12. And you can find all this in the Claude help article, understanding usage of length limits.
[00:04:04] I have good news if you’re a paid user like myself. If you’re on a paid plan in Claude, that would be Pro, Max, Team, or Enterprise, and you have Code Execution enabled, Claude now automatically manages your long conversations. And this explains why I’ve seen my little message when I’ve seen it.
The Context Window Explained (200K Tokens)
[00:04:22] When you approach a context window limit, Claude will summarize early messages to make room, and it doesn’t count against your usage limit when it does that summary. How large is that context window on paid plans? It’s 200K, meaning it can ingest 200K plus tokens, about 500 pages of text or more when you’re using a paid Claude plan. Okay.
[00:04:45] And you’ll see it, like Claude will say, I’m organizing my thoughts. That means the feature’s working, your full chat history is preserved. So Claude can still reference it even after some summarization. So one thing I wanna show you all what to do is how to go find that. Okay?
[00:05:00] So I’m gonna go here into Claude. I’m on the browser version, not the desktop version. This works just the same. To get into this, you’ll wanna go into your Settings. Here.
How to Turn On Auto-Compression (Paid Plans)
[00:05:11] Once you click in settings, you’ll see, and again, I’ve got a paid plan, so I’ll see all these options. You’ll wanna click on Capabilities. All right? And scroll down a bit, and you’re gonna run into the Code Execution and File Creation. You wanna have that toggled on. That will allow you to have that happen, where it can automatically compress longer conversations so y’all can still chat without running into your usage limits.
[00:05:36] There is a caveat to this, by the way, on very large first messages, like uploading a massive document at the start of a conversation, that could hit your limits in edge cases. So just be aware of that. Alright, so let’s talk about the five habits that will change how fast you hit limits. So first, I bet you can guess this one.
5 Habits That Change How Fast You Hit Limits
Habit #1: Start a New Chat When You Switch Topics
[00:05:57] Start a new conversation when you switch topics. Like I said, I’ve had different chats where one chat is about a particular topic, but I can skate into others. But yeah, just switch. Just do a new chat and start going. It’s the single highest impact habit change that you can do to lower and make sure you’re not doing all that usage.
[00:6:16] To hit limit, you need to treat conversations like individual work sessions. Finish a task and start fresh for the next one. This might be a little bit of a retrain for you. I think right now, about like my Slack that I’m involved with, for any of my clients or for my own contractors, where, you know, within a Slack, we’ll probably start a message, and there’s a thread going. That thing might go and go forever.
[00:06:41] Obviously, that’s topically still the same. But you can imagine the longer that thread goes, it becomes a little bit more unwieldy. It’s taking up more room. Instead training yourself to say, Hey, lemme go start a new chat if the topic deviates at all. Bring up a new chat in AI, and that will help you with the usage limits.
[00:06:59] So don’t use one massive ongoing thread for every piece of work you’re doing. It’s gonna get expensive, fast. All right.
Habit #2: Front-Load Your Context (Don’t Build It Message by Message)
[00:07:10] Secondly, this has to do with prompts and being good at prompts. Front-load your context instead of building it message by message. I am bad at this. I’ll just be honest. I’m getting better. But we’ve talked about prompts a little bit, and here’s a reminder: so before you start a task, before you even start engaging, you need to ask yourself: what does Claude need to know to do this well? So I can reduce that back and forth. And remember, if you’re thinking about usage limits, this is in your favor.
[00:07:36] You might be like, yeah, but I like the back and forth. It’s fun, it’s conversational, it’s pricey. It’s gonna cost you, so you need to consolidate and think of it this way. If Claude was your coworker and you’re like, okay, but I’m in a hurry. I’ve gotta leave for doctor appointment in five minutes. I don’t have time to go back and forth. I’m gonna be driving.
[00:07:53] Think of it like that. Let me give you as much and front-load as much information as can give you from the get-go. That way, you can be thinking and processing, and delivering, in the background while I’m driving or going to do whatever I think.
[00:08:06] Because a big thing about AI is to give it stuff, let it process that. Switch tabs, switch apps, go to something else, and work on it. And you can actually set notifications. So it’ll be like, I’m done. I’m ready with my response when you’re ready to jump back. Okay.
[00:08:21] So put all of that front-loaded context in your very first message. What is Claude’s role? Think about that.
[00:08:28] If you think of it like a coworker, what’s their title? What’s their background? What do they have expertise in? How are they going to be good at this? What’s the context of the task that you’re asking ’em to do? Are there any constraints? I only want you to be looking at these sources. Give it all of that upfront.
