32dots HEIDELBERG AI
Session 5 easy

Organize your work: Projects, custom instructions, memory

LESSONLesson 5 · ~18 min

🎯Goal. Stop repeating yourself: set persistent custom instructions, let ChatGPT remember useful facts, and group related chats and files into a Project.

▶ Try this prompt

Set as a standing instruction: I'm a molecular biology researcher. Always use precise scientific terminology, define acronyms on first use, give SI units, and flag when a claim needs a primary-source check. Apply this to all future chats.

Open Settings → Personalization → Custom instructions and put your standing context there. It then applies automatically to every new chat — you stop re-explaining who you are.

Steps
  1. 1Custom instructions (Settings → Personalization). Tell ChatGPT, once, who you are and how you want answers. It applies to all chats. Custom instructions are available on every plan, including free.
  2. 2Memory. ChatGPT can remember facts across conversations (e.g. your field, your preferred figure style). You can review, edit, or clear remembered items in settings — useful when you switch projects.
  3. 3Projects. Group related chats and reference files under one Project with its own instructions — e.g. a 'Thesis chapter 3' project holding your draft and analysis threads. Privacy gate: consumer chats may be used to train the model by default; turn that off in Data Controls, and keep unpublished or patient data on Team/Enterprise tiers, not the free tier.

You'll see. New chats that already know your field and answer in your preferred style without being told — plus a Project that keeps one piece of work's chats and files together.

💳Cost. Custom instructions: all plans. Memory: limited on free, expanded on paid. Projects: available on free, with expanded Projects/tasks/custom-GPT capacity on Plus and Pro.

💡Takeaway. Set your context once with custom instructions, let memory carry useful facts, and use Projects to keep a body of work together — and lock down privacy in Data Controls before uploading sensitive data.

How was this lesson?