ChatGPT Projects: A Workspace That Keeps Your Context
Most people run their whole life through one endless ChatGPT thread. Projects fix that by giving each piece of work its own files, its own rules, and its own memory. Here is how I use them.
- #chatgpt
- #workflows
- #productivity
The single worst habit I see in heavy ChatGPT users is running everything through one infinite conversation. Your tax questions, your hiring plan, and your weekend recipe all bleed into the same thread, and the model carries the wrong context into every answer. Projects are the cure, and they take about two minutes to set up.
A Project is a scoped workspace
Think of a Project as a folder that holds three things together: the chats about one piece of work, reference files you upload once, and instructions that only apply inside that folder. You can attach PDFs, spreadsheets, docs, and images, and every conversation in the Project can see them — so you stop re-uploading the same brief each morning.
The instructions field is where it gets useful. You set how ChatGPT should behave for this work — "be concise, use bullet points, ask before assuming" — and those rules override your global custom instructions, but only inside the Project. Different work, different posture, no copy-pasting a preamble every time.
Why the isolation matters
By default a Project's chats are walled off: they can reference each other but not your general history or a different Project. That sounds like a limitation and is actually the feature. The model stops dragging unrelated context into your answers, which is the usual reason a long-running chat slowly gets dumber. A clean, scoped context is a sharper context. OpenAI pushed this further with project-only memory: turn it on and the Project ignores your saved memories from outside it, and nothing from the Project leaks into other chats — a fully sealed workspace for sensitive or long-running work.
How I structure mine
I keep one Project per ongoing responsibility, not per task. "Hiring," "this quarter's roadmap," "the website" — each gets its files and its rules. New question, I open the matching Project instead of a blank chat. The context is already loaded, so I skip straight to the actual question.
This week
Pick the topic you talk to ChatGPT about most. Create a Project for it, upload the two or three documents you keep re-pasting, and write a one-line instruction for how you want answers. Next time, start there — and notice how much setup you just stopped repeating.
Sources
Keep reading
More guides like this one.
ChatGPT Work vs Claude Cowork: the answer is in your admin console
Two workspace agents shipped 48 hours apart. The thing that actually decides which one your team runs isn't the model — it's a toggle an admin flipped before anyone got to try either. Here's the seam, and my routing rule.
What Anthropic's Own Teams Do Differently in Claude Code
Anthropic published how its internal teams actually run Claude Code — and the loudest wins came from lawyers, marketers, and data scientists, not senior engineers. Here are the five habits worth stealing, with the real numbers behind each one.
The Three Codex Upgrades Actually Worth Your Time
I run both Claude Code and Codex. Codex just had a big month — here are the three changes worth the ten minutes it takes to set them up, with copy-paste prompts and the one myth to drop.