Ep 756: Claude's Desktop Updates: How-To Guide that Turns Claude into A Proactive Personal Agent
16 Apr 2026 · 0 min
In this episode of the Everyday AI Show, the host walks listeners through Anthropic’s rapid rollout of Claude Desktop updates in early 2026, contrasting the desktop app with the more limited Claude web experience. The discussion outlines the three core desktop modes—Claude Chat, Claude Cowork, and Claude Code—explaining how they differ in sync behavior, capabilities, and target users. A large portion of the episode is devoted to new desktop features released over the past eight weeks, including Routines (local and remote), Scheduled Tasks, Dispatch (iOS‑to‑desktop computer‑use), integrated terminal, in‑app file editing, diff viewer, side‑chats, shortcut keys, and UI redesign. The host demonstrates practical, non‑technical use cases such as daily email triage, automated podcast‑stats analysis, and AI‑generated design tools, showing how Routines can pull data from connectors (Gmail, Google Calendar, Slack, Canva, Figma, etc.) and run on a schedule or via webhooks. The episode also highlights Anthropic’s strategic shift toward a self‑hosted, OpenClaw‑like ecosystem, the upcoming Opus 4.7 model, and a new AI‑powered website/presentation design tool that threatens competitors like Adobe, Wix, and Figma. Downsides are candidly discussed: silos between Chat, Cowork, and Code; buggy implementations of Dispatch and scheduled tasks; throttled model performance leading to service outages; and UI quirks in the new desktop layout. The host wraps up with calls to action—sign up for the Everyday AI newsletter, repost the show for early access to an upcoming “Everyday Cowork” cohort, and stay tuned for future AI updates.
This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life.
Anthropic has already shipped more updates in the first 4 months of 2026 than than they did in all of 2025. True story, I counted. Actually, I was about 5 hours into planning this very show yesterday on Claude's desktop updates. And what did Anthropic do? Well, of course they came out with a completely new Claude desktop app. That's just how it goes. But aside from being indicative of how fast AI moves, these fresh features from Anthropic that we'll go over today signal a much broader shift from 2026's hottest AI company. They're moving everything to your local computer and simple tasks are getting an agentic glow up. And today we're going to carefully dissect the differences and mainly the advantages of using Claude on the desktop versus using it how most people are probably using it, which is just on the web. And now update after update, if you haven't noticed, Anthropic has kind of put this emphasis on becoming like a version of OpenClaw, the very open source software that it kind of severed official ties with. So we're going to put AI to work this Wednesday and bring you up to speed on everything that's possible in Claude Desktop. All right, so If you're new here, this is Everyday AI, but let me first give you the big picture. Most people are using Claude on the web, and if so, you are missing out on the majority of what Claude can actually do because most of its more impressive capabilities are being rolled out now almost exclusively on the desktop. And in the past 8 weeks, there's been more than a dozen meaningful updates to Claude, most of which are coming to the desktop. Desktop. And since reportedly using Mythos internally since February, Anthropic has been on a straight-up tear in terms of new features and capabilities. And with features like scheduled jobs on desktop and being able to launch literally and control your desktop from your mobile phone via the Claude apps, Anthropic is turning their desktop apps into their own version of OpenClaw. but not without its own fatal flaws. So on today's show, stick with me for the next 25-ish minutes and you will learn how Anthropic is using its desktop app to turn Claude into a proactive, always-on agent. You're going to see a handful of updates over the past couple of weeks that have completely changed how Claude Desktop works. You're going to know why routines are actually going to turn non-technical users into Claude code fanatics. Even if you've never used Claude Code. And I'm going to talk about the only downsides that no one's really shining a light on when using the desktop version that you can't overlook. All right, let's get into it. Welcome to Everyday AI. If you're new here, yeah, we do this every day. This is your daily livestream, podcast, and free daily newsletter helping everyday business leaders like you and me make sense of these nonstop, quite literally nonstop AI updates. I tell you what's important, how to use it to grow your company and your career. So if that's you, sweet, we're on this journey together. It's better to do it together. So make sure you go to youreverydayai.com, sign up for the free daily newsletter like thousands of other smart people are doing. And we're gonna be recapping the highlights from today's show as well as all of the other AI news that you need to know to stay ahead. All right, So Claude Desktop, are you using it? Chances are you're not, right? I talk to a lot of people in AI. That's really all I do, both on this show and not on the show. And most people I talk to are using Claude on the web, right? Claude.ai. And if that's you, don't feel bad. But if that's you, I think today's show is one of those that you might wanna watch twice. You might wanna take some notes and you're definitely gonna wanna read today's newsletter because for a couple of reasons. One, I mean, we have to look at and realize what's currently possible and what's currently possible, um, on the Claude desktop app is almost night and day. In terms of what's available on the Claude web app, right? Going to Claude.ai. It's kind of, I'd say, you know, old school chatbot on the web and on the desktop. We are now entering the full autonomous agent, right? That is going to do work for you, right? You can set the schedule, but it's going to run on your command. And the great thing is you don't have to be technical to understand how to do this. And hey, FYI, I'm thinking about doing an everyday co-work. I don't know if this is going to be a class like our Prime Prompt Polish or this— if this is going to be more of a cohort. So going along live, either way, just repost this on LinkedIn and I'm going to send you access on when we're going to be doing this. It's going to be later in quarter 2 because I do know that between OpenAI and Google, we're going be getting some really big updates in the next, uh, 2-ish weeks, uh, so we're gonna be pretty busy covering those. But I'm gonna guess that probably in, I don't know, probably May or June, we're gonna launch this Everyday Cowork class. All right, so if you want early access, it's going to be very limited, um, just make sure to repost the show and I'm going to make sure that you're on the front of the waitlist, FYI. All right, So today, today's show is going to be slightly less hands-on. And one of the reasons, like I literally said, I was 5 hours into planning this show on Claude's desktop and they completely changed how desktop works. And obviously I've been playing with it for a couple of hours and testing some things. There's some things that are buggy that weren't buggy before. So we are going to do a little bit live demo at the end. But a lot of this is not going to be as hands-on as we would normally do on our AI at Work Wednesday show. All right. So Yeah, we do the AI News on Monday. We do AI Working Wednesdays, which is normally a very hands-on demo. Today, a little less hands-on. And then on Friday, we do Friday Features, which is a collection and a roundup of all the biggest AI updates of the week. All right. And then Tuesdays, Thursdays, we kind of rotate our shows. So, but we're still going to be very much demonstrating the different capabilities and what's new in Claude Desktop. So first of all, why would you want to use Claude in the desktop versus on the web? And we're going to explain all of these in more depth later, but I want to give you what I think are probably the 4 biggest reasons and the main benefits. So number 1, you can use Claude Chat, Claude Cowork, and Claude Code all at the same time simply by toggling the tabs. All right. And if you don't know what Claude Code or Claude Cowork are, Don't worry, I'm going to be explaining those to you here in a bit. All right, so that's number 1. Number 2, uh, Cowork and Cloud Code allow for scheduling tasks and computer control, and we're going to be going over both of those, uh, later. But that's really the— I'd say one of the biggest separators, um, on the capabilities of using this on the desktop. All right, number 3, well, by essentially leaving your computer on, Cloud Desktop can access your personal files and complete tasks and save them on your computer autonomously. All right. A lot of times on the show, I talk about the human duct tape that still exists a lot with even people who are very fluent in AI. A lot of the work that we still have to do requires downloading for, you know, downloading, reformatting, copying and pasting, uploading. Right. So, yes, there's some permissions and privacy and safety elements that you, uh, have to pay attention to. Um, but that's the big thing. By— because this is a downloadable program that runs on your desktop, you can give it access to certain folders. So different projects can have different access levels, right? So if you are using this for work purposes, make sure you go through the proper channels to get this approved. But that's number 3. If you leave your computer on and combining that with number 2, which is scheduling and computer use, You can literally do anything that a human would do around the clock. And then last but not least, number 4, desktop can control a Chrome browser via the Chrome extension. And yes, you can technically just do that manually, but you can also technically schedule this now with Clone Desktop. So those are the main reasons. And we have to understand what's next because there was another report that just came out yesterday. Aside from everything that is already updated and live, well, we also just got word from reporting from the information, and I'm going to read this here quickly. Uh, so this is from their report that just came out yesterday. They said Anthropic is preparing its next flagship model, Claude Opus 4.7, along with a new AI-powered tool for designing websites and presentations, according to a person with knowledge of the products. Uh, those new products could be released as soon as this week, the person said. News of the upcoming AI design tool sent the share prices of Adobe, Wix, and Figma down more than 2%. In the hours following this report. Little typo there, not house, in the hours. All right. And then it says that tool would also pose a threat to startups like presentation maker Gamma and AI design tool Google Stitch. It aims to help both technical and non-technical users create presentations, websites, landing pages, and products using prompts in natural language. All right. So there's more on that, but that's really what I wanted to focus on. So aside from what's already here in live, and we're going to be going over the updates, Obviously Anthropic has a lot of big plans for the future desktop versions, right? Aside from we're getting word that Opus 4.7 may be close and around the corner, it essentially looks like they're going to have some sort of Figma slash Gamma slash Lovable type, type of setup that you can literally create, whether it's full stack applications, right? So a lot of the things that many people are vibe coding in Claude Code, for an example, you might still have to take that project and finish it off, off elsewhere to build it into a truly full-stack application. Well, that might be changing. The same thing with design. You might be able to get a, a pretty good design, but if you want something a little more polished, you might want to use something like Gamma. So, uh, not just keep in mind with what's available today, but also, uh, you know, reporting from the information is usually spot on. And we've seen a lot of this already in testing. The desktop version of Claude is going to become a powerhouse. So first, before we get into this, I want to also, especially for our non-technical users who aren't using this a lot, there's a lot of different places that you can use Claude. All right. And this is kind of the first time we've been covering desktop software. So I do want to differentiate and explain things a little bit. So the place where most people probably use it is on the web, going to Claude.ai. Then there's also on the web, a backend platform. So that's platform.claude.com. So as an example, the new managed agents feature, which I asked everyone in Monday's newsletter, what did you guys want to hear? I was actually surprised that the managed agents didn't win the poll for today's show. It was overwhelmingly cloud desktop updates. All right. So if you hate today's topic, start reading our newsletter and voting. We usually put that poll in on Mondays, what you want to hear on Wednesdays. All right. So you have the web, Claude.ai, you have the platform.claude. Dot com. That's just kind of like a playground sandbox, etc. Then you have the desktop app. That's what we're going to be going over today. It's available for Windows and Mac. The features are a little bit different. What I'm going to be talking about is Mac. Then you have the command line interface tool, the CLI, so you can use Cloud Code in the desktop app or in the command line interface. I do know a lot of, you know, serious people in the software side Up until yesterday, we're still very, you know, hardcore committed to the command line interface way to use Claude code. But I think some of these newer desktop updates that might be shifting it. Then you can obviously also use Claude via the API and build off it on the backend. And then last but not least, you have the third-party integrations, as an example, using Claude in Excel. So there's a lot of different places you can use Claude, but we're mainly going to be talking about today using it on the desktop app. So before we get too granular, let's kind of explain kind of what's, what's what here. All right. So Claude Code, if you don't know, here's the basics. It's an AI coding agent that builds software from natural language. All right. But the difference is it has access to your computer. All right. So for Claude Code, even if you don't know how to code, you can pick up a coding project that's already in progress. You can start one from scratch. It's going to write write those files for you. It has access to the different folders, commands, etc. And unlike WebChat, it does and can control your local project files and can run code on your computer. It can run your terminal. It can essentially run anything that you give it access to. And then it is available on Pro plans and up. So then Cloud Code, diving a little bit deeper. So it also gives you a visual code preview, which I'm going to show you what that looks like because that is also new. So you can test a live app window without leaving the tool, which is really good. You don't have to, you know, render it in Xcode or anything like that. The terminal version unlocks parallel sessions so you can automate scheduling and background tasks for bigger projects. There's also computer use, which I wanted to really demo that, but that's so hard to demo computer use and do an unedited, unscripted livestream at the same time. It just wasn't going to work. I did like a 30-minute test and I'm sitting there, I'm like, this isn't going to work anyways. Computer use lets Claude open apps on your Mac or Windows machine, click through interfaces, and test what is being built. So in Cloud Code, you can obviously, uh, there's a window within Cloud Code that you can preview whatever you're building, uh, but then also it can literally control your computer, right? It can launch, uh, launch other apps, it can use other apps. It usually does so via screenshots, uh, it is getting faster, uh, but it's still very slow. And then you have Dispatch. All right. Dispatch is technically a connection between the Claude iOS app, so the app on the iPhone and Claude Mac app, so you can run computer use on your phone. All right. So I want to do one Claude code use case. All right. And showing screenshot on my computer here, it's not a very interesting screenshot because I had to blur out a lot of it. Right. But a daily Email triage, right? AI moves too fast to follow, but you're expected to keep up. Otherwise, your career or company might lag behind while AI-native competitors leap ahead. But you don't have 10 hours a day to understand it all. That's what I do for you. But after 700+ episodes of Everyday AI, the most common questions I get is, Where do I start? That's why we created the Start Here series, an ongoing podcast series of more than a dozen episodes you can listen to in order. It covers the AI basics for beginners and sharpens the skills of AI champions pushing their companies forward. In the ongoing series, we explain complex trends in simple language that you can turn into action. There's 3 ways to jump in. Number 1, go scroll back to the first one in episode 691. Number 2, tap the link in your show notes at any time for the Start Here series, or you can just go to starthereseries.com, which also gives you free access to our inner circle community where you can connect with other business leaders doing the same. The Start Here series will slow down the pace of AI so you can get ahead. You will 100% find value in this, especially if you're like me and you struggle keeping up with email. Yes, the chat version has a connector for Gmail, so you can do this manually. But the great thing with the new updated scheduled kind of sections of Claude Code, and there is a— they kind of updated it. It's called Routines now, another fresh update, right? Now you can schedule this. So anything that you would normally do, and I know Claude Code might sound scary because you're like, I'm not a coder, well, don't worry. There's a literal— you can build these with natural language. But in this example here, I can have this go through my email every single day at, let's just say, 6:30 AM and triage. And here's, here's what it did. It listed out 4 different emails that needed action that day. So kind of, you know, said like red, yellow, green. So red ones were ones that I needed to apply to immediately. They were pretty urgent. Yellow ones were ones trying to keep on my radar, and green was just like FYI. So Obviously, I'm still going to go in and, you know, read those emails, but at least I know right away what needs my attention first without having to waste 20, 30 minutes going through, you know, dozens of messages. All right. So now let's talk about Claude Cowork. So essentially, Claude Code came out at the end of 2025. And what Anthropic noticed is a ton of people were using it to do non-technical work. They weren't using it to code things. They were using it to organize files on their desktop. So Anthropic, big credit to them because I think this actually kind of shaped, um, agentic AI in 2026 completely, is they created essentially a non-technical version of Claude Code called Claude Cowork. So it gives non-technical people Claude Code's power without ever having to open a terminal or technically even staring at code. So you can essentially point it to a folder, You can set the permissions and then describe in plain language what you want it to do. So as an example, you could say, go to this website, copy and paste these stats, put it into this spreadsheet, and save the spreadsheet to this folder, right? Which might be a task that you do every single day. So you can assign it to Claude via Cowork. So it creates real deliverables like Word docs, spreadsheets, presentations, and then saves them. And that's also available on the paid plan. So the desktop app on Mac and Windows. Diving a little deeper into Cowork. All right. So it connects almost 40 tools as well. So there's non-technical connectors the same way that you can connect them in Claude on the web. They work in Claude Cowork as well. So everything from SharePoint to Google Drive, Outlook, Slack, etc. So Claude can dynamically pull all that data across your workflow. Then you can also schedule recurring tasks in Cowork. I will say this. Uh, I'm loving the brand new routines, uh, which is a little different than the schedule, um, because at least for me, the scheduled tasks in Cowork are very clunky, and they have been since they started. And in my brief testing so far, the brand new routines, at least for me in my use cases, have been much more stable even though they haven't even been out for 24 hours. All right. Also, Cowork does have computer use as well. So computer use lets Claude open apps and navigate your browser when there's no direct connection to a tool available. So obviously, if Cowork has a connector, you should use the connector because if you're using computer use, yes, it's a great way to technically navigate the web or, you know, open other programs on your computer. But it is a little clunkier and slow because it's literally using computer vision and screenshots and navigating, but it can literally move your mouse and click on things. Uh, right. So that's really cool. And then Dispatch, which I referenced earlier, it technically lives inside of Claude Cowork. All right. So there's a dedicated Dispatch section inside of Cowork, and that's kind of how you can control your computer via the Claude iOS app and the Mac Claude app. All right, so here's just a simple use case that I took a screenshot here. I had this weird problem on my Mac and I said, use Computer Control and find this macOS security thing that's open on my computer. It's kind of between the monitors. I have two, I can't click it and I definitely want it closed, right? So this was an actual thing that was going on a couple of weeks ago, I think. There is some system error on my computer, but I couldn't just shut it down because I, well, I had Codex and Claude code running and I didn't wanna lose the progress on that. And I had like some meetings coming up and I'm like, okay, I just gotta, you know, let's use Cowork here. I did that. I literally think I went to the bathroom, grabbed a coffee, came back and it was fixed, right? I had no clue, right? Traditionally it's like, oh, you go to Google or Stack Overflow and you're like, what the heck is this? I tried everything I needed. I knew how to do, you know, I tried to kill the processes, nothing worked. So that's a good example. Give Claude your computer, walk away, come back, and, you know, the project is usually done. One thing to keep in mind, right, about any computer use. So whether you're using this via Cowork or using it via Claude Code, computer use itself, it hijacks your screen, right? So this isn't something where even if you have two monitors like I do, we're like, okay, I'm going to let Claude you know, quad computer use, work on this monitor and I'm gonna work on this monitor. You can kind of do that, but not really because it needs your mouse, right? So when it knows it needs to click on something, it's gotta go do that. Or, you know, it might open a new tab or a new Finder window and oh, that popped up on the other monitor. So now it's taking over your monitor that you thought you were working on. Using the computer use, it is good for situations like I said, if you need to do a quick 5 to 10 minutes step away from your computer or the majority of time if you're not in front of your computer, right? My main machine, I'm using it 10 to 12 hours a day, my Mac Studio, right? But I have it doing things obviously overnight when I sleep. So now let's talk about some desktop customization, right? I do want to go over all the features and this really to be a desktop 101. All right. So there's different things that you can do. To customize Claude, and not all of these are available on the web, FYI. So there's kind of 3 tiers. All right. There's apps, there's skills, and there's plugins. Okay. And with these, you can kind of do them out of the box or you can customize them. So the first are apps, also known as connectors. Anthropic uses the name interchangeably, which is kind of confusing because OpenAI, as an example, used to call them connectors. But now they don't call them connectors anymore. They call them app. Anthropic just flip-flops and they don't seem to actually have a strategy for what these are called because when you customize Claude, you say connect your apps and then it brings up a screen for connectors anyways. Nuance, right? But, you know, as an example, you can connect your Gmail. That's a connector or an app that I use all the time. And then it's dynamically, you know, connecting to your Gmail, you have to give it access to, right? But it's not screenshotting or anything like that. It's indexing. But then it can also, you know, you can do deep research across just your Gmail. So that's an example of what an app or a connector is. And like I said, I think there's almost 40. Then you have skills. I did a whole show on skills before, but if you missed that one, skills are essentially markdown instructions that teach Claude how to do a specific task by giving it a structured workflow, formatting rules, domain knowledge that it follows, right? So it essentially has these different files that are in a folder, and it's kind of like a complex task, right? So if you ever spent a ton of time building a custom GPT, but you really customized it, maybe with API calls, you know, uploading multiple files, you know, like very heavy-duty GPT, that's kind of like what a skill is. The good thing is, is while skills are an open-source standard, so if you do use them in Claude as an example, you can take them to anywhere else that supports skills. Google supports skills. OpenAI supports it on some plans, right? But that's essentially what skills are. And then last but not least is plugins. So the cool thing is Anthropic actually has some great plugins that they made. They have ones that are done by Anthropic and their partner organizations that are really good out of the box. But then you can customize those skills to make them more for you, or you can create your own skill. But essentially, the easiest way to think of it is plugins are a combination of of connectors and skills. That's all they are. Or you can stack, you know, as an example, multiple skills or a handful of connectors and skills. That's all it is. So the easiest way to think of it is, you know, as install packages for developers. They can contain skills, connectors, and other extensions all wrapped together. So extremely powerful, right? So actually, when plugins came out and were announced, a lot of financial institutions and software companies lost a lot of money on the stock market because, you know, the, like, everyone's looking at this and they're like, wait, right? Like, a, like, a plugin essentially can do the job of a junior researcher, of a junior associate, if you give it the right access and if you know what you're doing, right? Because it's essentially repackaging a set of skills and apps. All right, so now we know the modes, we can explain where they can be used. So here's where it starts to get a little tricky, but stick with me here. All right, so there's technically 3 different desktop products: Chat, Cowork, and Code. So Chat and Code are available on the web and desktop, whereas Cowork is desktop only. But even more confusing, Chat syncs. So if you open a new chat in the desktop, that chat will be on the web. If you start a chat on the web on Claude, it'll be on your desktop. Okay, but Cloud Code does not work that way. Cloud Code is different. So if you start a project Cloud Code on the desktop, it is not going to be in the web version. It's kind of like how Codex is with OpenAI as well. All right, so Cloud Code is not sync. And then there's new features that are web only, like the new routines, although they still kind of work even though Anthropic said they don't work on the desktop, but I'm going to show you that they do work anyways. But Claude is also integrated, like I said earlier, Claude is also integrated into other desktop programs. So don't get confused when I'm talking about Claude Desktop, right? I'm not saying the, you know, Microsoft Word or Microsoft Excel that has Claude in it. We're talking the actual Claude Desktop app. So some of the main benefits that we already talked about, but I think the biggest one is Claude is, is co-work dispatch, right? Because one thing I always think about keeping up with AI, right? What are those features that are true differentiators? Right, that you can't get anywhere else. Obviously, I think with some of the announcements that were just announced in the last like 24 hours, uh, with the new desktop update and the new routines, um, I think that, that gives Anthropic a big upper hand. But one of the most powerful, I think, is Cowork Dispatch, right? That to me, right, especially if you run a lot of agentic flows on a local machine, what happens when you're not in front of that machine, right? You have all these stories, uh, right, that, you know, people in Silicon Valley, they're, they're out at, you know, parties or whatever, and everyone's glued to their phone looking at like when their agents are done running on long runs, and then they like go home, right? They leave the party, they leave the networking event because they got to go home and, uh, you know, tend to their agents. That's what Cowork Dispatch is great. The downside with Cowork Dispatch, well, it's buggy. I'll show you that, my version, uh, right, of that. Uh, but it's all in one thread, uh, so I think the context gets a little drowsy, um, But for basic things like retrieving files, you know, when you're on the go, if you need to, you know, launch something. Again, a lot of this is when you're on the go, because if you're— if you have a full computer somewhere, you can just do it on that computer, right? Oh, normally I use a desktop, but now I can jump on a laptop. This is like when you're truly on the go, you're on a walk, you can just open up your phone and say, oh, you know, I need to send this over to a client. I know this file's there on my local machine. I can download it and send it. So what syncs and what doesn't? This is also important. So your custom skills do sync between the web and the desktop. So you can use your custom skills in Cowork and in Claude Code. That's cool. Also, your custom instructions do sync as well. That's another difference between OpenAI's Codex and ChatGPT. It doesn't sync. So here your custom instructions do sync. Downside, your chat history does not sync between the 3 different platforms. So your conversations stay siloed. Most connectors do sync across the web and the desktop, although some of their capabilities are a little different. And then plugins right now are for desktop only and not for the web. All right. Now, my gosh, 30 minutes in, we can get to what's new. I'm going to have to make this one go a little faster. All right. But there's lots to cover. So here is yesterday, brand new desktop app. Like I said, 5 hours in, new desktop app. But here's some of the things what's new in the new desktop app. So it's been redesigned, Anthropic says, from the ground up for parallel work and is a lot faster. So some of the newer features, it's an integrated terminal inside. There's in-app file editing, which I really like. There's a rebuilt diff viewer so you can see the differences between you know, your code as it goes along. There's side chats and adjustable side panel and then a large amount of other what they call, you know, quality of life. So it does look completely different. One thing I wasn't a huge fan of is toggling between chat, co-work, and code. It's not the smoothest UX, at least for me, and I'm so often using all three of them at the same time. I liked how it used to be, so they used to be kind of bigger buttons in the middle of the app, which always was very easy to kind of know and understand what you were working on because it was big, right? So now it's more of it's on the left-hand side. So, you know, if you update, you know, Claude Desktop and you're like, wait, what's going on? Yeah, the, the three little things and the words aren't always there. Maybe this is like old man Jordan shaking his fist. You only see the words for the one that you're clicked on. So at least for me, Not a big fan of that, but I know some people will be. All right. So yeah, kind of have that, that screenshot here for our audience, right? It's kind of those live in the sidebar, but it is very quick. There's also shortcut keys that you can switch between them, you know, Command 1, Command 2, Command 3. So that's good as well. If you don't just want to take your hand off the keyboard. I know a lot of people are like that. All right, so let's go over some other things that are new. Some of these things I already referenced, so I'll go through them quickly. But these, again, most of these are the past like 4 weeks, some of them the last 8 weeks, but this is like a year worth of updates. All right, so this one also from yesterday, and this is routines in Claude Code. All right, so these are kind of similar to schedules, um, that they had before. But a little different. And at least right now, these are just in Claude Code on the desktop. And I think I'm going to show at the end, I'm going to show a live-ish demo of this. We'll see how well it works. So essentially, you can configure a routine once, which whether that's a prompt, pulling something from a repo on GitHub or your connectors, which that part's huge. You can also do this with computer use. I was trying this out as well, although they didn't say that in their release. I always try to try things out so I can tell you all, right? So you can configure it once and then you can run it on a schedule from an API call or in a response to an event, which is really cool. So routines run. This is the great part. You can run them locally on your machine or you can run them in the cloud so you don't always have to keep your laptop open. So that's one thing with Cowork and the scheduled tasks in there, right? Sometimes even though I like— I didn't understand it. I would run it one day, you know, click bypass permissions and then, you know, oh, have it run at 2:00 AM because it's a hard task and it would ask for permissions each and every time, even though I said always allow permissions, I already ran it once manually approved it, right? So sometimes, you know, these permissions technically they're a good thing, right? That companies, you know, are always safe with security and permissions, but sometimes with some things it's like, yes, just please run it so you can run overnight. So some of these things, at least for me, running these scheduled options weren't always as good. Routines seem to be a lot better. And the fact that you can trigger them from an API call. So essentially that's a webhook. And there's so many possibilities. I mean, we could do a 5-hour show just on that. I'm not going to, don't worry. Right. Or in response to an event. All right. And then, yeah, I posted this. There is no mentions of these running on desktop. but they obviously do, right? So, uh, in the announcement, this is just that they were, uh, on the web and they showed it on the web, but they work on desktop. All right, next, um, another new update. Already talked about this, not going to take too much time, but computer use via dispatch. All right, so this is essentially a link between the Claude app iOS on iOS and the Claude app on your computer. So the way that Claude or Anthropic says it online, they say to assign a task from your phone, turn your attention to something else and come back to see finished work on your computer. Tell Claude once to scan your email every morning or pull a report every Friday, and it handles it from there. All right, sounds great in theory, right? And I had a great 1 or 2 days with this batch when it worked. Then it just broke, right? So I have a screenshot here. This is from, from my phone, right? It says, request too large, max 20 megabytes. Try with a smaller file. And the bad thing is there's only one ongoing thread in dispatch on your phone and on the Claude Cowork dispatch tab on your computer. So essentially I just got a weird error, right? I didn't upload a file that was more than 20 megabytes, but it says I did. Maybe because I used it too much. I have no clue. Right. But to get this to work, I'm going to have to probably wipe my account clean, uninstall everything, which I didn't really want to do, even though this was one of my most favorite features of 2026. Again, some of these things, I don't know if it's because Anthropic is shipping so fast, it's too fast. A lot of these things are freaking buggy, right? The updates that came out yesterday, again, haven't given them the full run, but they've been a lot better so far. But at least for me, dispatch was one of my favorite things I used, but it broke right away and there's nothing I could do to fix it. Right. Regular updates don't fix it. Logging in and out don't fix it. Right. I tried some, you know, Claude helping me fix it. That didn't work. So if it works for you, it's amazing. But little, little buggy. All right. Next, computer use in Claude Code. All right. So this was from the end of March. So this has only been like 2 weeks. But computer use, this is what they said on Twitter. Computer use is now on Claude Code. Claude can open your apps click through your UI and test what it built right from the command line interface. The cool thing about this, so obviously Cloud Code on the desktop has a preview option, so it runs, it can render and run the code inside the program, right? But if you've ever done anything in development, you know that, well, you really just want to actually run it on your computer, run the actual thing, not a simulation, right? So this is really cool. Cloud Code can literally, even if you, and this is great if you don't know coding at all, this is a great way to help. It can literally build you an app, launch the app, and use the app. Right? Yeah. Kind of crazy. But think of that for anything because, you know, you can have it launch, you know, if you're having a problem, I don't know, in Slack, right? Instead of trying to describe it, just send a screenshot, have Claude Code open Slack and say, go find the setting or go fix this right? Or at least tell me what's going on. So we can, you know, literally use and see anything on your computer and control it. All right, so yeah, they say Cloud Code on Desktop can now preview your running apps, review your code, and handle CL failures and run PRs in the background. All right, next, projects available in Cloud Cowork. This was a very welcomed update. So they said keep your tasks in context in one place, focus on one area of work. Files and instructions stay on your computer, so you can also import existing projects in one click or start fresh. So the short way of how this works is this is a better way to organize inside Cowork, because like I said, when Cowork first came out, it was a little scattered in the same way that you can use projects on the web. To better organize things. Same thing here in Cloud Cowork. All right, next. Yeah, I told you there was a lot. Scheduled tasks. This was technically at the end of February, and scheduled tasks are really for Cowork. And for me anyways, they've been really buggy, right? And some of the things I wasn't as comfortable doing in Cloud Code just because I'm not as familiar, right? Or I'm not as experienced of a developer, even though I probably, I don't know, 15 different apps. At least 4 of them I use pretty consistently. So scheduled tasks were in Claude Cowork. So what they say, Claude can now complete recurring tasks at specific times automatically. A morning brief, weekly spreadsheet updates, Friday team presentations, et cetera. Right. So all these different things in Claude Cowork, the non-technical version of Claude Code, you can schedule things. All right. So those are all the updates. That's what's new. That's the differences between A, B, and C, using it on the web, Claude Cowork versus this, apps, skills, plugins, the whole bit. Now let's dangerously skip permissions. All right, let's go in YOLO mode. Let's try live. FYI, y'all, I don't have great confidence this is going to work very well, but let's try it anyways. All right, it is AI at Work on Wednesdays. This is gonna be the shortest demo, I think, because I actually can't do a ton without accidentally exposing private information that I don't want out there. All right, but let me just kind of show everyone a little bit, and hopefully you can see my screen. Livestream audience, let me know. But kind of, you can at least see here the new interface. You can switch easily between chat, Co-Work and Code at the top of the screen here. But one thing I really wanted to focus on, a lot of these are, you know, more like nuanced, just better display, right? Coding things. I think there's a lot of them in Code, but I actually wanted to show Routines. And first, I'm actually going to run this. All right. So I'm actually going to have to reshare. Sorry, to do this correctly. I'm gonna have to share my entire screen because if we get this correct, we should be launching, we should be launching something in a browser if we get this right. All right, let's see how it goes. What could go wrong? All right, so now I am sharing my entire screen. Sorry if I'm jumping a little bit all over the place here. So This right here is a routine. So this is a brand new feature. The thing I really like about this is there's a calendar view, which is really cool, right? Because, uh, at least sometimes when you're trying to schedule things, uh, you know, automated runs, like in ChatGPT, it's like hidden. Like you have to either go into your settings, like 4 clicks deep, or just, I know the URL, I just put in the URL. Right. I actually love this new feature in Cloud Code. You just click Routines. And again, if you're just listening on the podcast, don't be scared by Cloud Code. They've actually made the interface even friendlier for non-technical people. All right. But the cool thing I like about routines— so these say create templated routines that can be kicked off on a schedule by API or by webhook. So you can see the list of yours. I I've been deleting them and tweaking them, uh, and I didn't want them to all show on my screen. Uh, and then there's a calendar view too. So you can click on the day. So whether you want to run something once a week, run something every day, twice a day, whatever it is. All right. So I want to show you how to click a new routine. So in the upper right-hand corner, you're going to click new routine and you have two options, local and remote. There's pros and cons, advantages and disadvantages to these. So first let's talk about local. So when you set this up, Couple things to keep in mind. All right, if you're running it tasks, right, there's different permission levels. You're gonna again make sure you have permission to do this, blah, blah, blah. But you're gonna wanna go to bypass permissions because this is literally the only way that you can use this. If you're scheduling something, you're not gonna schedule something when you're sitting in front of your computer. So there's different permissions that you're gonna set. So you should probably put bypass permissions. You can select a folder. So essentially that gives it access to read and write any files in there. So you probably shouldn't give it access to like your home drive, right? You should have a dedicated sandbox, you know, with guardrails around it depending on what you want to do. And then there's also, you can choose the model. All right. And then you can set the frequencies. You can run it manually, hourly, daily, weekends, whatever the time. All right. So this is the local version. I wanted to Okay, so that's how you would do it. You give it a name, you give it a description, and then you actually write what you want it to do in the kind of main box there. And then you can also choose the model that you want to use for each run because maybe it's something a little simpler and you don't necessarily need to use kind of your rate limit budget on Opus 4.6 Extended, right? Because that, if you're on a $20 a month plan, that's going to eat through your limits pretty quickly. All right, now let's go ahead and do a new routine remote. And I think here's where some of the true capabilities happen. So these are not happening on your local computer. However, you can see here, one of the things that you can do is run it via GitHub. So if you don't know GitHub, think of it as like the Google Drive for your code, right? You can run your code there, store it privately, whatever. Same thing, give it a name, give it a description. But here's some of the cool things. You can schedule it like a normal cron job, right? So whether you want to run it hourly, daily, weekdays, weekly, custom, etc., right? So that's cool. But then you can also, let's just say I want to run this every 10 minutes as an example. You can, you can run multiple triggers. Right? So you can do a GitHub event. So anytime something, right? So this for obviously people in software development, you know, coding, you can probably understand, you know, if a GitHub event happens, running that as a trigger is huge, right? You, you could technically automate most of your job if you're in software development or, you know, automate a good chunk of it just by that alone. Right. Also, you can do the normal schedule, a GitHub event, or API, right? So here, without getting too technical, this is essentially webhooks, right? So anything that happens on the web, right? And being able to connect with Zapier, amazing, right? So essentially anytime anything happens on the web, right? You get a new form submission on your website, a new row is added in a spreadsheet. Right? Think of any if, you know, if this, then any if thing. If there's an API, which is like everything, then you can do blank anything, right? So that's cool. But here's the real power. The real power is, well, you can use all of your connected connectors automatically, right? So I have some of my connectors down here. I have Beehive, I have Canva, I have Clay, I have Figma, I have Gmail, I have Google Calendar, and I have Slack. Right? So anytime, you know, a new submission comes in to your website, you can send a message, but then you can also use the power of stacking these together. Right? So let's just say, I don't know, as an example, when you get a new email, you have to create a calendar event, update people on Slack, and then go into Canva and, you know, update an invitation. I don't know. Right. With a little tweaking, you can literally automate all of that and you don't even have to trigger anything because you can connect it via an API. So just a ton, a ton of in just this one feature alone. Combining this with the desktop is huge. So obviously if you do the routine that's remote, you're running on Anthropic's infrastructure, right? But the good thing is, is you have different trigger events and then you can use your connectors automatically. Locally a little different. You have to make sure your computer's on, number one, right? But then the real cool thing there is, well, it has computer use. So anything that you would normally do on your computer, well, you can just do locally. All right. And here we go. Let's try it. All right. So I'm actually going to— I have an example here. All right. And I'm just going to click run now. So I have this set up to run every day at 4:30 a.m., so I'm going to run it now and we'll see how it goes. So when I click run now, I can go watch it run and hopefully it's going to work. So what I said, every day at 4:30 AM, look at my Buzzsprout stats from the day prior. So I host this podcast on Buzzsprout. All right. Did what I thought it would do. It opened it in a different window. So I'm bringing it back over. Hopefully it'll still work. All right. So it launched. And right now, if you're watching the video, I'm not touching anything. It's just doing this on my computer. It's running a series of commands that I I told it to, right? I essentially said, look at the prior day's stats, go through, compare it, you know, as an example, compare it to the last Monday, go look at the last 5 Mondays. So I know if my stats those days are going up or down. This is what I would normally do, right? Most days. So I know, okay, are you all liking these shows? Are you not? It's a lot of manual work that I have to do. And generally I do it every single day because if I'm gonna make adjustments, on the show, right? I wanna base it in stats, in real stats. So I asked it to not only compare the last day versus the previous 4 weeks, but then also each day go and look at which shows, right? 'Cause we have 750 episodes, which shows are getting plays each day. So I know, oh, a show I did 2 weeks ago is still getting hundreds of downloads a day from 2 weeks ago, right? Or I can find anomalies like that and say, okay, I should focus more on that because there's clearly a lot of demand for something like that. To get that kind of insight, I would have to spend at least an hour every single day. So let's see if it's done. I'm scrolling down to the bottom here. Yeah, it's already done. So we did that obviously in 30 seconds and I can run that every single day, right? And I can also have it save it to Google Drive. I can have it save it locally, right? Because I gave it a folder on my desktop. So I go through here, And yeah, it looks like we had a good, you know, good, good day, you know, 18, 18% better than normal. Great. All right. So that's just a very quick example of what you can do with the new routines, but went over a ton of new updates in Claude Desktop. I know this was a longer episode, but there was a lot to cover. All right. But here's the recap. The biggest downsides that no one's talking about. COWORK and Claude have no memory of your normal chats or each other. All of those are in silos, right? Which stinks because within Claude Chat, well, Claude Chat knows everything that happens in Claude Chat, but in Claude COWORK, it doesn't know what you're working on in Claude Chat, or Claude Chat doesn't know what you're working on in Claude Code. So that stinks. All your information stays in those respective silos. All right. Another downside, the web, the web app features usually are pretty stable, right? So yes, the desktop is way more powerful, but some of these new features, my example like Dispatch, just buggy, right? Usually online you don't get that. More with more power comes more bugs. All right. And then last but not least, Anthropic's model performance has reportedly been throttled, which impacts desktop as well. So there's been plenty of times when I go into the desktop and literally nothing works, right? Yes, Anthropic's seen a surge in demand. Yes, that's technically a good thing if you're trying to invest your processes into a platform that's going to continue to grow. But it doesn't really help when you're trying to finish a project that's due in a couple hours and you're like, oh well, gotta do this in, you know, ChatGPT or Gemini or whatever instead. If you did prefer doing something in Claude, the service has been extremely spotty, right? And this isn't a horse race, but if you look at the stats, people always accuse me of being in OpenAI homer or something. I'm not. I just tell you the truth, tell you how it is, right? I've used Claude recently a lot more than in previous quarters or in previous months, but I'm still a heavy ChatGPT user. ChatGPT doesn't go down, literally doesn't go down. Go look at their uptime monitors on each company's respective websites. Anthropic's services are routinely down across the board. All right, so that is a downside that not a lot of people are talking about. All right, quickly recapping the upsides that you need to take advantage of right now. If you're only using Claude on the web, I just gave you a dozen or so reasons to stop using it just on the web, right? If you're only using it on the web, you're missing out on 80% of what Claude can actually do with these skills, plugins, connectors, and desktop features. Anthropic is turning into, well, in OpenClaw competitor, right? They cut off the ability to use your paid plan on OpenClaw. So now you have to pay, pay via the API, which a lot of people didn't want to do. Well, one of the reasons why they're doing that is, well, they can charge more for those people that do still want to use their models. But I think ultimately, well, they're building their own version of OpenClaw via the desktop app. All right. And the new app or the new update that came out yesterday only solidifies that with these routines. All right. Claude Code is not technical. All right. Just because it has the word Claude Code in it, it's not technical. You can just go in there and say, build me an app that does A, B, and C, and it's going to build it. You don't have to do anything else. All right. And now with the routines and connectors, it's one of the most compelling use cases right now, right, to do some of those things that I showed you just automated and you don't have to go in and trigger them yourself, it's huge. And Claude Dispatch with computer use, when it works well, felt like a cheat code, right? That you had a true AI assistant that could essentially access your life no matter where you were, even if you weren't in front of a computer, right? It's buggy for me, very buggy, but when it works, and for those people that are using it, it is really, really good. It. All right. That is a wrap. Remember, if you want early access, whenever we do launch the Everyday Cowork, uh, whether it's a cohort or a course, make sure to repost this show. So if you're listening on the podcast, we always put a link to the LinkedIn livestream in the show notes. So make sure to check your show notes, go repost this, and I will essentially put you at the top of the list to notify you once we do open that up. It is going to be like everything we do. It's going to be free, right? Other people literally charge thousands of dollars to do this. And I will venture to say ours will probably be a little better and hopefully more fun. All right. So if you do want access to that, repost this. I don't know if it's going to be May, if it's going to be June, because I know we have a lot of big updates coming here over the next 2 or so weeks that are going to probably have my hands tied a little bit. All right. That's it. If this was helpful, please go to youreverydayai.com. Sign up for the free daily newsletter. Thanks for tuning in. Hope to see you tomorrow and every day for more Everyday AI. Thanks, y'all.
And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.
More from LARRUZ


Ep 882: Rogue AI Agents: Why Breakouts are Happening More and How Companies Should Prepare (Replay)


Ep 881: Faster AI Agents, Fewer Human Coworkers: The Overly Productive Future of Managing Agents? (Replay)


Ep 879: New: OpenAI Presence. Has The AI Customer Service Takeover Finally Arrived? (Replay)




