New ChatGPT Library Explained & More AI News You Can Use

Welcome to yet another week in generative AI. We're going to be looking at various stories, including a ChatGPT update where they added a library tab. It's a real upgrade to how it handles all the different Word documents, Excel sheets, and how to work with them. Then we have more Claude updates like every single week. It's insanity. They're just shipping. Now there's an auto mode where it just works by itself while trying to avoid the risky actions and updates to Google's AI Studio where they made it a more user-friendly and capable platform than ever. So the stories from all the big players and much more in this week's episode of AI News. You can use to show that pulls together all the releases in the AI space. And we filter for the ones that we think matter and that are going to be useful to you. Let's begin. And starting with this ChatGPT feature is a good idea because honestly, I've seen a change in my behavior with all of these LLMs over the past months. And that's the fact that I'm drafting way more documents like Word documents or Excel sheets. For PowerPoints, I still generally go to Gamma AI, which introduced a new feature. We'll talk about that later. But for Word docs and Excel sheets, these models have just become good enough for it to be really worth it to start with an LLM. And I love that ChatGPT clearly sees this, and they introduced a new feature solely focused on these. It's a library of your various documents. And while that might sound simple on the surface, I think it's a bit of a shift in thinking, especially when it comes to managing context, which, as you might know, is one of the essential things that we're doing when working with AI. And before we get into it, what is this? Well, It's a new tab here on the side that collects various files you've been using. So if you took a screenshot and added it, it'll appear here. But also if you create something like a personal finance tracker as an Excel sheet, it too will appear in this library. As you can see, this doesn't work retrospectively. It only works for all files starting March 25th. But here is the fun thing and the use case I want you to be aware of. Because obviously this is an easier way to get to these files. Nice, I can go in here and just download it from here rather than looking for the chat. That's good. But I think the amazing thing is that you can kind of select this and start a brand new chat with it. Now, obviously, this personal finance tracker here is empty. But where my mind immediately went here is that I would be adding some of my essential context files that I work with regularly in here. So then it's really easy to get to them and to reference them when I need them. So for example, I have this personal DNA file, which tells ChatGPT about my identity and different goals. And what you'll also notice is that they changed the interface that these files are displayed. Before you couldn't even open up a Markdown file. Now it shows the fully formatted thing. It's gorgeous. And I could directly chat with it in here. So it's just a shift in ChatGPT. And therefore all other LLMs will also be following this, where we're really moving away from just walls of text. And we're moving into different types of outputs like Excel sheets, Word documents, infographics, interactive visuals, like we looked at in last week's video, rather than just text. And I think most people should really shift their workflow away from text in, text out to thinking, okay, what documents could I provide as context to this conversation or task that I'm working on. And then once you have them there, finally, I want to show you that you can even go into new chats here, just type and then I could link those files from within here. And if we go back to the original chats where I created the personal finance tracker, you could even do that within a conversation, right? So this is what really makes it amazing and why I think everybody should start using this right now. And if you want to learn more about context management, I actually created a separate video this week focusing on Claude cowork projects. Another incredibly useful feature when it comes to managing context, but I created a standalone video on that. So let's move on to looking at what else came out this week. So the next one is a big overhaul to Google's AI Studio. And this one is actually interesting because they had these separate products. And as we talked about, there's this trend with Google where they just create all of these experiments, different applications, and then over time they merge them. They take the ones that work the best and they turn them into singular products. Same thing here. They had Google AI Studio, which was basically their developer-facing Gemini. That's a good way to describe it. But also a lot of non-technical consumers used it because it just had a lot of these presets. You could use it completely for free. And it was just a nice and easy way to, let's say, spin up an app or a website. Now they completely overhauled it and merged it with this other application that you might have heard about. We talked about it on the show before. It's called Antigravity. And that was really their agentic app builder. It was very autonomous, very capable. It went mini viral on X, but I feel like it tapered off over time because just everybody keeps returning to Claude Code or Claude Cowork, OpenClaw, Lovable. I think those are the apps that I would recommend over Antigravity. But now they made this merch, and it's interesting because it's really user-friendly and there's a lot of presets. So I want you to see this. There's some really impressive demos of things that are technically complicated, like a multiplayer game or a multiplayer experience using 3D particles. As you can see, they're really leaning into the multi-user aspect with these demos because these things have been notoriously hard to do with other models. And that's because it's not just code, it's also the server architecture. But the cool thing is this includes databases and authentication allowing you to create shared tools or games for multiple users. And the reason for that is because, as I said, they're merging it with multiple things. And what I did not mention is that they merged a third thing into it. You may or may not be familiar, but Google also had Firebase, which in simple terms is basically the database that would attach to an app that allows you to create accounts and more. Okay, if none of that made sense, fair enough. Let me just show you. So when you arrive in here, it might look similar, but when you tap over to build, look at that, immediately I get a pop-up with, hey, you can really build multiplayer experiences easily now. This is a very unique thing in the But let me tell you that. So if I scroll down, this is what I meant. If nothing that I said made sense to you, here you have a bunch of example applications. And here's the really fun part. When I click one of these, immediately I get put into a version of it where this is live on the internet and there's other players in here. So depending on their activity, for example, in this application, Cosmic Flow, I would see their cursor going around. And if you press the button that creates this, it's just a little interactive game. Now, What if I want to make this app my very own? Well, it couldn't be any simpler here. All you do is go up to Remix, you can give it a name, you say apply, and now it's going to create a copy of this that is going to be yours. This copy you can, well, test right away here by pressing launch, or you can use the chat here on the left side to make changes. There you go, it's a functional app at this point with one player that is not hosted on the web yet. Also, I didn't make any changes. I could change the particle lifetime or add new functions, whatever you want. The point is this: if you have this ready and you want this multiplayer application, well, really publishing it is as simple as pressing publish up here. And I did do one thing in advance here. I set up a billing profile. I just had to add my address and a credit card, and that allows you to see this screen where you can publish the app. You simply say publish app, and there it is. Now I have a live online application that we could play together if we wanted to. This has never been that simple. This is insane. Like, Seriously, multiplayer applications were just not doable at this level of ease. So this is really AI moving forward in a big way. We can collaborate in this. And sure, this is a game, but you could probably easily imagine a productivity-based app where your team collaborates and all of a sudden it doesn't just have a multi-user interface, but it has features like a live canvas where multiple people can use a whiteboard at the same time. You can make things like that really easily with this now. Now you know. Now I do want to add that, you know, hosting and using it is going to cost you a little bit of money. That's where you need to set up the billing account. The next story is by a company that is kind of legendary in the space, but hmm, sentiment of this is very negative across the internet. It's Midjourney V8. Yes, the brand new version of what used to be by far the best image generator. But at this point in time, there's so many others and they're so good at not just generating but also editing. And turns out that editing is a use case that many people actually want. Well, nevertheless, they launched V8 in alpha. You start out by rating, I think, 300 images. And what we did is right away we ran our typical test prompts that you might be familiar with. And the big conclusion? Meh. The prompt adherence here is really weak. I mean, in this image, we literally asked for a Marvel superhero and we got a DC superhero. In others, it's missing elements. What I did notice though is that the composition is a bit off axis. Look, usually this couple in the cinematic still is in the middle. Here, they're off to the side. So, you know, it's, I guess it's a bit more of a cinematic model. But overall, I don't see a way in which I could present this to you as a step forward. Just a variation of what we had. And that's been the story of many image generators, including Midjourney lately. They just kind of reached perfection a while ago. How are you going to be better than perfect? You can just be different. If you find this useful, make sure to subscribe to the channel. It really helps us out. Okay, let's move on. So for the next one, this I found so, so interesting, but maybe that's my bias as an AI creator and trainer. And it's a study that Anthropic conducted by interviewing people through the web interface. And there's many insights here. And I actually do recommend reading this. It's amazing. This