useful tools, especially for programmers. I use LLMs every day, and I can't imagine a world without them. However, there are a few strategies you can use to achieve even greater results with LLMs.
I've covered a few different techniques in previous articles, such as:
- Using Slash commands
- Using program mode
- Updates agent.md continuously
In this article, I will cover how you can use a few prompts to make your LLMs perform even better.
Why use a few prompts
First, I want to cover why you should use multiple notifications. A few briefs are incredibly useful because they allow you to show the LLM your purpose without having to explicitly list the purpose in your brief.
For example, let's say you want a website designed in a certain way, similar to a previous website you made. And without a few prompts, you can try to define a previous website that you want duplicated and have LLM create that new website. However, this will lead to a lot of ambiguity in your information, where the LLM has to make some assumptions. Therefore, you will probably not achieve the desired result.
If you instead provide the LLM with the actual codebase, or at least screenshots of your previous website, and simply ask them to replicate the website, you will achieve much better results. This removes all the ambiguity in your knowledge and helps the LLM to get the greatest results.
I argue for the fact that you should use this approach to a few shots in everything you do. As long as it is not the first time you are working in a certain job, always look at one of your previous jobs to see how an LLM should do something. For example:
- Doing marketing stuff? -> show LLM your previous work
- Adding a new feature to your app? -> show LLM your previous features
- Creating new slash commands? -> show LLM how you edited your previous slash commands
I can almost guarantee that by referring to your previous work and showing the LLM how to do something not only quickly, but in real use, you will achieve the greatest results.
How to use it to inform a few
Now I want to discuss how to use a few prompts. Few information is not something you can use often. Some jobs are just new, and it's very difficult to take advantage of or use previous work you've done because the new job isn't similar enough.
This is absolutely fine and something you should embrace. However, you should always look for opportunities to use a few promotions. First, I'll discuss how you should plan your work to maximize the surface area for fewer shooting opportunities, then I'll show you how to do a few photo tutorials in practice, using examples.
Planning your work
First, it's important to organize all your work into accessible folders on your computer. Personally, I keep almost everything I do inside the main editing folder. I then have a folder structure for code collections that I often work on. Another folder that includes personal projects that I occasionally access. Another folder with marketing stuff I'm working on, like LinkedIn posts and short videos, and another folder for all the presentations I've held on AI.
Now, whenever I start a new project, my first task is always to find out what folder this project belongs to. In general, scheduling a job like this is just routine computer cleaning. However, being organized in this way makes it much easier to take advantage of the next few information. I always recommend that you spend some time checking where your work belongs in the beginning so that you can use it later.
In addition, you should always contribute your work to GitHub. The reason for this is that it allows you to save all your progress and gives you version history. So if something happens to your computer, or you make changes that you want to restore, you can easily restore them using Git.
In addition, if you have no knowledge of using Git, it is not really a problem, as you can use LLM to work with Git for you. You don't really need to interact with Git yourself.
A few caveats apply
Now, if we think that you have planned your work well, it is time to start taking advantage of a few information. The concept of a few values is very simple. Whenever you start a new task, you simply refer to the folder or file of the previous task that you want the computer to repeat or follow the same or similar style.
I think it's easier if I show you, if I explain some examples of how I use a few information when I work.
Writing code
Perhaps the most common use case for me if a few prompts is to write code. Let's say I want to run a GitHub Actions authentication script on a new repository. I actually never asked Claude Code to come up with this script from scratch. Instead, I just tell Claude's Code, “This script is in folder X, copy or paste the script directly into my current workspace. However, just make this one change where you don't use the validation script part”.
This has two main advantages. For one, I'm almost certain I'll get the GitHub Actions validation script I'm expecting, because I know it's running on another repository. In addition, this is good because even if I copy the text to another repository, I can still make changes. And in this example, the change was that I didn't want to use a full authentication script. I want to skip one part of it in this new repository.
Code Claude is great for dealing with these types of tasks, where you tell it to duplicate another piece of code and make a few custom changes. That's why this works so well.
Creating marketing materials
Another common use case I have for small compensation is creating marketing materials. Creating new marketing materials can be a time-consuming task. You should, for example, create brand new presentations or carousel views for use on LinkedIn.
In addition, it is often difficult to define an exact preference when it comes to presentations. You may want a certain type of font style or a certain type of alignment of text and images in your presentations. This is easy to explain in natural language, but it's much clearer to the model if you show it an example of what this text font looks like or how the text and images align with your previous work.
So, when I do a new presentation these days, I always show Claude Code my previous presentations and tell him the things I want to change from those previous presentations. The things I want to change are usually the actual content of the presentation, of course, where I describe each page in my presentation in as much detail as possible. This, of course, is important to keep the content your own and not AI-generated.
In addition, I am simply multiplying the Claude Code. I told it to do the first draft of the presentation for me. I then revise the draft, writing all the changes I want to be made with MacWhisper while revising the presentation, and have the AI do a second draft. I will continue like this until I am happy with the presentation.
Slash Commands
Creating slash commands is also something I do regularly. Slash commands are actually stored commands that you can have in code that allow you to access information quickly. I usually have slash commands for commands like creating a pull request in dev, creating a pull request in main, simplifying code, or running a PR update.
However, I usually want my slash commands to follow some sort of structure. Layout is a tag layout with a few points that I tend to share across my different slash commands. Therefore, showing Claude Code my previous slash commands makes the generation of new slash commands much easier, faster, and easier to follow the preferences I have.
The conclusion
In this article, I have discussed how you can use a few tips to get the best results with your LLMs. Effective use of a few pointers by showing LLM examples of your previous work can make your LLM more effective in your use cases. I recommend that you always strive to use a few prompts whenever you work with LLMs to achieve the best results. The best part about less information is that it gets better when you do more work. The more work you do, the more previous examples you have to show for the LLM, and the more it will work for you, which is what makes it such a great strategy.
👉 My Free eBook and Webinar:
🚀 10x Your Engineering with LLMs (Free 3-Day Email Course)
📚 Get my free ebook Vision Language Models
💻 My webinar on Vision Language Models
👉 Find me on social media:
💌 Stack
🐦 X / Twitter
