Lesson Content

Long AI sessions can become messy. Organization commands help, but they also create risk if the student does not understand what is being copied, summarized, or authorized.

Forking a chat is useful when you want to branch the conversation. For example, one branch might continue debugging while another explores a safer design. A side chat can help ask a quick related question without interrupting the main task. Starting a new chat can be better when the old chat contains stale instructions, emotional noise, wrong assumptions, or too many unrelated topics.

Compacting context is useful when a long chat needs to keep moving. The risk is that a bad summary can preserve the wrong details or drop important constraints. Before relying on a compacted session, ask the assistant to restate the current goal, protected files, tests, and definition of done.

Memory controls require extra care. A memory can help the assistant remember stable preferences, but students should avoid saving secrets, private client details, medical information, passwords, or temporary anger as durable project truth. Feedback commands are for reporting product issues; students should remove sensitive material from screenshots or logs when possible.

Approval commands are serious. If automatic review blocks a command, that block may be protecting files, privacy, or system stability. Students should not approve a denied action just because the assistant asks. They should read the command, understand why it was blocked, and ask for a safer alternative when unsure.

Professional Use Case: A developer is halfway through a plugin bug fix. The chat is long and contains old failed ideas. The developer compacts context only after writing a short checkpoint. Then the developer asks the assistant to repeat current scope, changed files, test plan, and what must not be touched. The developer refuses to approve a broad delete command and asks for a narrow file-by-file cleanup plan.

Student task: Write a "session recovery checklist" for yourself. Include when to fork, when to start a side chat, when to compact, what not to store in memory, what to remove before submitting feedback, and what you must understand before approving a blocked action.

?This is a personal operating checklist, not a technical essay. Keep each line short enough that you would actually use it during stressful work. Sample clarification prompt: "Make this checklist more practical for a tired student. Keep it direct and include one warning about approving commands."

Starter scenario: "A chat has been working on three projects at once, and now the assistant sounds confused about which site it is editing."

Starter clarification prompt: "Help me decide whether to fork, compact, or start a clean chat. List the information I should preserve before moving."

Troubleshooting: If a compacted chat forgets an important rule, restate the rule clearly and ask the assistant to confirm it. If a forked chat repeats old mistakes, check whether the mistake was already in the copied history. If an approval request is unclear, ask what the command does, what could go wrong, and what safer command could achieve the same goal.

Quality Rubric:

• The checklist protects privacy and project scope.
• The student distinguishes organization from execution.
• Approval is treated as a review step.