Modern ChatGPT Workspaces and Multimodal Tools By the end of this lesson, students should be able to: Four different kinds of context. A strong user knows where information lives. A single chat has a working context. A Project groups related chats, files, and project instructions. Saved memory may carry selected preferences across conversations. Temporary sessions are appropriate when you do not want the session to draw on or create persistent memory. Projects are best for long-running efforts such as a business plan, course, job search, research program, or content series. Keep the Project instructions focused: purpose, audience, quality standards, preferred formats, and verification rules. Upload only material you are authorized to use. Periodically remove outdated files because stale source material can produce confidently outdated answers. Context boundary rule: put information in the narrowest scope that accomplishes the task. A one-off detail belongs in one chat. Reusable project rules belong in Project instructions. A durable personal preference may belong in memory. Confidential or unusually sensitive work may not belong in ChatGPT at all. Failure recovery: when output starts blending unrelated requirements, summarize the current goal and constraints, start a fresh chat inside the same Project, or split the work into separate Projects. Do not keep adding corrections indefinitely to a contaminated thread. Enroll to continue this lesson. The preview above shows the lesson objectives and opening lesson content. Enroll to view the full lesson, complete the practice work, and take the lesson quiz.
Lesson 1: Projects, Memory, and Context Boundaries
Lesson Objectives
Lesson Content
