Creating a Memory Block
This page helps you create a memory block with an AI assistant. If you want to know more about what they are and why they’re useful, check out Give Your AI Assistant a Memory.
There are really only two things you need to do:
- Tell the assistant what you want to work on
- Ask it to return a memory block you can paste into a fresh chat
That’s the whole thing. Everything else is details, and the assistant handles those.
Don’t Overthink It
Open a new chat and say something like this:
I’m working on a project. It’s a react front end that talks to an api. I’d like your help working on it and want you to have all the necessary context you need to do a good job. ask me for whatever information you need to have a full understanding of the project. I’d like you to return a memory block I can paste into a fresh chat with you. With that memory block you’ll have all the context you’d need to work on the project with me.
Don’t worry about the wording. Don’t pre-edit things in your head or try to optimize it. The assistant doesn’t care if you missed a capital letter or two. You’ll work out the details in the conversation.
Talk with the assistant about your project. It’ll figure out what’s important and turn that into the memory block.
Once you have the block, give it a quick once-over to make sure there aren’t any big mistakes. Don’t worry too much about it though, this is a first draft. If you find a mistake, ask the assistant to fix it. When it looks good, paste it into your notes. Then open a fresh chat, paste in the block, and start chatting.
Improving the Block With Each Chat
At the end of each chat, say something like this:
look over this conversation and figure out what we’ve learned. I’ll paste in the memory block, tell me how you’d update it. Once I confirm the changes, return the updated memory block.
Don’t overthink the wording here either. Tell the assistant what you’re hoping for and it will figure out the rest.
Paste in the memory block you used to start the chat at this point. The one that’s already in the conversation is all mixed up in the context and harder for the assistant to work with cleanly.
Over time, that memory block will describe your project and your preferences better than anything you could have written in one go.