Basics · 5:16
Splitting tasks
One job at a time.
- 36 tips from beginner to master.
- Asking many things at once spreads Claude's output limit thin.
- Split tasks into separate prompts for depth.
| Approach | Result |
|---|---|
| All at once | Shallow answers |
| One at a time | Detailed answers |
Do this in your own tools, step by step. Ticks are saved on this device.
Why split?
Output limit per answer.
Tip number?
One.
0:00 Now, even if you've been prompting for a while and using large language models, I highly recommend you stick around to the end because I myself have prompted for hundreds of hours alone perfecting my craft and there are a lot of things in this video that I share with you that I've learned over the years of prompting in large language models. So, if you're ready to get into it, grab out your coffees, get out your notepads, and let's learn how to use Claude.
0:22 For those of you who don't know me, my name is Drake. I've been teaching artificial intelligence for a couple of years now and I have an AI Foundations University where it's a community, a classroom, a network, and a live call group all in one platform. This is the best learning experience for AI. I've learned more in the past 3 months than I have in the past 2 years studying artificial intelligence alone thanks to this community and thanks to getting all of these people from different industries together in one platform and talking about AI.
0:45 Now, I'll leave a link to join this in the description or the top pinned comment. So, if you like what you see in this video and you want to dive deeper into learning with us here at AI Foundations, highly recommend starting at the community. It's an amazing experience. You'll gain a network of people to talk about AI with and it's meant to be your final stop. In the community, we're talking about AI news all day long.
1:08 I mean, people are posting multiple times per hour. We have a classroom full of different courses to learn large language models, AI image generations, automation, and more courses to come soon. We have a calendar even with a bunch of different live calls and we do multiple different styled calls per week. We have live Q&A's. We have money making methods with AI. And of course, you get the community, classroom, and calendar. So, I highly recommend you join if you are interested in anything artificial intelligence.
1:31 I'll leave a link in the top pinned comment and the description below if you're interested in joining and without further ado, let's get into these 36 Claude tips and tricks. Tip number one is to split your tasks when using Claude. Often times we send off these long prompts that are very very detailed and we can't get as much out of Claude when we don't split up our tasks. I'm going to show you an example of what that means to split tasks.
1:54 Let's say I'm dealing with this long prompt in Claude and I want to create a YouTube channel in the artificial intelligence space and I want Claude to help me along the way. Well, here's my prompt. It says, I need help creating a YouTube channel in the artificial intelligence space. Please help me come up with 30 catchy names and describe why each one is good. Also, I would like 10 channel description variations that are general in regards to AI and also describe why those are good as well for each one and why they might work better than others.
2:17 Lastly, I would like you to put together a content calendar and strategy with mock video ideas and days of the week I should be posting on. Now, these are all amazing tasks for Claude to complete and each one has their purpose. Let me split them up real quick. So, we have our goal up top. We need help creating a channel. Next what we want is 30 catchy names and we want Claude to describe why each name is good.
2:41 The next thing we want is 10 channel description variations. And then finally, we want a content calendar and strategy with mock video ideas. These are all great tasks once again, but when we send off this prompt, Claude can only go into each one of those tasks so much because of its token limit and how much it can actually generate per output. As you can see, it's completed my request. It has the 30 catchy channel names with describing why each one is good.
3:04 It gave me 10 channel description variations. It gave me a content calendar and strategy, but it's only going so in-depth on each task because it has to think about and move on to the next task. So, tip number one is to split up each of these tasks so Claude can expend more of its output limit on that specific task. So, instead of me asking all of these questions at once, maybe I should ask them one at a time.
3:29 And this is if you're trying to get detailed and good responses. So, I can keep my goal, and then I can start a new chat, and then I can just give it one command at a time. And I can even tell it to be more detailed and go into specific things about that task. Maybe instead of getting detailed about your entire project, get super specific into one task. That way you can expend its energy and its output limit into that specific task.
3:52 So, now I'm just sending off this 30 catchy names prompt. I still have the same goal, but I'm just having it do one task at a time, and I'm having it be super detailed on the descriptions of why each channel name might be good and attractive for viewers. And once I send this off, you can just see how detailed it went into every single name. And it really helps me when it comes to making a decision for my YouTube channel based in the artificial intelligence niche.
4:16 I mean, just take a look at this one task, how many output tokens it expended. So, what I've done is I've taken these two prompts and I've put them in a Google Doc. And what I want to do is compare the word count when we split a task versus when we don't split a task. So, when we didn't split a task, and we wanted our 30 catchy channel names, this is how in-depth it went. Let's take a look.
4:38 I can go to tools and word count. As you can see, it gave me 334 words into this specific task. When I sent this task off to Claude on its own, it went much more in-depth. As you can see, we have 1,259 words compared to 334. So, this is the more in-depth explanation you're going to get when you start to split up your tasks in Claude. Tip number two is to use pictures for context. You can do this for so many different use cases.
5:04 Let me show you a few examples now. Let's say you have a Notion database or a Google Sheet where
Get the other 14 chapters, the exam and your certificate for LinkedIn, plus all 700+ AI courses.