Teach ChatGPT Who You Are: Memory & Custom Instructions

Out of the box, ChatGPT is a permanent day-one assistant: every chat starts with a stranger, and you pay a re-explaining tax on every task — or skip it and get the generic answer. This course removes that tax for good.

You'll set up the personalization stack layer by layer: pick a base style, write standing instructions that pass the stable-general-consequential filter, and learn how memory actually works — the auditable notepad of saved memories versus the impression it forms from your chat history. Then the ownership half: open the list of what it has quietly remembered about you, prune it with four flags, draw a fence around what should never go in, and use temporary chats for sensitive work — or for the days you need fresh eyes instead of your own groove. You finish with a fifteen-minute monthly routine that keeps the whole system honest.

What you'll learn

• Why every chat starts from zero, what the briefing tax costs you, and the day-one test for what belongs in a standing setup

• How to write custom instructions that sharpen every answer — and why stuffing the boxes dilutes them

• How ChatGPT's two memories actually work: saved memories you can read and edit, and the chat-history impression you can't

• How to audit and prune what it remembers with the four flags: stale, wrong, sensitive, noise

• What should never enter memory, and how temporary chats keep sensitive work off the record

• When personalization backfires — and the three ways to override your own setup

• The fifteen-minute monthly review that keeps your ChatGPT accurate as your life changes