|
Happy Sunday,
Today I've got the craziest ChatGPT use case I've seen all year, how ChatGPT can find lawsuits that owe you money, why people are changing the way they write so AI detectors don't flag them, and how to have Claude answer your Instagram DMs for you.
Let's get into it.
|
IN THIS EMAIL
» ChatGPT Found a Lost Earbud in a Pile of Gravel
» ChatGPT Finds Lawsuits That Owe You Money
» People Are Changing How They Write So AI Detectors Don't Flag Them
» Claude Can Now Answer Your Instagram DMs For You
|
🚨 THIS IS WILD
ChatGPT Found a Lost Earbud in a Pile of Gravel
This is one of the craziest ChatGPT use cases I've ever seen.
A dad was leaving the park with his son just as the sun was setting. His son took off toward the parking lot, so he sprinted after him, and on the way his white Bose earbud fell out of his hoodie pocket into a huge patch of white gravel. White earbud, white rocks, getting dark. He searched for a solid 10 minutes and couldn't find it.
So he took a photo of the gravel, uploaded it to ChatGPT and asked, "Can you find my headphone? It's a white Bose earbud."
ChatGPT zoomed in and scanned the photo, and on the first try it couldn't find it. But instead of making something up, it told him it wasn't sure and asked for closer photos. So he took a few zoomed-in shots and sent them over. This time ChatGPT sent the photo back with a red circle around the exact spot. The whole thing took less than five minutes, and his Reddit post about it blew up.
I posted this one on Friday and it's already at over 2.4 million views, and I get why. It's the kind of thing that makes you realize ChatGPT can actually see, not just read and write.
For those of you who might not understand what's happening here, let me explain. When you upload a photo, ChatGPT doesn't just glance at it the way you do. It breaks the picture into small sections and zooms in on each one, and it knows what an earbud looks like from every angle. So it looks for the thing that doesn't belong: one smooth, curved, shiny piece of plastic in a sea of rough, jagged rocks. Your eyes see "white" everywhere and give up. ChatGPT checks every inch for that one shape. And when it wasn't confident, it asked for better photos instead of guessing. Just check whatever it circles against your original photo, because it can occasionally get it wrong.
I know this might be an obscure use case that you'll only need once in a while. But it gets you thinking about what else you can do with this kind of intelligence. Spotting the one thing that's off in a photo is the same skill as catching the scratch on the rental car before you drive off, finding the mistake in a long receipt, reading the tiny print on a label, or checking a photo of your room for the outlet you need. If you can take a picture of it, you can ask ChatGPT about it.
Watch the video →
💰 DO THIS TODAY
ChatGPT Finds Lawsuits That Owe You Money
There's a good chance a company owes you money right now, and you have no idea.
Companies settle class action lawsuits all the time. Your phone carrier, a store you shopped at, an app that had a data breach. The money goes to anyone who qualifies and files a claim, and most people never claim because they never even hear about it. So I asked ChatGPT to find every settlement I qualified for, and I made $452 from settlements I never knew existed.
Here's how to do it.
Step 1: Connect your Gmail. In ChatGPT, click your name in the bottom left, then Settings → Integrations → Plugins → Browse directory. Find Gmail, click + and sign in.
Step 2: Open a new chat and paste this:
Go through my email and find every company I've bought from or had an account with, then match them to every open class action settlement. |
ChatGPT goes through years of your receipts and accounts and lines them up against every lawsuit that's paying out right now.
Step 3: Stay in the same chat and ask:
Which of these do I actually qualify for, how much could I get from each one, and when's the deadline? |
Step 4: Now here's the trick. At the top of ChatGPT, switch from Chat to Work, and say:
Go file the claim for every single one I qualify for. Only use the official settlement website for each one, and stop before you submit each claim so I can check it. |
ChatGPT opens its own browser and fills out every claim form for you while you just sit there. Most only ask for your name, address and email, and the proof of purchase is usually already in your inbox.
And here's the bonus. Say "Check for new settlements every week and tell me the second I qualify for one," and ChatGPT will keep watching for you.
