You nailed the perfect prompt. The output sang. You saved your Super Prompt. Then, you opened a new chat the next morning and AI acted like you'd never met.

• It's not broken. You just haven't taught it to remember YOU, Jim VandeHei writes.

Why it matters: Master what AI stores, what it can reference from your past and how to direct both — and your chats will start smarter. This is how you unlock next-level prompting and results.

• The basics take 10 minutes. The payoff compounds forever.

A quick primer on AI memory: Think of it as two layers.

1. Inside a standalone chat thread, AI can use what you've said there. That's called working memory. Close the chat, that brain resets.
2. Lasting memory is different. It's what AI remembers about you — your job, your style, your preferences. Most people accidentally rely only on working memory and wonder why every chat starts cold. Don't.

• There's nuance to how memory works on different platforms. Easy trick: Just ask the LLM to tell you how its memory works and how to get the most out of it based on your specific needs.

🫡 Teach it on the way out. You can explicitly flag what matters to your AI's saved memory. End important chats by telling the robot explicitly what to keep — and what to ignore.

• The prompt: "Save this preference for future chats: I'm CEO of Axios. I write a Saturday newsletter for CEOs in Smart Brevity. I prefer short, punchy paragraphs with bold labels and concrete stats."

🔍 Audit the AI's file on you. You can see the memories AI has stored about you — and edit them. Read them like an HR file on yourself. Delete what's wrong, sharpen what's vague, add what's missing. Do this every few weeks.

• The prompt: "What saved memories do you currently have about me? List them, then suggest what I should edit, delete or add."

🔮 Mine your past. This is a new, underused move. ChatGPT and Claude (on paid plans) can search past chats if you enable the ability in your settings. That's not a chatbot — that's a coach who can review your recent thinking and spot patterns you can't.

• The prompt: "Look across my recent chats, especially the last 30, and tell me what patterns you see in how I think, write, get stuck — or repeat myself?"

🎓 Graduate to workspaces. This is where it gets powerful. Projects (in ChatGPT and Claude) and Gems (in Gemini) are lasting workspaces built around a single topic. Anything you revisit more than twice deserves one — not a standalone chat.

• You can add specific files and write the rules in plain English. AI can help with that.
• For my CEO newsletter project, I dropped in every column I've written, my book on leadership and speeches I've given. Then I wrote out the audience, the tone, the length, plus data sources to use and ones to avoid.
• The more useful context you add — and the more work you do inside the project — the sharper the output.

The bottom line: AI without context is a stranger. AI with memory is a colleague who gets sharper over time. Spend 10 minutes today teaching your AI who you are.

📬 Send Jim the weirdest (or smartest!) thing your AI has ever remembered: [email protected]. ... Share this column.

• 📈 If you're a CEO or on a CEO's team: Ask to join Jim's new weekly Axios C-Suite newsletter.

Axios San Diego celebrated the opening of Torrey Pines' scenic coastal trails this spring after months of closures.

• If you go: There are several trails to explore that don't make you work too hard for panoramic ocean views.

The steep hike down to the rocky beach is worth it, Axios San Diego's Kate Murphy writes.

• Get Axios Local.