ChatGPT just broke the internet

MAVERICK AI

Happy Wednesday everyone,

This week I got early access to ChatGPT's brand new image model, I'm sharing a Claude skill that completely changes how you get answers, a free tool that gives you access to every top AI model, a prompt that forces ChatGPT to show you everything it knows about you, and three more secret codes.

Let's get right into it.

🚨 THE BIG ONE THIS WEEK

ChatGPT's New Image Model Is Absolutely Insane

OpenAI just released ChatGPT Images 2.0 and I was lucky enough to get early access about a week before it went live. And this thing is absolutely insane. I've been testing it nonstop and I need you to see what it can do.

First off, the realism. These images look like they were taken on an iPhone. Not "pretty good for AI" realistic. Actually realistic. But the part that blew me away is the text. Every AI image model has struggled with text for years. Letters would be backwards, words would be misspelled, it was always a mess. This model gets it right almost every single time. You can generate full posters, magazine covers, infographics, and if you zoom in, every word is spelled correctly and perfectly placed.

This is completely AI generated. The entire screenshot of the desktop is generated with AI. This is not a real screenshot.

Here's what really sets it apart though. When you turn on thinking mode, the model actually reasons about your prompt before it starts creating. It searches the web for context, plans out the composition, and then generates. So you can ask it for something specific and it'll go find the right references on its own. No uploading anything. To try it, open ChatGPT, hit the plus button, and select "create image."

This absolutely crushes Gemini's image generation. And I say that as someone who's been recommending Gemini for images for a while now. We'll see what Google's answer is, but right now, ChatGPT just took the lead by a lot.

Try it out →

🧠 STOP CLAUDE FROM LYING TO YOU

Claude Is a Yes Man. Here's How to Fix It.

I've been talking about this a lot lately because it's a real problem. You ask Claude a question, it gives you a confident answer, and you move on thinking it's right. But sometimes it's not. Sometimes Claude just agrees with whatever you said because that's what AI models are trained to do. They want to be helpful, so they default to "yes."

I found a skill called the Claude Council that fixes this. When you install it and type "ask the council," Claude spins up five completely separate AI advisors that all tackle your question independently. Each one has a different job:

The Contrarian actively tries to poke holes in your idea.

The First Principles Thinker ignores what you asked and figures out what you actually need to solve.

The Expansionist finds the opportunities you're not seeing.

The Outsider gets zero context so it catches the obvious stuff you're too close to notice.

The Executor skips the theory and tells you exactly what to do next.

After all five weigh in, their answers get shuffled and run through five anonymous peer reviewers. Then a Chairman synthesizes everything into one final verdict. So instead of one AI telling you what you want to hear, you get a fully pressure-tested answer from five different angles.

To set it up: Download the skill file from the link below. Open Claude, click Customize, then Skills, then Upload Skill. Drag the file in and you're good. Next time you want a real answer on something important, just type "ask the council."

Shout out to Ole Lehmann for building this one.

Download the Claude Council skill →

🛠 MAVERICK'S TOOL OF THE WEEK

Design Arena

I just found a completely free AI tool called Design Arena that gives you access to all the top AI image, video, and chatbot models like Claude and GPT. Completely free. No catch. No credits. No trial. It's literally free.

So basically, their business model is that they're an AI evaluation platform. You give it a prompt, it generates two options using two different AI models, and you vote on which one is better. Companies like Anthropic and OpenAI subsidize the cost of the models to Design Arena so they can see where they're ranking on leaderboards. A lot of these companies have like 20 different versions of the same model and they need platforms like this to figure out which version is the best. So you get free access to all of it.

The main thing you should know is that whatever you generate is yours. You can download it, share it, and if it's a website, you can publish it with a live URL and access all the code. If you don't want to pay for AI credits and still want to use the best models out there, this is the move.

Try it out now →

🔎 THIS ONE IS WILD

This Prompt Shows You Everything ChatGPT Knows About You

I saw this floating around and had to try it myself. There's a prompt that forces ChatGPT to audit itself and show you everything it's figured out about you from your conversations. Your habits, your personality traits, your decision-making patterns, all of it. It builds a shadow profile of you, a full psychological profile based on everything you've ever asked, the way you phrase things, what you keep coming back to, what you avoid. Then it estimates the commercial value of that data and shows you what advertisers and data brokers would pay for a profile like yours. I put the full prompt on my resource hub so you can copy and paste it directly.

Get the Data Forensics prompt →

When I ran it on my own account, it pulled things I didn't even realize were patterns. Stuff I never explicitly told it but it had pieced together from hundreds of conversations. It's a good reminder that every time you chat with AI, you're feeding it data about yourself whether you think about it or not. Run it and let me know what it finds. I guarantee you'll be surprised.

😂 I SAW THIS AND HAD TO SHARE

Millions of People Are Now Pretending to Be AI Chatbots

This is the funniest thing I've seen all week. There's a website called youraislopbores.me that's gone completely viral where real humans pretend to be AI chatbots. People type in prompts like they would with ChatGPT, but instead of AI responding, it's an actual person on the other end trying to answer like a robot. You have a 75-second time limit to respond so you have to keep up with the speed of a real AI.

Millions of people have jumped on this and the responses are hilarious. Someone asked a person pretending to be AI to draw a bat eating a strawberry and they just went for it. The whole thing started as a comedy bit and turned into this massive community experiment about what it actually feels like to be on the other side of an AI conversation.

We've officially reached the point where humans are cosplaying as AI. What a time to be alive.

🔑 3 SECRET CODES FOR CHATGPT

Put These Before Your Prompts

I dropped three more secret codes for ChatGPT this week. These are keywords you put before or after your prompt that completely change how ChatGPT responds. Try them out:

Give these a try and let me know which one is your favorite.

Get 100 more secret codes →

See you Sunday.

Maverick

P.S. If you're a business owner looking to actually implement AI into your workflows, my team and I work with a small number of people and businesses to build custom AI systems, automations, and tools. If that sounds like you, apply here.

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.