Skip to content

Projects, Memory, and Longer-Running Work

What Projects are

Projects are ChatGPT's answer to sustained work. They let you group related chats, files, and instructions into a single workspace organised around an ongoing objective. OpenAI describes projects as having built-in memory: ChatGPT can remember the chats and files inside the project and use that shared context in later work. Projects also support the same built-in tools you may use elsewhere, including canvas, image generation, study mode, voice, and web search.

When to use them

For humanities researchers, projects are useful when the work has continuity. A book chapter, article, grant application, digital edition, teaching module, reading group, or recurring administrative role all benefit from a stable context space. Instead of re-explaining the topic every time, you can keep working in one bounded environment with the same files, the same instructions, and the same evolving conversation history.

Caveats

There are a few caveats. Project memory is helpful, but it is not magically transparent. OpenAI says there is no list view of project memories in the way there is for personal memory. If you want ChatGPT not to draw on a particular conversation, you may need to delete that chat or move it elsewhere. Temporary chats cannot be added to projects.

In practice, this means projects are excellent for continuity, but less good when you need extremely fine-grained control over what the system remembers.

How Projects, Custom Instructions, and Memory relate

A useful distinction:

  • Custom Instructions tell ChatGPT how to behave in general.
  • Personal memory lets it remember details across chats if you allow that.
  • Projects give you a bounded workspace for one stream of work.

If you care about intellectual compartmentalisation — keeping a grant bid separate from a teaching project, for instance — projects are usually a better tool than relying on memory alone.

Leif's Notes

Projects are best for sustained, bounded work. If you find yourself re-explaining the same context at the start of every chat, that is the signal to create a project. But be aware that you have less granular control over what the project "remembers" than you might expect.