Prompting Basics

Last update: September 30, 2026

We need to spend a quick minute going over the basics of prompting. The most important thing you need to provide to an LLM is context. If you’re unhappy with the results you’re seeing when interacting with AI, chances are you could’ve provided better instructions or more context to the model. (Hint: they’re essentially the same thing.)

Imagine you have a task list that you need someone else to do — but they’ve never done it before. They’re probably not going to do it exactly like you want them to. But did you tell them exactly how you like it done?

The more specific you can be when talking to an LLM, the better. Being specific will reduce the amount of back-and-forth you have to do when reviewing outputs. This works the same way with humans, too… the more specific you are about what you want = (hopefully) less time correcting someone to do that way.

And just like a human will remember one of your preferences, an LLM can too. So if you have a specific way or style or convention that you want things done in — tell your LLM to capture a memory of that.

Here’s a few examples of memories that you could tell your LLM to remember:

Ensure that you cite your sources. If you do a web search and present information back to me, you must cite the exact website you got that information from.

It is critical to double-check the spelling of names in client-facing documents.

I work as a [job title] and regularly work on tasks like [this], [that], and [the next]

I prefer technical concepts to be explained to me like I'm 5

The key takeaways

• Be specific and provide context for what you want, how you want it done, etc.
• LLMs can capture your preferences or context as a memory and reference it before replying to you

This leads perfectly into our next topic. Check it out: