Anthropic Finally Solved AI Memory & More AI Use Cases

Welcome to another week in AI. Things have been busy. We'll be looking at big releases from big companies this week, including claw shipping, their version of memories. Spoiler alert, it's the best implementation I've seen so far. Google AI Studio releasing a plethora of models, all-in-one interface, and the revamped Google AI studio. Open shipping a browser, and so much more in this week's episode of AI News You can use the show that rounds up all of this week's AI releases. Filters by the ones that are available, and then we discuss and test the ones that are relevant.

Let's go. So let's talk about memories in general. You probably know these from within chat GPT. In chat they're under personalization and down here memories. They're automatic way of gathering information about you and then your new chats include that information. Claude now ships their version of this but it is way different from what we've seen within chat GPT and I would even argue different from any other AI out there. If I go to settings then capabilities and scroll down somewhere in the middle you should find this new memories feature. As you might already notice from chatb there's two options. one of them where you allow it to look into past chats and the second one where it proactively creates new memories from past chats. And this is kind of the big difference. The first time I saw this, all of these memories were pre-generated in here from the past chats. And then you could edit them by talking to Claude and then saying something like remove this and then it will remove it. But let's get to the differences here and why I think this is superior to the chatbt implementation as of now. First of all, it separates memories that are just about me from ones that are about recent things I might have been working on. You can see that here in this top of mind category, it talks about some recent chats that I had. If you saw the chatbt Atlas video, you would have seen that I was generating images of Conor McGregor with Claude as a demo for Atlas, which we'll talk about in a second. And it shows that here. It says that, hey, Igor is working on content creation projects, including developing social media captions for MMA fighter content. This is not what I usually do, but I absolutely love that it's categorized under top of mind here. Whereas in ChachiPT, this would just be another memory that in terms of hierarchy would be evenly as important as this core work context for me or personal context. Especially the personal context doesn't change that quickly. Whereas my latest chats changes on a daily basis. And this is why if you've been following the channel, you will know that I'm not the biggest fan of memories in chat because me trying stuff will often pollute all my other chats. I'm not a content creator for MMA fighter content, but chat might assume this very quickly if I have memories on, but here it's categorized under top of mind and it shows. I tried this on a bunch of chats. If you ask it about yourself or your work, it's never going to pull up these top of mind things. And if it does it, it will know that this is just something you've been trying recently and not a core part of your identity or your work. If I ask it what it knows about me, it recites the memories. And you can see it does point out that I've been working on a content creation project relating to this, but it knows that I've been only working on this rather than it being a part of my core identity if you ask cla. Okay, so you might consider that a nuance, but here's the fact that makes this so amazing in my opinion. These memories get rebuilt on a daily basis. You can see it up here. The summary is regenerated each night and does not include projects which have their own specific memory. So if I work on different things, this will update every day and claude will simply have context on what I've been kind of trying lately which changes every day. Whereas in chat static once you work on something, it goes into the memories. And yes, they added a feature where things might get deleted last week. But I found it to be a bit random. And this just works so much better for me. And yes, this also works for projects. So if I go into my AIA education project which I set up to manage a lot of the B2B initiatives that we're doing now you can see it's generated memories in here too that it also updates every day if I do anything within this project. Interesting thing I want to highlight that also here these categories of context are dynamic. So depending on the project depending on what you're working on these change and morph depending on what kind of memories would be relevant to the given project. And most of all, and this is just a personal feeling, when you work with Claude, it just seems to work and it knows the right type of information. It's not too much, but it's also not too little. And as I said, the stuff that might be relevant, Claude knows that it might be relevant cuz it's just a top of mind things that you've been doing and not a core part of your journey. So yeah, I really wanted to get my thoughts on this out because I think this is an essential part of your experience in tools like this. And context management is everything when it comes to getting good results here. And honestly, this is the first time I see memories where I'm excited about using them rather than slightly anxious on what will chat save and how will that distract my next chat experience. That's why by default, I kept memories off within chat GPT and I just managed my contacts manually. Long story short, this feels different. Cloud memories are available as of now for all Max users. That's their $200 plan and it's coming to the pro plan which cost $20 a month soon. Free users don't get this. unfortunate, but that's how it usually goes with these features that actually make a difference. Honestly, just this is probably not worth the money. You can just craft the context yourself. If you're enjoying this, make sure to leave a like. It really helps the channel. But now, let's move on. The