This quote up here really resonated with me. I think it's so accurate. AI is sort of like money. It just makes you more of what you already are. Just amplifies, right? So good. Such an accurate take in my experience. Anyway, as I said, there's a lot in here. But what I liked is seeing what people hope for with AI. With personal transformation, life management being actually ranked 2 and 3. So it's not all just professional. This actually maps onto my experience too. Time freedom is obviously a big one. And then by the way, you could open these up and see more quotes of people about professional excellence. They show where AI actually delivered on their vision. So it's a big productivity tool. And people use it for cognitive partnerships, like a thought partner, clone type of situation. And then the main thing that people worry about, and this all is very true from my experience too, is the unreliability. It's just, you do a task, but then you do it 5 more times and one of those times it's just not what you would expect. And then you can't rely on the tool. So you always need a human in the loop. Yes, unreliability is the biggest one. Super interesting article, check out more. Okay, so let's talk about this week's quick hits. One of them is Gamma Imagine. So this is also a theme that I noticed in the space lately. If you're not familiar, Gamma is what I would consider the best presentation creator in the space. And then there's also another app that you might know, Lovable, which I would consider the best website creator, at least the most user-friendly one that is the most aesthetic. Both of these apps lately started launching other features. So here Gamma is trying to do Gamma Imagine, where it wants to turn into a graphic design tool. And on the same note, Lovable last week announced that they're now moving to be a general purpose assistant to create documents for you too. I mean, I understand that they have to keep growing and they probably plateaued. But yeah, you're trying to compete with the big boys and that's going to be hard. Anyway, if you want to try it for social media content creation, Gamma Imagine is available to you now. Then we have OpenAI releasing GPT-5.4 Mini and Nano. These models won't be relevant to most consumers. But for developers, if they're building agents, some of the sub-agents could be using these models. They're quicker, they're cheaper, as per usual with these smaller models. And the benchmarks are impressive for the size of the model. But again, most consumers won't be using this. Then Figma also introduced AI agents to compete with last week's Stitch release that we're making a dedicated video on. Look out for that on the channel. Claude Code shipped even more features with their new auto mode, which is actually really cool. Because if you're not aware, Claude Code used to have this one command that I even showed off on launch right away. And which basically said is like, don't ask me for permission on anything. Just do it. And it did, it went wild. In some cases, it deleted full folders for people. In other cases, it just got the thing done. And auto mode is basically a lighter weight version of that, where a classifier looks at what Claude Code is about to do. And if it ranks it as dangerous, then it asks for permissions. But in most cases, it just gets it done. So yeah, Claude Code also evolving into a more autonomous system. And then also, I want to point out that people have been hyping up the remote cloud computer use. But really, that was them just remarketing what at first I think was probably a failure in labeling. A feature that we covered in last week's episode, Claude's dispatch feature where you could remote control your computer from your phone. This is the same thing with just better branding and people are going crazy now. It's like, what is coming out? Last week dispatch came out and people were like, oh, all right, that's good. So it's funny, just the power of words at display. Anyway, we've had that feature since last week. Another quick story is that Gemini is rolling out personal intelligence to free accounts now. Something that was announced a while ago, now it's finally coming. Their way of gathering more context automatically and bringing it to free accounts. So they're really moving forward on a more personalized experience, as are all their competitors. But Google has more of that personal data than most others do. And then finally, this might go over the head for many people, but I want you to know that this exists. It's an MCP server. So a way to connect external applications into things like Cloud Cowork relatively easily. And this one is for a service that I've just been using for a long time. It's called Fall AI. And it's basically a collection of all the different image generation and video generation models in one place. And the cool thing about Fall is you connect to it once, and then you can use all the image models, all the video models. And now with this MCP server that can link into ChatGPT. So if you want to half automate the creation of images and videos with AI, this MCP server makes it easier than ever. I'm saying it's super easy, just easier than before. And that's all. Did you find something that was helpful? We'd love to hear in the comments below which one of these was your personal favorite. And with that being said, my name is Igor Pogany, and I hope you have a wonderful week.