[00:08:45] And y’all, I wanna encourage you, voice to text is your friend here. You don’t have to type all that in. You literally can narrate it into the prompt field and just go on. If, as long as that’s all in one particular message, you do wanna have your thoughts kinda organized. But that’s at least all that front-loaded is all gonna be in one message, okay? So use voice to text, it’s your friend.
[00:09:06] Every additional clarification message that you have to send, because your first prompt was vague, is a token burn. So remember it behooves you to do that all in the first. And maybe what you need to do is set up yourself a template somewhere. It can be in a text app or a notes app or whatever, where you’re like, it reminds you, okay, what do I need to put in this prompt?
[00:09:28] Like I said, read off that script. Know that you need to voice to text it into the prompt field, and away you go. Okay.
Live Demo: Writing a Front-Loaded Prompt
[00:09:36] I’m gonna give an example of this. So let me go over to my Boss Level. And actually, I’m gonna start a new chat. Lemme go to my projects, how-to videos for marketers, and I click on that. I’m just gonna start a new chat here in my Claude, and I’m gonna say, okay, help me write a video script about Claude usage limits.
[00:10:02] Now, sure. I could hit enter. Claude is inevitably going to come back to me with questions, as it should, because it’s like, I would love to help you. And a coworker would do the same thing. They’d be like, I would love to help you. I need a little bit more information. What are we talking about? So think about that. instead I’m going to be like, you are a video expert in the ways of YouTube and how to videos, who is going to help me script a five to 10-minute-long video about Claude’s usage limits and how to optimize whether for a free or paid plan.
[00:10:38] Okay. I’ve already given it context there, given it its role. I want you to pull from Claude’s help articles, but also guide me where I have an opportunity to share my screen that can be helpful for viewers. All right. I really tend to like how-to videos that go step by step, taking me through each phase of navigation or reasoning so that the explanation makes sense as well as demonstrates what the user should do on screen.
[00:11:05] Okay. So, given it some example there, my goal is to really answer the question I’ve seen on Reddit a ton. I have just entered a very lengthy prompt. Please notice too, and this might drive some of you nuts, if you really love paragraphs, I didn’t do any paragraphs.
[00:11:24] Now sometimes, I am a paragraph person, and I tend to do that in like social captions and whatever. If I hit shift enter, that means I can keep going. I can put hard returns and paragraphs in here if that’s what you wanna do. So that’s a quick hack.
[00:11:38] If you wanna just I need this to look pretty to your eyeball, but you literally can create a run-on like this. Or as I mentioned, do a voice-to-text that does all this stuff, but I’ve given it all those things. I told it what its role is. I’ve given it a lot of different context, not even in the same parts of the paragraph. I gave it the goal, what do we wanna accomplish here? I’ve given it constraints, telling it what to look at, and even an example of what we want from this. Okay?
[00:12:06] So I am gonna hit enter on that, and Claude is now gonna think through all that stuff. So it’s thinking through all this, I can go do other stuff while it’s doing this, but now I wanna point this out. I have everything I need to write you an accurate SEO optimized script. That is how you know you won. Okay. Because so often, and that’s one thing I appreciate about Claude and about AI and good, well done ai, engineered AI helpers is that.
[00:12:35] It will often come back to you say, I need some more information in order to proceed. You didn’t give me enough. I run into this sometimes, too, again, when I’m being super quick and just tapping out something I need it’s help with. But the fact that it said, I got what I need, let’s go. Okay. So that demonstrates that we did a good job.
Habit #3: Use Claude Projects for Ongoing Client Work
[00:12:54] Okay, going back to the five habits that will change how fast you hit the limit. A third one, use projects for ongoing client work. I do this, I love this. So if you’re working with the same brand, client, or content series repeatedly, projects are built for this.
[00:13:11] I’m gonna hop over into one of my projects. I’ve got Rachel, the CEO of a podcast company. It lets you front-load a lot of different context and things like that. I think I actually have a video telling you how to create a project in Claude, but when you upload files to a project, Claude uses a system called RAG, which stands for retrieval augmented generation.
How RAG Works in Projects (And Why It Helps)
[00:13:36] Instead of loading all of your documents into every conversation, it only retrieves what’s relevant to a conversation. That means your brand guidelines, client briefs, and tone of voice docs don’t eat your context window every single time. This is why you wanna do that.
[00:13:50] So right here is a place where you can upload files. If you do have some brand guidelines that have your writing style, your brand colors with hexadecimal values, RGB, CMYK values, or whatever the fonts you like to use, the tone in your voice, in your target audience, and things like that. You can also put in instructions, and it builds memory based on what you’re talking about. Okay.
