How to get the most out of ChatGPT "projects"
(oh and Claude projects too)
If you’re like me and use ChatGPT more than Google, you’ll quickly find yourself juggling multiple chats, custom instructions, and scattered references for a single AI-powered task. ChatGPT’s new(ish) “Projects” feature is about to become your new best friend.
By housing all of your related conversations, files, and instructions under one cohesive umbrella, Projects can save you time, reduce confusion, and help you see your work at a glance.
What are “projects” in ChatGPT?
The “Projects” feature (of ChatGPT, Claude, and other LLMs) is designed to group together:
Chats: Keep relevant conversation threads together so you can quickly pick up where you left off.
Custom Instructions: Define or tweak instructions specific to that project, making sure ChatGPT responds with the right context or style.
Files: Store documents, code snippets, or key reference material in one central location without losing them across scattered sessions.
Projects help you tie things neatly together under a single project name.
Setting up projects
Starting a project is usually as simple as hitting a “Create New Project” button and giving it a descriptive name. If you’re working on a SaaS web application, you could have a “Backend API Development” project and a “Frontend UI/UX” project, each containing relevant chats, custom instructions, and files.
Here are some tips for getting started:
Name projects clearly: Choose descriptive titles like “E-commerce Checkout Flow” or “Internal DevOps Tools” to keep your workspace scannable.
Create a project right away: The moment you realize you’ll be referencing the same files or instructions repeatedly, consider making a new project. Don’t wait until you have half a dozen disorganized chats.
Define roles or goals: In the project description or pinned instruction, specify why this project exists. A short sentence like “Refine code for microservices and automate the build pipeline” keeps the entire project focused.
Tweak your custom instructions in projects
One of the most powerful parts of Projects is the ability to override your general ChatGPT settings with custom instructions that apply only within that project’s boundaries. This is where I get the most benefit out of projects. My custom instructions use to be really long, with lots of “in this situation, do this.”
How to make the most of custom instructions:
Specify tour tech stack & style: For a project dealing with Python microservices, instruct ChatGPT to assume Python code examples and PEP 8 formatting. For a project building iOS apps, you might prefer Swift with Apple’s recommended style guidelines..
Include common constraints: If your code must follow a certain pattern or avoid certain libraries (e.g., no external dependencies beyond the standard library), remind ChatGPT upfront.
Custom instructions allow for more context-rich, relevant answers without needing to retype the same constraints in every prompt.
Using separate chats inside projects
Inside each Project, you can have tons of chat threads. Resist the urge to shove all your questions and answers into a single chat, cluttering up the context window. Instead, think of each chat as a micro-topic.
By compartmentalizing your conversations, you make it easier to review past discussions and keep the threads from becoming unwieldy. This is especially handy if you need to share certain chats with colleagues without giving them every detail.
Final Thoughts
ChatGPT “Projects” isn’t just a way to group chats. It can help you maintain consistency and quickly access everything you need for a given endeavor. By creating well-structured projects, leveraging custom instructions, and keeping your files organized, you’ll get a much better experience.
If you got something out of this, you might like some older editions of the newsletter:
