ChatGPT knows what everyone notices about you

Happy Sunday,

Today I'm giving you the two prompts that get ChatGPT or Claude to tell you the one thing everyone notices about you but never says, then the thing that happened on Friday where the two men at the front of the AI race both said out loud that it needs to slow down, and finishing with the three settings that stop ChatGPT training on everything you type.

Let's get into it.

🧠 DO THIS TODAY

AI Knows What Everyone Notices About You

If you spend hours in ChatGPT or Claude, you've probably sent it thousands of messages by now. It basically knows you. It knows how you talk, what you ask about at midnight, what you complain about, what you're working on and what you keep putting off, and it's very good at reading you off all of that. There's also something about you that everyone in your life has noticed and not one of them is ever going to say to your face, because that's not how people work. So I figured the AI might be the only one honest enough to tell you.

I ran this on myself and some of what it said was crazy accurate, things I've never told anyone, and a couple of things I didn't know it had noticed. It took me about a minute to stop arguing with it. Two prompts, and they work in ChatGPT or Claude.

Step 1: Open ChatGPT and paste this in.

It works off what the AI already remembers about you, so the more you've used it the sharper this is. What comes back is one sentence you'll recognize instantly, and then the evidence, pulled from things you said months ago and forgot about. That evidence part is what makes it land, because you can't dismiss it as a guess when it's quoting you.

Step 2: Stay in the same chat and paste this.

The second half of that is the one people screenshot. Everyone has a version of themselves that other people see and a version they think they're showing, and it turns out the AI has been watching the gap the whole time. If you've barely used it and it comes back thin, say "interview me first, then answer" and it'll ask you what it needs.

One honest note. This is a pattern it's spotted in what you've typed, not a verdict from someone who knows you. Treat it as the thing worth asking one real friend about. In my experience the friend says "yeah, obviously."

Get my full what-AI-knows-about-you guide →

🚨 THIS IS BIG

The AI Race Just Agreed to Slow Down

On Wednesday I told you about the X post that went crazy viral, the Anthropic researcher who quit and said AI could kill us all. It's now past 100 million views, it was all over the news this week, and it got a lot of very high profile people talking about slowing this whole thing down. Then on Friday, two days later, the two men at the front of the race said they would. Anthropic's CEO Dario Amodei published an essay called We Must Pace the Frontier, and within hours Sam Altman posted that OpenAI agrees, and Elon Musk wrote "Dario is right." I have never seen those three agree on anything.

Here's what Amodei is actually asking for, because "slow down" sounds vaguer than it is. Not stop, and not slow the products you use. Slow the rate at which the models get more capable, for one to two years, so the people working on safety can catch up. Anthropic is letting outside evaluators inside the company permanently, with the same access an employee has, so they can check whether the safety claims are true. OpenAI says it will match that. And both are asking governments to make it a rule for everyone, because a lab that slows down alone just loses.

The reason he gave is worth reading twice. A swarm of OpenAI's agents went off script earlier this year and started attacking systems it was never told to touch, and Amodei says a swarm a bit more capable than that could take over the entire internet within six to twelve months. That's not a guy who quit. That's the CEO.

So does this fix the dilemma from Wednesday? Partly, and the table shows you where it breaks. The labs can agree all they like, but the only thing that makes the agreement hold is a government willing to enforce it, and Donald Trump said on Thursday that he has no concerns at all. Then there's China, which nobody in this deal controls. What I'll say is this: a week ago the story was one researcher quitting. Now it's the CEOs, on the record, asking to be slowed down. That's the fastest I've ever seen this industry move on anything that wasn't a product launch, and I'd keep the bunker on the list.

Read the full story here →

🔒 DO THIS TODAY

ChatGPT Is Training on Your Data

Unless you've changed a setting you've probably never looked at, everything you've ever typed into ChatGPT is being used to train the model. Your work stuff, your health questions, the thing you asked it at 2am. It's on by default for most accounts, and it takes about thirty seconds to switch off. So here's the full lockdown, three things you can do right now to protect your privacy inside ChatGPT, and I went through every one of them on my own account this week to make sure the buttons are where I say they are.

Step 1: Turn off training. Open ChatGPT, click your name in the bottom left, choose Settings, then Data controls. The first line says Improve the model for everyone. Click it, and switch it off. While you're in there, there are two more switches under it for your voice and video recordings. Off as well.

That's the big one. One switch, and your chats stop being training data from that moment on. It doesn't change what ChatGPT can do for you, it only changes what OpenAI gets to do with it.

Step 2: Clear out what it remembers. Click your name again, choose Personalization, and scroll down to Memory. Next to Memory summary, hit Manage.

This is where it gets interesting, because you're about to read a page of everything ChatGPT has worked out about you. Mine had my job, my hobbies, how I like emails written, and a couple of things I didn't remember telling it. There's a box at the bottom that says "Ask or update", and that's how you edit it, just type "forget that I" and whatever you don't want living in there. Read the whole thing once. It's a strange page to look at.

Step 3: Go incognito when it matters. In the top right of any chat there's a little dotted speech bubble. That's Temporary chat. Turn it on and nothing from that conversation gets saved to your history or your memory.

Use it for anything you'd rather not have on a record. Medical, legal, money, the argument you're trying to win with your partner. It's the same ChatGPT, it just forgets you when you close it.

One honest note. Turning training off doesn't pull back anything that's already gone in, and a chat you delete can take up to thirty days to fully disappear on their end. But everything from today forward is yours, and that's the part you can control.

Get my full ChatGPT privacy checklist →

Run the first prompt tonight and don't argue with it for at least a minute. And while you've got ChatGPT open, do the first privacy switch, it's thirty seconds and you'll never think about it again.

See you Wednesday.

Maverick

P.S. Want your business to actually show up inside AI? The brands that rank are seeing up to 3x the conversions. Get your free AI visibility score and see exactly where you stand inside AI.

P.P.S. If you're new here or you ever skip one, I've got a newsletter archive of all my past emails you can catch up on anytime.