next release is also from Claude. Super interesting one. If you're watching this channel, you probably already know about Claude codes. what many people consider to be the best AI coding assistant and many people also use it for marketing tasks product management chat GPT alternative but it has been notoriously challenging to use especially for nontechnical people because you needed your terminal to run it which can be intimidating that's very understandable well now they released a web version all you need to do is connect this to your GitHub and there you go you don't need a terminal you can just do things in here the one requirement here is you do need to know the very basics of GitHub because it works within existing repositories. So for example, I could take the no machine which was a project on a previous video and for that you need to claw GitHub application installed in your repository which is as simple as this. Okay, so that was the setup and that's it. Now I can work with that repository build up my project right here in the web browser and yes this also works on a mobile device which for the first time makes it really simple to work with clot code on the go. Furthermore, you could go into other repositories and start chats in parallel and do multiple things at once really easily without your desktop looking like this. As per usual, you should be careful with multiple instances of cloth code on one project. And as already mentioned, I think the best part here is that it's really easy to use on the go now. So now it can keep vibe coding useless applications wherever you go. Okay, that's a joke. There's definitely a lot that you can do that makes a ton of sense, but every good joke has a bit

of truth to it. Okay, so one of the biggest gaps with AI right now is the following. It's great at thinking, but when it comes to actually doing things like sending email or updating an ocean page, most people still end up clicking through 10 different tabs or apps. And that's why I'm excited to show you Rube by Composio, the sponsor of today's video. It's basically a super assistant that lives within any AI chat application that you already use. In other words, it's like glue between the apps and something like Chat GBT, Claude, Cursor, or VS Code, you name it. and it lets you control over 500 apps with just natural language. So instead of building an automation and struggling with connecting various APIs, I can just say find emails from the last 24 hours, summarize the important ones and mark promos as spam. And Rube just does it. It even remembers you across various applications. You just authorize once and then run Gmail actions straight from cursor, claude, manos, whatever. For creators and builders, this can be huge. I can pull comments and likes from various platforms, drop them into a Google sheet, and flag the top posts. All with one prompt, no context switching required. And if you want to, you can go way further. For example, you could publish a Figma design, auto resize to Instagram, Tik Tok, X, and YouTube, or scrape Reddit trends and create posts based on the findings. As you know, I'm a big fan of open standards and MCPS. And if you want a quick way to start with Rube, you can add the streamable NCP endpoint rupt. app/mcp or just hit a rub. app app and plug it into something like NAN or any other automation app that you might already be using. It's SOC2 compliant with audit logs, so it's enterprise ready, too. If this sounds interesting to you, try RP for free at getroup. link/Iadvantage. Thanks again to Composio for sponsoring this video. And now, let's look at some more AI news that you can use. The next

big release this week has been chat GBT atlas. And if you're not familiar, that's our brand new browser joining other AI browsers in this category like Perplexity's Comet. If you're interested in the details and how it performs in practice, I created a separate video that you could check out on the channel. But I did want to follow up after a few days of using it. And that is the fact that for me personally, this is not a sticky product. It's not something I really find myself wanting to use. And I've been even looking for specific ways to utilize this. And usually the most interesting ones are the AI browser using other AI tools and then maybe moving it over to something like a Google sheet or a Word document. Again, I show that workflow in the separate video, but here this example actually combines another story we'll be looking at today, which is Google's new AI studio. And it just prompted Atlas to open the new Google AI studio and then build the app within it automatically. I'm sure you can get creative and find ways to get some value from this, but it's probably not the game changer that many people made it out to be when it came out. Oh, and there's one more prompt that I wanted to share with you here. Daniel from the team actually sent this to me. He said, "This can be quite useful. Read this page. consider how it relates with my areas of interest and then give me a list of key points and insights. This way you can kind of match what context chat has on you and the content that you're looking at. This way you would kind of adapt it to you. Yeah. Here it talks about agentic system requiring memory hygiene similar to what clot does with the memory release, right? But other than that, this is kind of good, but nothing that you couldn't have done with chat GPT. It's just a bit simpler in the browser as this site is right in here. And I personally do find these results here interesting. We'll put the prompt in the description below. You can try it for yourself as this works on the free plans too. For the

next story, this could have been a quick hit, but I actually want to make it a proper thing because this has been my number one feature request for a while in ChachiP and it's the ability to share projects. Up until now, this was only accessible to the enterprise plans as far as I'm aware and now all the plans including the free users can share projects. Meaning, if you built out your context in a project, you can go ahead and finally share it with other people. Just make sure to change the accessibility of the project to anyone with a link. Or you can put an email in here and they'll get an invite. As you might have heard me say many times before on this channel, I think projects are the most powerful way to manage context. But with the sharing ability, this feature is a powerhouse and now you can finally collaborate with people. So one person gets to build it out and everybody gets to use it as an assistant on any sort of project. It's a minor feature within chat with a big impact at least for me and anybody else who has

been using this on a daily. Another thing that is new this week is Google overhauling its AI studio. If you're not familiar, this is their developer interface for their various models. But really, a lot of users and especially power users have felt at home in here for quite a while. Reason for that is all the new models come right into here before they even hit the Gemini web interface. And usually when new stuff comes out, it's free to use in here. And now they overhauled the entire interface, so it's even easier to use various applications all in one place. If I go to the chat interface here, you can see that I can switch between all their models. Nano Banana for image generation, Gemini 2. 5 Pro for the smartest model. There's even a video model. Although, I do have to note that last week's release of VO 2. 1 is not in here yet. I'm sure that'll follow shortly. But yeah, you can just do all of these things in one place. So, if I want videos and images of cats with hats, I can be so efficient at creating them now. This is incredible. Seriously though, the chat is kind of nice to have all the models in there. And then here in this build section, you can build apps. So, I could build an app that adds hats to various pictures of animals. You could build this and it's going to code up the whole thing. You could save it right to GitHub or deploy it. There you go. Works as expected. And of course, this remains the place in which you retrieve your API key to build applications on top of Google's AI

offerings. Okay, this week we have a bunch of quick hit stories that are worth mentioning, but maybe we don't need to spend multiple minutes on them. Starting with the general availability of the Claude desktop app. I feel like this has been generally available already, but they've updated it with a few features like voice inputs, a shortcut where you can invoke it from wherever you are. Just double tap the option button on a Mac. You can easily bring in context by screenshotting different windows. One note about the voice inputs. Honestly, it's quite bad. It just gets punctuation, capitalization, and some words wrong regularly. Nowhere close to what you might be used to from ChachiP 11 Labs.

Next up, there have been some changes to the guardrails on Sora 2. This is kind of an ongoing story that will keep developing as this has made a big impact in the mainstream. So, one thing that happened is that Martin Luther King's estate complained to OpenAI for disrespectful depictions and OpenAI banned the usage of Martin Luther King. If you haven't seen them, there were a bunch of videos where he was in a grocery store. — My sister, I have a dream that one day this slushie will be free for all who thirst. — Sir, hey, SIR. THIS ISN'T ABOUT DREAMS. IT'S ABOUT PAYING FOR YOUR SLUSHIE. — I MEAN, I thought it was hilarious, but I entirely understand the point here. So now OpenAI is apologizing for unintentional generations and there have been a bunch more celebrity complaints like this. Now OpenAI is slowly reeling it back in. So yeah, as per usual some of these AI releases are most exciting and free on the day of release. Now the restrictions are coming in. Then we also

have runway releasing various applications. I mean things like changing image styles is not novel. Others are more interesting like removing objects in videos. It's just making things that were possible before more userfriendly. And I think for filmmakers, these runway apps are something you should definitely explore. On a similar note, they also released this workflows view. We've seen this many time before, but runway is just trying to be this platform that has all of it. So, this is new also along with the new model fine-tuning for which you do need to sign up, but I reckon this is going to become a bigger thing over time where filmmakers can fine-tune their own models and specific aesthetics and

really make it their own. Then we also have Perplexity moving in every niche imaginable. I mean, at this point, they're just trying to be the Google of AI. They have a language learning feature now, so you could be learning languages with Proplexity. Next up, I

wanted to briefly feature something that you might not be able to use, but I thought was really interesting. Channel 4 in the UK used the AI presenter for the first time to kind of make the point that people wouldn't even notice these days. Just thought this was interesting and I wanted to share it with you. At the end of the segment, she kind of revealed that she's AI. Viewers were

left in shock. And then last, but certainly not least, I wanted to show you this Google Skills page. And this is brand new, but basically it's a platform to acquire AI skills. And a lot of this is free. So, you can just go into the catalog, look at the different topics here. Some of them are going to be very technical and very concrete. But if you go to these collections, for example, and go to AI boost bites, they have these 10-minute lessons on various topics, including little workflows or hacks. As you know, this channel is all about uplifting people by sharing this information. And I think Google Skills do this really well. So, feel free to check it out. And that's really everything we have for this week. A lot of practical things that I will be implementing in my workflows and testing further. I hope you found something useful or inspirative. And with that being said, my name is Igor and I hope you have a wonderful