Why your AI needs a memory (and the easy way to give it one)

Why bother? (Read this part even if you skip the rest)

Picture a super helpful assistant who forgets you every single morning. "Hi! Who are you? How do you like your coffee? What's a 'spreadsheet'?"

Every day, you explain it all again. They're still smart. They just aren't your assistant.

That's an AI with no memory. And here's the sneaky part: it feels fine at first. In week one, you're just playing around. By month two, you use it for real stuff. And it keeps:

• writing in the wrong tone,
• asking things you've told it five times,
• forgetting your rules, like "don't send anything without asking me,"
• making the same mistake you already fixed.

So people decide the AI "got worse." It didn't. It just never wrote anything down.

Memory turns a smart stranger into your assistant. Keep it simple, and you get three things:

1. Less repeating yourself. It already knows your job, your people, and how you like things.
2. Better results. Answers fit your life, not some made-up average person.
3. Safer behavior. Your ground rules, like "ask before deleting," follow it into every task.

Maybe you've seen fancy "second brain" setups online, with special apps and webs of linked notes. You don't need any of that. The simple version below gets you most of the good stuff.

The easy version: one note, two minutes

AI memory comes in two flavors:

• What it picks up on its own. Handy, usually already on, and you can check it.
• What you write down for it. This is the part you can count on. It's also the part most people skip.

So write one short note. Give it to your AI. Then update it whenever you catch yourself fixing the same thing twice. Anthropic gives its power users the same tip: write it down when you'd otherwise explain it again.⁵

Your "About me" note (copy it, fill it in, keep it under a page)

About me: [your job, what you use AI for]
Formatting: [e.g., short paragraphs, US spelling, Word files not PDFs]
Never: send, buy, or delete anything without asking me. Don't touch files outside the folder I give you.
People and places: [names the AI should know]
Current projects: [1–3 lines each]

One big rule: never put passwords or account numbers in it.

Where to put it, app by app

Claude (Chat and Cowork)

• Auto memory is on by default for Free, Pro, and Max. Check it at Settings > Memory → "Generate memory from chats."¹
• See what it remembers: Settings > Memory shows what Claude knows, topic by topic. You can edit or delete any of it.² Touchy topics, like health or beliefs, aren't saved unless you turn that on.²
• Chat and Cowork share one memory when Cowork runs in the cloud.¹
• Your note: paste it into Settings > Cowork → Global instructions → Edit → Save. In the new "one Claude" version, it goes in Settings > General → Instructions for Claude.³
• Want one private task? Turn Memory off in the + menu before your first message. Or use an incognito chat.¹⁴
• On a Team or Enterprise plan? Your admin decides if memory is available. It stays off until you turn it on.¹

Claude Code

Code keeps its own memory in plain files on your computer.⁵

• CLAUDE.md: a plain text file of instructions in your project folder. Claude reads it at the start of every session. In the Code tab, just ask: "Create a CLAUDE.md with instructions for this folder." Paste in your "About me" note.⁵⁶
• Auto memory: Claude saves its own notes when you correct it. It's on by default. Type /memory to see, edit, or delete them. Say "remember that…" to add one.⁵
• Keep CLAUDE.md short. Anthropic suggests under 200 lines. Be specific.⁵

ChatGPT

• Go to Settings → Personalization → Memory. Depending on your account, you'll see Reference saved memories and Reference chat history.⁸⁹
• Saved memories are things you told it to remember, like "Remember that I'm vegetarian." Chat history is what it guesses from past chats, and that changes over time.⁸⁹
• Manage memories lets you see and delete them. Heads-up: deleting a chat does not delete a memory saved from it.⁸⁹
• Temporary Chat doesn't use or update memory.¹⁰
• Your note: tell ChatGPT "Remember this about me:" and paste it. Or put it in a Project, which keeps related chats, files, and instructions together.⁹¹¹

Grok Bot

• Each Bot keeps its own preferences, its role, and summaries of past work. Other Bots don't get it.¹²
• Your note: put the key rules in the Bot's description when you make it.¹³ When you correct it, spell it out: "Use this format for future weekly reports…"¹³
• Shared files: all your Bots share one cloud computer and its files. So a "notes" file saved there is visible to every Bot.¹²¹⁴
• For big decisions, ask the Bot to check the real source. Don't lean on memory alone.¹²

Your 5-minute monthly memory check

1. Open each app's memory screen. Delete anything old or wrong.
2. Update your "About me" note.
3. Make sure nothing private, like account numbers or passwords, is stored. Passwords never belong in memory.