ChatGPT was down because we made too many Studio Ghibli pictures lol #chatgpt #ai #news #2025 #future #story #llm #mentorship #learnontiktok #learn

So it wasn't just you, ChatGPT was hard down today, it came back up. Why? Because so many people were signing up for ChatGPT because they wanted to make images with the new image model with ChatGPT 4.0, that's why. Sam was sharing that they got a million users signed up in just an hour yesterday, so it's been like an absolute flood, hard to keep the servers up. I think one of the things I take away from that is that we have spent the last year plus here on TikTok emphasizing how big a deal AI is, and mostly what we've talked about is language, language in code, language that you can talk and speak back and forth like English or French or German, and images are like 10 times bigger as a use case. Getting them right is enormous, and this is the first model that has gotten them right in the sense that it can do photorealism, you can edit things with text, you can write text right in the image, I've showed a few examples here on TikTok. It's become obviously easy, which is the usual bar for massive product adoption. And because of that, it means that ChatGPT has hit another steepening of their growth curve. They're growing much faster now because they've been able to put images on the table as well as language. Now you might say, and you're right, this is not the first image model. We've had image models for a while. Why did Midjourney not have this moment? Why did the Gemini model for images not have this moment? I will give you two answers. The first is, winners win more, and that's just how software works. They have 400 million active users. When they update something, and it's really good, a lot more people notice, tell their friends, bring more people onto the platform, share more, etc. That's boring reason number one. The more interesting reason is because it actually is better in ways that matter to users. Midjourney is not great at text. The new Gemini image model is not as good at inferring intent from a prompt as ChatGPT is. In fact, there's some evidence that ChatGPT takes your prompt, and part of how it infers well is it quietly expands your prompt on the back end to make it a richer text prompt that it then feeds to the image engine. And so I did like side-by-side comparison. I looked side-by-side at Gemini. I looked side-by-side at ChatGPT. You can look at it on the sub stack. I think I wrote it about a week ago. And it is very clear that prompt fidelity, that understanding the prompt is something ChatGPT does very well. And I think if you do that well, it unlocks a lot more trust from users because they know if they give you a prompt to make an image, it's going to come back exactly the way they want it. And my feed is full of people who have built incredible ads with long, like two and a half paragraph prompts for an image generator that specify the aspect ratio and exactly the angle of the camera and the lighting and the style. And if it's prompted like that, it can produce an incredible image. And even for casual use, it produces lots of fun Studio Ghibli pictures. And that's why ChatGPT was down. And that's a good prompt to have. It's a good prompt to have the more people are hearing about AI and understanding it. And if images are a way to do that, that's great.