AI headlines. July 25, 2026.

Saturday morning headlines. Opus 5 and Cowork were able to curate some really good articles today, so let's see what's going on in the world. First and foremost, Opus 5 was released yesterday, and just about everybody had it rolled out to their apps, and of course the web interface has it as well. So far I'm getting great results, and we'll see what everybody else thinks about it. I'm not a software engineer. I don't do a lot of full software. When I code, I do scripts, so my opinion on the coding side of things is not going to be the same as software engineers' opinions, so we'll see what happens. ChatGPT has rolled out its full duplex voice in the desktop app. You can speak while it does things in the background, so you can have conversations with ChatGPT while it's doing other things, and the demonstrations I've seen make it seem very useful. I don't have a subscription at the moment, but I might renew just to test this out. Meta and Facebook have now launched a new free Facebook verified badge. The benefit of this is you can tell who's a fake account and who's not, at least, I mean, for the most part. I'm sure somebody's going to find workarounds, but for the most part, a lot of people who are going to be verified are going to be actually real people, and this is great for Facebook because Facebook is still the social media where the most scammers live, unfortunately. So while a lot of people have concerns about privacy and things of that nature, this is not a requirement at the moment. This is just an optional feature to get verified. I like this feature. Other people may not. And while we're still talking about Meta, there is now a new feature where Meta's seller app writes your marketplace listing from a photo. So you basically just upload a photo and it writes everything for you for your ad. Seemed promising. I don't use the marketplace, but for those who do, this might be helpful to you. In Germany, Flux3 has rolled out. This is an AI that makes images and videos with sound, and it seems very promising. I haven't read too much into this, but since we need options as far as diffusion models, I like variety. I like choices. This will help, in my opinion. Many of you may have already seen this, but Jensen Huang opened a X account in his very first post. He's talking about his support for open models, and he believes everybody should support open models. A lot of people liked this posture, and a lot of people are really behind him for this, especially with what's going on with the Chinese models and the tension between the United States and China. We'll see what becomes of this. Google has put out a lot of analytics suggesting that AI is actually not taking people's jobs. At least the data they've collected from Gemini sort of suggests this strongly, and it says Google analyzed 15 million Gemini chats and found that under 10% actually automate a task, and 86% of AI use happens outside of work. So for those who are at places that use Gemini, it appears that Gemini, at least, is not replacing people's jobs. I know a lot of companies are suggesting layoffs are due to AI, but there's a lot of speculation in the community that this may not be accurate. This may be more of an excuse. Again, we'll see what happens. Moonshot and DeepSeek are now seeking IPOs. This is a touchy topic right now, because there is a lot of attention on this and what's happening there, but the way that Moonshot came about, I'm not going to make any accusations here. It seems highly probable, at least from the evidence and data, that Moonshot may have actually been distilled from either Claude or ChatGPT or perhaps both, and now both companies are preventing that from happening in the future. So while Moonshot and Kimmy are really interesting right now, the future improvements, it may not be sustainable if all of what's happening plays out the way that I think it will. I might be wrong. We'll see what happens there. And more with the diffusion models, it appears that NVIDIA, or at least NVIDIA Research, says that San NVIDIA 2.0 matches current quality and we're running 120 times faster than rivals on a single GPU. So NVIDIA is making advancement for their diffusion models. It's completely expected considering that NVIDIA is the graphics king. So this is something I'm going to watch. I'm not going to comment too heavily on this, but it seems promising. Claude's voice models now run on more capable models. Previously, I think it was limited to just Sonnet, but now they have allowed Opus and I can't see anywhere yet that it's confirmed that Fable 5 is allowed, but basically previously with voice chat you were limited to the lesser models and now apparently they're opening up some flexibility with that. So for those who use voice chat, at least with Claude, this may help you. Now this article is paywalled, but it turns out that Midjourney has acquired an astrology app. Apparently the plan is AI meets the Zodiac. They want to integrate these two. I'm not sure the direction they're going with this, but for those, because there are a lot of people out there who pay very close attention to astrology and things related, this may end up being a new experience as far as astrology and things of that nature. I just wanted to include this for those who may not be completely technical, but still want to see what's going on in the AI landscape. And one more. I suspect this headline will be immensely popular. Instagram is now banning the Metaglasses Creeps. So the people who go around recording other people and those who even try to turn off the light that tells people that they're recording, they may not be able to use Instagram the way they are using Instagram. And I suspect other companies such as YouTube and TikTok will catch on to this. It's about time. They've been allowed to do this for too long. And if you're one of the people who do do this, why? Why do you do this? Anyway, that's all the headlines I have for today. Have a great Saturday.