Don't Use Claude (Until You Watch This)

This is literally the video I would have wished to get when I started using Claude as a professional, not as a coder, as a professional who needs to get real work done in the business, and that's what I will share with you.

This video is split into two parts. In the beginning, I will go really deep into the Claude ecosystem because this is completely confusing, and I think I found a way to explain this very simply. Claude co-work versus code and so on. And then after this, I will dive into Claude code, the terminal version. Bear with me and show you how I use this because after going through this environment, you will clearly see and understand why you shouldn't use Claude co-work or the Claude code on the desktop version, and it is much simpler than you think. In fact, I think it's much simpler to use the Claude code terminal version than anything in the Claude desktop version. It's always easier to show with images. So, let's dive right in.

All right, let's start with Claude code CLI or the terminal version. Okay, this is where people are scared of, and it looks so much for coders. But this is once you are professional using it, and in fact, we have thousands of our members who switched from co-work to Claude code for this exact reason. And this is what you get when you just open your Finder on Mac and you search for Terminal. You get this terminal. And if you don't know how to install it, you can just search for it: "How to install Claude." And you come up to this page here for the quick start, and you just need to use this command that looks like code, but in fact, it is just telling it to go to the website to install Claude code for you in the terminal. So you just copy this, you go to the terminal, and paste this in with Command V. Now you see it's setting up Claude code installed. It added this to this local location. So I just need to write Claude now, and it will launch Claude. And here we are. And now I can start simply asking Claude any questions, as you might be used to from the Claude desktop.

But before we dive into using it in the terminal, I wanted to start at this end of the Claude ecosystem because on top of this, the rest of the ecosystem is built. So there is Claude code, and then we have the next layer on top, there is Claude desktop. And the Claude desktop is what you know with the chat where you can switch to co-work and code if you downloaded the desktop version. But with this visualization you see here now, that Claude code on desktop and on mobile, and Claude co-work, both are based on the original Claude code. It is just stripped-down versions where you have less and less features available. So in Claude code, it's less than in Claude. It's even less features than you would have in the raw version. And this is once you start using this version, you have features at your hand as a professional that are really helpful to get work done that you will certainly miss inside the desktop version once you know about them, and that's what we will get in the second part of the video. So that's why it's a good entrance point to use Claude code work to get a few things done, and or even Claude code to make small things to work on smaller projects. But the moment you really want to scale this and become really efficient with Claude, this is the moment where you should go to the terminal, no doubt about this. And we have thousands of people who agree exactly with this, and they are non-coders, they're not even tech-savvy, and yet they, once they get the point, started using the terminal, never looked back.

And here you see now Claude chat, and this is on the outside, and this is the least functional version of Claude that you could use. This is the ChatGPT of Claude, if you will, where you simply have a chatbot that you can talk with AI, and you can give it a bit of context. And we come into this in a moment. But now, let's add the browser. So there's also Claude in browser, and that might be the entrance point for you where you started signing up for Claude and you started there. And this is where you only get access to Claude chat and Claude code, but not to Claude co-work. And on top, if you're using it in Chrome, you can install the Claude Chrome extension, which is even less functional than anything before because here you have no context whatsoever. So this means here you can really use Claude just to interact with your browser. However, that's the beauty. You don't need to use this Claude Chrome extension directly on the browser. You can let Claude code and also the others use this extension to work for you. So this means you just give it whatever you need in here. Look up something on the browser, do something, go to this website, and use the Claude Chrome extension. That's what you tell it. And then it will use this Chrome browser extension installation, opens the browser for you, goes to the website, and does whatever you like to, but with full context in mind. And that's what we always want to have working with Claude. We want to have context available so it gives you the best results possible for your specific needs.

Now let's talk a bit about the functionalities of all the different items here. Let's first talk about projects. That's something you might have seen in co-work and in chat. You have this project feature where you can upload static information so you can give it context. It's like if you ever used NotesLM, it's the same principle. You upload whatever you need for the project to work on, and then this Claude version has this context available for you to work with. So therefore, Claude co-work for some reason they added there too, which makes no sense at all in my opinion. And Claude chat have this project feature available. Claude code doesn't. Okay? And it doesn't matter because I'm not using projects at all, and you will see in a moment why.

Because when we look at what versions have access to a local folder, this is where Claude code, the raw version, the normal Claude code, and Claude co-work have all access to a local folder on your machine. But Claude chat doesn't. And this is where it becomes now to me crystal clear on this image why you should never use Claude chat if you really want to get work done because of context. A local folder will be always dynamic working with Claude, and projects will be static. What does this mean? I told you, you upload information to a destination on Claude that Claude can then use as reference, as context to work with. So let's say you upload business documentation, and you want to talk about your business strategy and think about this. It will reference to this business information, but it will never update anything that you uploaded. This is different if you use a local folder. It's much more accessible because you probably have some documentation there already. I can just drag and drop this thing into my local folder. I can start working inside this folder, and it can start updating the information there. So, let's say I have a document called "Business Strategy." I can now talk with Claude about my business strategy based on this original context. But then when we iterate, we will shift the business strategy, Claude can update this document for me. So we have the latest version of the strategy. So for the next session, I can reference this. This won't happen in projects. In projects, you can talk through these things, you can get an output, but it will never update your documents after you concluded something. This is the dynamic thing. This is the persistent memory that you can build for Claude over time, and that's what we leverage the most to build our AI teams. Something I'm not diving into in this video, but it's all based on this local folder access.

Then we also have GitHub access. So access to a repo. And what this does, it just means you can connect your Claude desktop version and your Claude code to an external storage in the cloud where you have versioning. So let's say, again, getting back to this example, if I talked about the business strategy and it updated the document for me with the new strategy, using something like a repo in GitHub allows me versioning. So Claude will always be able to go backwards in time to see the full state of the local folder that we had there, and even roll back things or extract things from the past. But this is actually something that in this video we don't worry about because you shouldn't upload sensitive information to GitHub or to a repo. This is more when we go into building an AI team, as I showed on this channel many times already, that it becomes useful because I have much more control of the different iterations that Claude built for me.

Okay, but this is it about context. And you see Claude Chrome, there is no context at all. There's no projects, there's no local folder, there's nothing. However, I can always access this extension through my context directly here. So the moment I'm using Claude code in either version, or even Claude co-work, I can have the local dynamic context available and tell it to access the Claude Chrome extension with this context in mind to do work for me on the web. And that's what you need to keep in mind here.

One more thing for Claude chat, I want just to quickly add here is the artifacts. And that's again another confusing thing compared to folders. And there's also a reason why it's not appearing here. If we go to Claude chat and we open artifacts, this is whenever you had a conversation with Claude chat and it created something for you, it will just store it here in an endless list of things. Maybe it was a mini app, maybe it was a document that it created, or whatever. And this is called an artifact. But this is just lost in the void, in my opinion. Hard to recover and find. And this is where it's just a playful thing to get these "aha" moments. "Wow, look what is possible! I built a game in just a few minutes." But in fact, it's completely useless as it is not persistent and building on top of my basic context. And this is where everything that I create with either of these three, it will be stored in this local folder, and therefore directly accessible, directly in context of it. I can have different local folders that I just opened because, as you can see, when you go to co-work or to Claude code, you have now here the option down here where you can choose a different folder, or you can pick a project in here, something I never did, as you see. But I can choose a local folder, and there are different local folders that I'm working in. So the moment I attach this folder, Claude knows that I'm now working on this local project, if you will, or in this environment, and so on. And the same applies to Claude code. You can just open Claude code, and here you have the button to work on local and choose a folder you want to work in. And it is the same principle as for Claude co-work. And the same applies for terminal. I can always just write down `cd` and then drag in any folder I want to work in and hit enter. And now I'm with this terminal inside the folder.

Another way would to do this if you don't have a terminal open yet, you can right-click, go to Services, and here you can say "New Terminal at Folder." And it opens this terminal directly on the folder. So there's really no excuse for a non-coder that you cannot open a terminal based on a folder. It even tells you in what folder it's in. And now I just need to hit Claude, and I'm with Claude working now in this folder. So I can just side by side open up this folder, and Claude is now working inside this folder. And this is all you need to know.

Now let's quickly talk about automations. We have something called schedules. And this is where usually things become even more confusing for people because there are different types of scheduling. So in Claude Co-work, we have something called "schedule." And in Claude code on desktop, we have something called "routines." I will use a different icon also. I'm not sure that there's a big difference. And in our terminal, if you hit slash, I have "schedule" again. So here I can schedule things or loop. And we get into this when I dive deeper into how you use Claude co-work. There are different versions there, but let's say they are much closer to each other again here versus this is a separate thing. And all this means is that you can use scheduled to schedule tasks like this. So let's say you have a local folder where you scan documents into, and you want every day everything organized by AI. You could perfectly set this up here. It would itself, and you can set up the timer and so on, and even manually run it to test it. And include co-work routines that work just similar. It's the same setup, but it's called routines. I don't know why people need to confuse other people all the time with this. I don't know if they make fun of us in some way or form, but they needed to name this differently, give it slightly different functionalities, but in the end, it's the same thing. And in the back end, as to my understanding, it's it's running a cron job. Don't worry what this is. It will just ensure that on your local machine, it gets triggered. So, this means your machine needs to be switched on in order for it to work.

But again, if you go to Claude code docs, you can go down here, and here you see a comparison between the different versions of the schedules that you have. And it says here for scheduling that survives independent of any session, use routines desktop. So it's all about that you can close your conversation, and it will just launch it itself. And that's where it becomes agentic and just works on autopilot for you. If you heard about Open Claude, this is where we get into this area here. And now you see here the comparison, and you have an option where inside the Claude, you can launch this too. And this is where we have the schedule command, and you see it's different intervals that I can choose from. It doesn't have access to your local files. It just accesses the GitHub that I just mentioned here. So it's pretty confusing, and the more you switch between all the different things, the more confusing it will become. That's why sticking to the raw version makes it much easier to stay on top of using Claude. No doubt about it.

And I showed you the loop command too. Again, this is something that runs on your local machine, but it will stop at a certain time. It's not something you set up and forget. It will be something if you're working on something and you're waiting for something to come in, for example, you could constantly let Claude check if there's an update, but eventually it will stop this work. And then you have proper cron jobs, which means you build automations on your local machine. This is where you can also use Claude. You don't even need to use any of these commands. So let's not dive deeper into this, as we just get confused. This picture says it all. We certainly have a way to automate things. Claude chat doesn't have anything. Claude Chrome doesn't.

And now something that we missed on this overview is Claude mobile. And again, that's where it becomes even more confusing because on the mobile app of Claude, you only have access to Claude code and to Claude chat, but not to Claude co-work, except for this dispatch thing that you can use to spawn Claude co-work. And yet, it's still confusing because you have this here too. There's this dispatch, and this is just one continuous feed where you can keep talking to your local machine this way. There's a dispatch in co-work, and then there's also dispatch in chat, and it is just confusing. I tested it, and it cannot even access the folder I'm looking for and so on. I don't know if it is actually using it properly, finding the right things. I cannot trust it because this is like having Claude chat on my root of my computer now, tries to find the different things I work on, and it's just confusing.

The way I do this, it's as simple as that. I'm in a Claude session working on a local folder. I can hit slash remote, go down remote control, and this starts a session right away. And there's nothing else I need to do because I'm logged in with my Claude account. And if I'm logged in on my Claude mobile app, I can directly access this session via my mobile phone. All I need to do is open up code here. And there you have the idle session. And here we are. And now I can say "hi." And you see it instantly appears on my terminal on my computer, and therefore I have direct access to my terminal this way. And if I open up Claude now on the browser, and I go to code, you see here is also the same session in a browser. This means I can now go to a different computer, open up my Claude on the browser, and I can access the local terminal session via the browser. So you see, if I now go here in the browser and I say "hi" again, you see it instantly sends it to my local version inside the terminal, and you see it's directly synced. So here I'm in the browser, and here I'm there. So there are many ways how I can access this. And you can also connect WhatsApp, Telegram, and so on to your terminals this way. Not something of the scope of this video, but it's nothing special, even if people try to tell you it this way.

One more thing I want to say about the Git repo that I mentioned here. There is no need to upload this to the cloud. You can create this locally. And if you just tell Claude to create a Git repo locally, you get this thing. It will be just in your local folder. It doesn't go anywhere. Nothing goes out to the web. But you will have this version control going on. And that's really useful. And I made a video about this already if you want to dive deeper.

So after this, you might have already got an idea that this was the most stable version to all the features. I can access it via mobile. I can access via Claude code the browser. I can access the local files. I can set up schedules, and everything stays the same since I start using it nearly a year ago. While here, it's a mess. Every day they update something else. Something is broken. UI is not working, and functionalities are not there. And this is where I want to show you the features that I use in Claude code that are not available in Claude co-work at all, and limited in Claude code and desktop for some reason. And what I mean by this are the slash commands. They are really powerful. There's nothing you need to install. They are built into Claude code right from the get-go. And those are the ones that I use on a daily basis.

So let's start with `/clear`. Okay. And you see I don't never need to write it out. I can directly select commands based on this. But what this does, it clears my context window. Okay. Everything, the whole chat history is now empty, and I can start from scratch. Why would you do this? Well, if this fills up over time, Claude becomes dumber and dumber if you stay in the same session all the time. So you could use one of the other commands, `exit`, to go out of Claude and restart Claude. But that's something you might don't want to do, right? That's just another restart of the whole machine, which is not necessary. That's where `clear` is enough to clear the context window. You can start on a fresh slate, but with the context of this folder in mind. So keep this in mind. Okay. Now we are back in the terminal. I also can write `clear` to clear the terminal itself. Okay, so it's the same command in both worlds.

Now I can say Claude again. And what I'm using on top of this are flags, which is just these two hyphens and followed by those words that just launches Claude in a different state. So `--dangerously-skip-permissions` sounds very dangerous. It bypasses permissions. It's the version I'm usually working in. The thing is, I'm pretty sure what he is doing, what I'm working on. But if I'm unsure, I can always now, inside the current session, switch between the different modes by just hitting Shift Tab. You have this new `auto` mode, which is like bypassing permissions but with more control and a bit more guardrails in it. We can just keep it empty, which essentially will just ask you for anything that Claude will do, any update on any file. It will ask for permission. And here you have `accept-edits`. This will just take into account things that you previously already accepted in working in this folder. It will keep it in mind and it won't ask again. But this is where I really use usually this or `auto` mode, the new one, because this gives me the most agentic workflows, as now I can kick off a task and come back later to see if it is finished. Also, when I remote access it, it's easier to do. And I'm usually working with at least six terminals open in parallel, working on completely different things, and therefore I don't want to hit "accept all" all the time. And that's why I want to have this activated. However, this is where it becomes important to use this `plan` mode whenever you are not sure about something. So, plan mode ensures that Claude doesn't do any changes yet. It will actually plan out the steps that it is going to do. That's why I always do the plan mode upfront, so it comes up with the plan, so I can review what it is about to do. And if I'm happy with it, then I switch to bypass permissions and I let Claude code do the work. And the plan mode, you will see it will make Claude set in a completely different thinking. It approaches the problem a different way. It researches, it gives, asks you follow-up questions that you might be used to from Claude co-work. So let's have a look here again. Where is the plan mode available? It is in the Claude code desktop version, and it's Claude code terminal. In Claude co-work, there is no plan work. However, it seems that they set this up with custom prompting in the back end. Maybe because code co-work is much more proactive, asking you follow-up questions or clarifying questions if it doesn't get the point what you want, and to ensure that you get the best outcome. Whereas here, I can actively switch into planning mode, then I plan things out, and once we want to execute, I have the plan. I can even store the plan as a text file locally, so I have it for later reference, and many more things this way that gives me more control.

Another thing is now new is the ultra plan mode. Something I'm not using yet because here it's actually using Claude code on the web to draft something out, and I'm not so sure about the context it has available at this point, and what it is actually uploading, because obviously I'm working on a local folder here. That's why I'm not using this mode at all right now. The same applies for this new ultra view feature. This just is much more expensive, and it makes sure to really dive deep into code bases. But for most use cases for busy professionals, this is overkill. I might use this for my coding work, but this is not yet necessary if you work just on your local documents, or you want to create documents and things like this, or note-taking, or digital journaling, all the things that I showed on this channel already.

Another command that I think is really important is `rename`. So you can just hit `rename`, and it will rename the session according to your conversation. So it says here, "Could not generate a name based on this" because obviously there's no conversation going on right now. However, you can also say `rename`, hit tab, and start writing "test session," and then you have here a "test session." This is especially useful if you have several terminals open and you're getting confused about what you're talking about in this session. This is what I do here. And you can even change the color. And now I have a very important thing that I need to stay focused on and so on. And there's another advantage. It's not just about easier overview with several terminals open, but whenever you accidentally close the terminal, which kills the session, or you leave the session for some reason, or you get a crash, or the computer crashes, the beautiful thing is if you come back into an empty terminal and you launch Claude again, you can hit now, and then here you see now all the previous sessions, even 17 hours ago. I can even hit Command A, and I have even sessions that go way beyond this. But you see on top, "test session." I hit test, and now I'm back in this session with the full context that I'm working on. Nothing was lost. And I think this is really crucial. If you ever worked on something and something happened, you will be happy that you know this command to get back to work. Obviously, that's something if you work on a desktop, you have on the side all the different sessions, and then you can just get back there to your session this way. But again, no excuse not to use terminal because this is the way you get back to.

Then, if you have a longer conversation, let's say write some random text of 100 words. Okay. So now we created something random. So let's say the context window is filling up. This means that you might have seen this already, that Claude starts compacting your work. This might happen at a point where it's really critical that it understands what's going on. You can always actively compact the conversation. So what this will do, it will go back and it will extract the key information and make sure to keep in mind what was still the open tasks and keeps the context in mind. But beware, with each compacting step, it loses context of the previous conversation. So now you can always see here, Ctrl O, I can see what Claude did here. Okay, I said compact, and here is the compact summary. And this is the new context that Claude has now available after compact finished. And this is the context that is just not visible for you now, but Claude still has in mind, so you can keep working. It still knows. So if I say, "What was the last thing that you did?" it knows now the last that it created this 100-word piece of random prose for me.

So talking about memory, this is your short-term memory, if you will, because when you leave the session, this memory is gone. And that's where we want to have persistent memory. And in my previous videos and inside our courses, I explain in detail how to build a persistent memory, a PKA, a personal knowledge assistance system where you have a local folder, and over time the AI agent helps you to build your knowledge base over and over. So whenever you launch now this Claude code inside this folder, it has the memory from before based on session logging insights that the AI does for themselves and so on. I'm not going diving deeper into this. I just want you to let you know the moment you leave the session, context is gone. If I would open now another terminal, both would have the context of the folder, but they don't have context of each other. So that's why again, it's so important to have a local folder where you have this knowledge built in the memory instead of here.

And yes, they provide something like global memory and auto-memory where Claude automatically updates any insights it gets out of your conversation. So it keeps this in mind. And I hate this because it interferes with my work inside this folder because there might be completely different topics I'm working on in different folders, and it keeps taking this global memory, which might come from a previous session that has nothing to do with the current session, and this starts to confuse Claude up. But luckily, there is an option again in the terminal, at least you can hit `/memory` and here you can, this auto-memory, you can switch it off. So I could switch it on, and this creates this auto-memory folder, and whenever you keep talking, it will update this folder. As I, for the given reason, switch this off. I have a project memory, which is the Claude.md file in this current folder here, and there is the user memory, which is the global knowledge. But I at least have control over these files on what goes in there. So I always recommend to switch this auto-memory off. And on desktop, there is this preference thing. You see it's empty. I don't want to have anything in there. But also in here, there's a memory function that keeps updating. I don't want to have this. I want to have the pure thing. So it is really focused on what I'm working on, and that's it.

There's also a neat feature that I'm not using often, but when I need it, it's really useful. Let's say I told them to plan something out, and it's currently working on something. Plan out how I can improve this folder structure that you are in. So I think he will take some time. I switch to plan mode, for example, and he starts working. And while he's working, that you always see based on this indicator here that he's working on something. I can hit now slash and hit "by the way," and I can now just ask a question. "How are you doing?" So you see it starts a separate session within here, and he keeps working on what I asked before. And here he answered me the other question. And now I could hit F to fork this into my current session. So this means while he's working on something, and I forgot to mention something about, I can use this "by the way" command to talk separate from the current work that he's on. I don't want to interrupt his work. Maybe he pulled off different agent than he's working in. And that's why I use this "by the way." I iterate on something, and then I hit F for fork. So he merges it back into the current session and keeps working along this way. Very neat feature.

Talking about agents, this is again something that is unique to Claude code. So if you use Claude code in the terminal, or at least in the desktop version, that's why I'm using it, because if you want to have a team working with you, this is where the agents come into play. Uh, the reason the last time I checked co-work, it didn't start separate agents and worked in parallel. So therefore, it is really sequential here. It starts really pulling off different agents doing work in parallel, and the output is always better. Terminal, no doubt. If you realize there's something wrong, you can always hit escape during the conversation, and it will interrupt the conversation. Another thing to keep in mind, you can use the up arrow to go to previous things that you shared with Claude code. And while he's talking, I can add new messages that will be below his working on. And the moment it fits, he puts it into it, and then it adds the context while he's working on something like the "by the way" thing, but without a back and forth in parallel. I can directly inject new messages while he's working on that. I see that he maybe is going off rail, or he needs to have additional context. That's where I just keep writing, and then while he's working, I can also when you ever add something, you hit arrow up because you you know think differently. This is the moment where I can just simply get it back and edit something more, and then bring it back in again.

You might be curious why it is showing this below, and it might look different for you. This is where you can use the status line command. If you hit tab, you have it here. And now you can just write what you want to see below the status bar, even the weather or folder count. And it will now create a section below that shows the folder count of the current folder, for example. And that's how you can update this. It is a local file. You can even manually update this. But if you don't know how to do this, you can always ask Claude. And that's the beautiful thing about it. I will cancel this because I don't want to have the count here.

Another thing is the effort that you can now adjust this way. And this changes costs, token costs, obviously. The lower the effort, the less tokens will be used, but the outcome will be also maybe not as good. So there might be different use cases where you really need to have high effort to dive deep into research, or you need to get something really right. And something that where you just do some online research or something like this, where you would lower this effort. That's where you can set up this. Another one is model, where you can select the different models. Though that's the cheapest, and this combined Sonnet, and then here the new Opus 4.7. And this combined with the effort, you can really save a lot of tokens. And if you're on a Pro plan, knowing about these features will make sure that you hit your limit for the 5 hours or for the week much, much later, not so soon as you might used to. And these are the controls, you know, you need to know about.

The last thing I want to mention is the custom commands. So if we are again in this folder, you see this Claude. If you don't see this either, it wasn't created yet because by Claude starting to work in this folder. And if you start working in a folder that it has no clue about, you could do `/init` and it will just go into the folder and check out what is there to initialize this Claude file and this Claude folder, which gives him the context next time when he starts what this is all about, what uh content is available in there, and so on. Something I usually don't do, as I actively say, "Create a Claude.md file," and this is what it should be doing and so on. So you see it analyzes it. It gives me some feedback and so on. No need to dive deeper into this. I clear again the terminal. So you know about this. The other reason why you might not see this is that it is a hidden folder. So on Mac, you just hold down Shift, Command, and dot, and this hides or shows any hidden files or folders in a folder. So that's might be useful also for other use cases if you haven't known this yet. But this is the way how you showed it. That's why it's also grayed out. And whenever there's a dot on front, this means for Mac to hide these things. And in here, we have the commands folder. Okay? And that sounds so complicated, but what this just is, is just a folder called commands. And here you can even create yourself these text files. MD is just a text file with formats, which is essentially just these hashtags that indicate there's a title, titles or subtitles, and bullet points. And that's why it's an MD file, and that giving this basic structure. But you see in here, this looks very complex, and this too. But you could just simply create an MD file that just says something that you keep saying on a recurring basis. Again, if you don't know how to create a file in here and so on, you can tell Claude again. So I you just say, "Create a /command that just has the prompt say hello," and this is it. And it will now go into the folder that we are working in currently. It will create a new text file here. See, there it is. Hello.md. I can open this. And you see all there is in there is "say hello." So this means whenever now I hit `/hello`, see it appears now here, "say hello." I hit enter, and it says "hello." Why? Because all this does is just injecting this prompt. So I can just open this with TextEdit, and I can just delete this, and I say "say goodbye," and I save this. Close it. So now in this hello.md, if you open it, it says "say goodbye." And now if I hit the slash command again and say hello, he says goodbye because I just injected this prompt into my conversation. So now you can imagine if there is a long thing that you keep repeating over and over, you know, before you close the session, do this and move the things over there and so on, and you keep recognizing there's something you keep repeating. That's the moment that you should create a command or let Claude create a command for you because you can describe what you want, and then Claude will create a much more comprehensive, structured version of it. And in fact, whenever I now say `close`, `close`, it will instead read all this what it should do to close the session. And that's what this is all about. So just be aware, the longer these are, this is what you feed into your terminal to carry out this command. This allows you to really have shortcuts for yourself now to really keep going.

And then, as I mentioned, there's this loop where I can say "every minute say hello." It will now set up this local cron job that will die over time or if I close this session. And now every minute it will tell me hello. And you see here it created this schedule task log that tells him that does this in the background. And then we also have schedule. Every morning I need a report of the changes in this folder. And here you can see that this becomes really powerful if you have information that gets into here. So let's say I have a document scanner and I keep scanning things. And here I have the folder where the scan documents appear. I could have now a schedule where he constantly just checks every day if there are new documents and starts organizing. It gives me a report. However, this only works if you have it in a Git repo. That's just something to keep in mind. And therefore, if you have some sensitive information, you probably don't do this. This being said, you can always ask Claude instead of using schedule to set up local cron jobs that will do this for you. So, you just say, "Hey, can you remind me every morning to do this?" And it will create this on your local machine for you to trigger itself to do this for you. Be careful with this kind of automations because if you overdo these things and Claude and entropic things, you do just automated stuff, and then you're not actually working with it anymore, it might ban your account, which is much less likely if you just use Claude to build all this. And because Claude itself will keep itself in the guardrails of this, no guarantee, but it's much more likely than using something like Open Claude, for example. And as you can see, the loop keeps going. So every minute we now saw hello hello coming on. So now I can say `stop loop` and it will now stop the loop, and we are back to normal. And once we completed the session, I exited, cleared, and that's it. And now I have everything that I worked in inside this folder. Next time I come in, I open my terminal in this folder and I start working again.

To me, so much easier to use Claude and much more efficient, and obviously also scalable, especially if you use a team that also lives inside this folder, therefore as an orchestrator, orchestrating the work to specialists and so on. If you're interested to learn more about this, their videos on the channel, or you join us inside my IOR if you're not there yet, because there we dive really deep into how to build this AI team, the personal knowledge assistant system, and so much more. We also do weekly coaching. So, if you really want to dive deep and your profession want to get real work done, we are here for you. And if you want to stay up to date on these things, make sure to subscribe to the channel so I can catch you up in the next.