A couple of quick rules. Only claim what you actually bought or used, because you're confirming your answers are true. Filing is always free, so if a site asks you to pay, or asks for your Social Security number or bank login, it's a scam. And payouts take a few months, since the court has to approve everything first.
Get all 7 settlement prompts, including my scam checker →
🚨 THIS IS BIG
People Are Changing How They Write So AI Detectors Don't Flag Them
AI detectors are now deciding whether people's writing counts as human, and it's changing how everyone writes.
Bloomberg just published a big story on this, and it starts with a book. Earlier this year, Hachette, one of the biggest publishers in the world, pulled a horror novel called Shy Girl right before it came out, after an AI detector called Pangram said 78% of it looked AI-written. It's believed to be the first novel from a major publisher ever pulled over AI. The author, Mia Ballard, denies using AI and says the flagged passages came from an editor she hired. But by then the internet had already made up its mind.
And now the regular people are reacting. Writers and students are changing how they write so a detector won't flag them. The em dash (that long dash —) is the big one, because AI uses it so much that detectors treat it like a red flag. People have shown that swapping out a single em dash can flip the exact same text from "AI" to "human."
I'm not going to lie, I think this is kind of crazy. We've gotten to the point where humans are editing their writing to sound less like a robot. You might have noticed I never use em dashes in this newsletter, and honestly, that's a big part of why.
Here's my take. An AI detector is a guess, not proof. It looks at patterns and gives you a percentage. That's useful, but it shouldn't be the only thing that decides whether a student fails a class or a book gets pulled.
So here's what I'd do if you write anything that matters, for school or for work:
Write it in Google Docs. It saves your version history automatically (File → Version history → See version history), which shows every edit you made over hours or days. That's your receipt if anyone ever says a machine wrote it.
If you use AI, be upfront about how. "I wrote it and used ChatGPT to check the grammar" is a totally different thing from "ChatGPT wrote it," and saying so up front beats having to defend yourself later.
Read Bloomberg's full story →
🛠️ BUILD THIS
Claude Can Now Answer Your Instagram DMs For You
You can now connect Claude to your Instagram and have it reply to every single DM you get.
It works through a tool called ManyChat, which is officially approved by Meta, so you don't have to worry about your account getting banned. You connect Claude once, tell it who you are and how you want it talking to people, and it answers your DMs in seconds, 24/7.
Here's how it works:
Step 1: Your Instagram needs to be a Business or Creator account (it's a free switch in Instagram Settings → Account type). Sign up for ManyChat and connect your Instagram. You'll need ManyChat Pro for the AI part.
Step 2: Get a Claude API key. Go to platform.claude.com → API Keys → Create Key and copy it. It's separate from your Claude subscription and you pay by usage, so add $5 or $10 of credit and set a monthly limit.
Step 3: In ManyChat, go to Settings → Integrations, paste your key into the Claude card and click Connect.
Step 4: Build an automation that runs on any DM or comment, add an action block and pick Claude. Then write out exactly who you are, what you do and how you want it talking to people. This is the part that actually makes the difference, so take your time on it.
Step 5: Now here's the trick. Before Claude sends anybody the thing you promised them, have it check whether they're following you. If they're not, it asks them to follow first, then sends it. Now every DM is a chance at a new follower.
And it's way more than freebies. You can have it answer the same 10 questions you get every day, recommend the right product for someone, book calls straight into your calendar, qualify leads for your business or handle basic customer support, all while you're asleep.
If you're a creator or you run a business on Instagram, I highly recommend ManyChat. I've partnered with them, and we've got some big things coming in 2027 that I'm really excited about. ManyChat is honestly probably how you ended up on this newsletter in the first place. Which means if you've ever DMed me on Instagram and gotten a reply in two seconds... I hate to be the one to break it to you, but that wasn't me. Sorry.
Get my full step-by-step setup guide →
Connect your Gmail and run the settlement prompt this week, because some of these deadlines are coming up fast. And if it finds you money, reply and tell me how much. I want to know who beats my $452.
See you Wednesday.
Maverick
P.S. Interested in sponsoring the newsletter and putting your product or service in front of 265,000+ people? I'd love to feature you. Sponsor an issue today.
P.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.
|