[00:14:18] And I wanna point something out too. Instructions are something you enter at the get-go as well as files. The memory gets automatically built as you go, but I could go in and adjust it. This has been built as I’ve used this project, so Claude has put this together based on the context of the conversations we have.
[00:14:38] If I decided to go in, and I actually did this one time because I had a project where I had a client’s two unique restaurants in the same project, and I realized I need to parse these out. They’re not the same. Their voice is different, their tone is different. The vibe is different, the experience is different. So I actually went into the memory of the project where I was bringing one of those restaurants out of, and I deleted references about one of the restaurants that I was putting in a new project. Okay. So you have some control here.
[00:15:07] Both you and Claude work on that memory. Keep your project instructions concise.
[00:15:14] What you wanna do is treat them like a standing brief, not an encyclopedia. Long project instructions load every time. I’m gonna go check what I did. I did about three paragraphs. That’s not terrible. If I were doing, if you’re doing like a saga in here, that’s like several pages long, too much. Remember, it learns. It builds that memory based on what you’re talking about.
Habit #4: Don’t Re-Upload Files Already in the Chat
[00:15:36] All right, fourth habit to avoid usage limits. Don’t re-upload files you’ve already shared in a conversation. Don’t need to. You can say, Hey, look back at that one. If you’ve uploaded a document earlier in the same chat, Claude didn’t lose it.
[00:15:46] It’s not like that coworker that’s like, I need that link again, which is very human. AI is not; you don’t need to upload it again. Re-uploading means that Claude is processing that entire document twice for no reason. So don’t do it, just reference it by name or describe and say, Hey, remember that CSV that I uploaded, or remember that PDF I gave you that’s got this title or whatever.
[00:16:07] And usually it’ll be like, you’ve uploaded a few files. I think it’s this one. Is this right? And you’ll be like, yes. Okay.
Habit #5: Turn Off Tools You’re Not Using
[00:16:13] Final habit to help you with your usage limits. Turn off tools and features you’re not using. Ooh, so think about this. Maybe you don’t need web search, extended thinking, or connected integrations like Google Drive or Slack for a particular chat. Those are token-intensive.
[00:16:30] When you’re doing a focused writing, for instance, or an ideation test where you don’t need live search or external data, turn those things off. Remember, I told you we’re gonna go back to settings here. So again, this is in settings, going to capabilities.
[00:16:43] Once you’re in settings and scroll down under that code execution and file creation is this allow network egress. So this is where you can turn this off in case you’re not really needing it to go out and about and find some things out in the interwebs. Okay? So you can turn that off.
Quick Recap: The 5 Habits
[00:17:02] So just to summarize again, usage limits aren’t a fixed wall. They’re a reflection of how much you’re asking Claude to hold in its head at once. Think of Professor Dumbledore, and he needed the pensive to take out memories and put them in there to store because his brain got too full. That’s what we’re dealing with. I love analogies.
[00:17:22] Start fresh conversations more often. Front-load your context. Use projects for repeat clients or purposes, and turn off the features you’re not actively using. Those habits alone will meaningfully change how often you hit a limit. That’s a Wrap — Subscribe for Moore
[00:17:38] I’ll put links to all the Claude Help Center sources in the show notes. That’s officially from Anthropic, so worth bookmarking. I would be willing to bet if you’re not using Claude, but something else, they have something similar, and that seems to be across the board, is that the shorter conversations is the way to go. Full stop.
[00:17:57] Thanks so much for watching, everybody. Good luck with your usage limits in your AI tool of choice. Let me know how it goes and if this helps at all, or if you have other suggestions for how we can all keep trucking with our AI chats.
[00:18:11] You’re still here. It’s over, but you should hit like on this video before you head out and subscribe if you wanna get “Moore” videos that’ll help you realize how much “moore” there is to marketing.
[00:18:22] See what I did there?
Whether you’re using Claude Free, Claude Pro, or a team plan, these best practices apply. And yes, this works on the browser version and the desktop app.
Claude Help Center — Understanding Usage Limits: How do usage and length limits work?
How to Create a Project in Claude: How to Set Up Your First Claude AI Project
But wait! (Yep, there’s more for subscribers.)
Exclusive content is available with Rachel’s livestream videos: Boss Level Mondays and Hot Mic Fridays. Start your week off with irreverence and real talk as you face down the work week of a marketer with Rachel at your side. Then, on Fridays, head to Hot Mic Fridays for unfiltered takes from the week, as well as podcasting and video tips you can always use. Subscribe here on Substack to access and join Rachel live!