Skip to content
← All free guides
Power-user Claude 6 min read · free

Set up memory and projects so AI remembers you

Stop starting from zero every conversation.

The short answer

Tell the AI who you are and how you work once — in its memory or custom instructions — and use Projects or Custom GPTs for recurring work, so it stops needing the full briefing every time.

The biggest upgrade in how you use AI usually isn't a cleverer prompt — it's the model already knowing your context before you type a word. Memory and projects turn a brilliant stranger who needs the full briefing every time into an assistant that already knows your role, goals and voice.

Where to set this
  • ChatGPTMemory + Custom GPTs for reusable, context-aware assistants
  • ClaudeProjects — standing instructions and reference files per workspace
  • Geminisaved info and Gems for personalised assistants
Beginner

Tell it who you are, once

Write down the durable facts about you — your role, what you're working on, how you like things explained, your communication style — and put them where the tool reuses them automatically, usually a 'custom instructions' or 'memory' setting. Now every answer arrives shaped to you without re-explaining. The setup takes five minutes and pays off on every reply after.

Intermediate

Use projects for recurring work

Beyond a global profile, projects (or custom GPTs) let you set context that travels with a specific kind of work — a project for your side business carries its rules, reference material and tone every time you open it, separate from your day job. This keeps contexts from bleeding into each other and means each stream of work gets exactly the background it needs.

Advanced

Curate it like an asset

Memory is powerful because it persists — which is also its risk. An outdated fact or wrong preference will silently shape answers for months until you notice. So review what it remembers periodically and prune anything stale, and keep genuinely sensitive data out of systems you don't control. The quality of your context now determines the quality of everything the model produces for you.

  1. 1Write down the durable facts: your role, goals, preferences, voice.
  2. 2Put them in memory or custom instructions so they apply everywhere.
  3. 3Use a Project or Custom GPT for each recurring kind of work.
  4. 4Review what it remembers occasionally and prune anything stale.

Setup prompts

promptHere's a short profile of me and how I like to work [paste]. From now on keep this in mind, and tell me if anything I ask conflicts with it.
promptInterview me with 8 questions to build a 'how I like to work' profile I can paste into your memory/custom instructions.
promptWrite the standing instructions for a Project that always helps me [recurring task] — role, rules, format, and an example of good output.
promptHere's how I write [paste 2-3 samples]. Summarise my voice into rules you can follow so future drafts sound like me.
promptReview what you currently know about me and flag anything that seems outdated or that I should update.
promptSuggest 5 separate Projects (or Custom GPTs) I should set up based on the recurring work I describe here, and the standing instructions each one should contain: [describe my work].
💡 Keep its memory tidy. An assistant that remembers the wrong things is worse than one that remembers nothing.

Want the prompts that go further?

Drop your email + WhatsApp — I'll send the weekly playbook and the full prompt vault.

Prefer to ask first? Message me on WhatsApp · see plans

More guides like this