|
Happy Wednesday,
Today I'm showing you how to find the money the government is quietly holding with your name on it (I did this and it found 500 dollars I had no idea about), then walking anyone who sells online through connecting their Shopify store straight into Claude in about thirty seconds, and finishing with the thing Anthropic just did to every single word Claude writes for you, which almost nobody has noticed yet.
Let's get into it.
|
IN THIS EMAIL
» The money the government secretly owes you
» Claude can now run your Shopify store
» Every word Claude writes is now watermarked
|
💰 DO THIS TONIGHT
ChatGPT Can Find Money the Government Secretly Owes You
Ten days ago I had you digging through your own inbox for forgotten credits. This one's better, because it isn't money you were sent and forgot about. It's money you never knew existed. I ran it on myself and it found 500 dollars from a deposit on a place I lived years ago.
Here's the part I want you to know. When a company can't find you, it isn't allowed to keep your money. It has to hand it over to the state, and the state holds it for you forever. Old deposits, a final paycheck sent to an address you'd already left, a refund on a closed account, dividends on shares someone bought you as a kid. Roughly 1 in 7 people in the US has money sitting in one of these databases, and states paid back over 4.49 billion dollars in a single year. That's just the people who thought to look. So naturally I did the only sensible thing and made ChatGPT go and look for me.
Step 1: Open ChatGPT and paste this in. Use ChatGPT for this one, I've noticed other models can get a bit sensitive about digging around your name.
|
Find any unclaimed money or unclaimed property being held in my name. Search the official government database for my country and region, plus every other place I have ever lived, and give me the exact link for each. Then check the places nobody thinks to look: retirement money left at old jobs, undelivered tax refunds, savings bonds, money from banks that failed, unpaid final wages, and any class action I might qualify for. Search my full name, my name without my middle name, common misspellings, and any other name I have used. Tell me exactly what to type into each site and what a real result looks like versus a dead end. My name is [YOUR FULL NAME], the places I have lived are [LIST THEM], and the companies I have worked for are [LIST THEM].
|
The old addresses are the whole game, and it's the bit everyone skips. Your money doesn't follow you. It stays with the state where that company last had you on file, so if you've moved twice there could be three separate piles of it. List every place, even the one you lived in for four months in your twenties.
Step 2: Finding it and getting it are two different things, and the forms are miserable. So let it fill them in. Switch ChatGPT over to Work and paste this in exactly as it is. The @browser at the front is what sends it out onto the actual websites:
|
@browser Now open each of these databases and actually run the searches for me. Where you find something in my name, start the claim and fill in everything you can. Stop and ask me before you submit anything and before you type anything sensitive. When you're done, give me one list: what you found, what it's worth, what stage each claim is at, and what you need from me to finish it.
|
One thing that matters more than anything else here. Every one of these databases is free, and you never pay anyone a cut to get your own money. There's a whole industry of "finders" who'll happily take 20% for typing your name into a website. If a site wants a percentage or a fee up front, you're in the wrong place.
And if you're outside the US you've got exactly the same thing under a different name. The UK has My Lost Account plus the government's pension tracing service, Australia has ASIC's Moneysmart with around 2.7 billion dollars sitting in it, Canada has the Bank of Canada's unclaimed balances register. Ask for yours by name in that first prompt.
Get my full unclaimed money guide →
🛒 THIS IS A CHEAT CODE
Claude Can Now Run Your Shopify Store
Claude connects straight into Shopify now, and if you sell anything online this is a genuine cheat code. Not "ask it for advice" connected. Actually connected, working on your real store:
|
+ Add new products
|
+ Create discount codes
|
|
+ Browse recent orders
|
+ Pull up customer details
|
|
+ Check your inventory
|
+ Run real analytics
|
|
All of it in plain English, all of it against your actual numbers instead of generic advice about e-commerce. Here's how to set it up, it takes about thirty seconds.
Step 1: Head to Claude. On the left, choose Customize, then Connectors.
Step 2: Click Add, then Browse connectors. Search for Shopify and hit Connect.
Step 3: Log into Shopify and choose which store you want connected. That's it, your store is officially inside Claude.
Step 4: Now you just talk to it. Personally I think this is best used for pattern recognition, because that's what AI is genuinely best at in the world. Once it has your actual numbers it starts finding things in your own store you were never going to spot yourself. This is the first thing I'd ask:
|
Based on my store's recent performance, find five ways I can increase my conversion rate that will lead to more sales and more total revenue. Use my real numbers, not general e-commerce advice. For each one, show me the data in my store that made you say it and rank them by how much money they're worth for how little work they take. Don't change anything in my store, just show me.
|
It reads the real data out of your store and comes back with changes that are specific to you. It's basically having a business partner who's already read every number you've got and never gets bored of being asked. And from there it's your call. You can go and make the changes yourself, or hand them straight back to it and let it make them for you.
Which is worth knowing before you start, because this connector doesn't just read your store, it can change it, including creating live products and live discount codes that real customers can use. So while you're getting a feel for it, end your prompts the way I did up there with "don't change anything, just show me."
Get my full Shopify setup guide →
🚨 NOBODY IS TALKING ABOUT THIS
Every Word Claude Writes Now Has an Invisible Watermark In It
Anthropic just quietly started hiding an invisible watermark inside everything Claude writes. Not a label at the bottom you can delete. Not hidden info in the file that vanishes when you copy the text out. The mark is built into the words themselves, so it travels with them when you copy and paste, and it can survive a fair bit of editing after that.
It's on every new model from the start of this month, they're going back and adding it to the older ones, and it's everywhere Claude exists. The website, the app, Claude Code, Cowork, all of it. You'll never see it or feel it, they say it doesn't change the meaning, quality or readability of anything. The reason it happened is Europe's new AI rules, but instead of doing it just for Europe they switched it on worldwide, so wherever you're reading this, it applies to you.
Now here's the part people online are getting badly wrong this week. There is no detector. Not yet, and not for anyone. Anthropic say they're building tools to read the watermark but haven't released them and haven't said when. So no teacher, no boss and no client can run your writing through anything today and get an answer. And when it does land it'll be blunter than it sounds, because heavily rewriting it, paraphrasing, translating, screenshotting, or writing something very short all strip the mark. They also admit finding it doesn't prove Claude wrote the thing, since plenty of people write something themselves and hand it over to tidy up. It's a signal, not a verdict.
At this point I think this is probably the right move, and I'll say the awkward bit out loud, because I'm the guy who teaches you to make AI writing sound human. Am I contradicting myself? I don't think so. I teach that because raw AI output is generic and sounds nothing like you, and because AI detectors are wrong constantly and have got real people in real trouble over work they actually wrote. A company being upfront that its own model wrote something is a different thing entirely. What I'll be watching is what people do once the detector exists, because "a machine touched this" is going to get treated as "this person cheated", and those aren't the same sentence.
Worth noting Google already do a version of this inside Gemini, and OpenAI have reportedly had an accurate detector finished for two years and chosen not to release it. Make of that what you will. Either way, the move for you is the same one I've been banging on about for a year. Use AI to think and to draft, then rewrite it in your own voice so it actually sounds like you.
Read exactly what they're marking →
Do the unclaimed money one tonight. Ten minutes, costs nothing, and worst case you find out you're all square with the government, which is a nice thing to know anyway. And if you've got a store, connect it before Monday so you've got a week of real numbers to point at.
See you Sunday.
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.
|
