You're Using Claude the Hard Way. This Is the New Way

0:00Forget Claude coowork and forget using Claude like a normal chatbot, right? Because the biggest Claude updates Anthropic just shipped are landing in a tab you've probably never opened. Partly because of its name and because it started as a black terminal window only engineers opened. But now it lives inside the Claude app and looks just like your chat except you type the job and it does the job. So this is the only Claude video you'll need. We go through

0:2514 things one by one in simple English. So, even if you've only ever used a chat box, you can follow along. Three of them only came out in the last few weeks. Number six is [music] design, which now sits right inside the app. Number 12 is mods, where you change [music] how Claude itself looks just by asking, and we'll build one together. And number 14 is the biggest. You type one prompt, and Claude runs a whole team for you. So, we've saved it for the end. But let's start with the one that shows you the difference fastest. Number one, you give Claude one line and it hands you this, a

1:01full research report on cryptocurrency written by five helpers working at the same time. Remember the tab we said you've probably never opened? It's this one, the code tab. And yes, that name is exactly why you've been skipping it. But you just watched it write a whole research report and nobody typed a single line of code. So let's rewind and see how we go into artifacts, start a new doc, and type do deep research on cryptocurrency. Run five parallel sub aents plus one small command at the end. The moment we hit enter, it writes itself a to-do list so you can see what stage it's at and what

1:40it's going to do next instead of staring at a loading screen and wondering if anything is happening. Then that new doc opens and it starts writing the report straight into it while the research is still going on in the background. And see this bar showing how far along the job is and what it would have cost if we were paying per use instead of on a plan. That bar is a mod, one of those new ones. And it comes from the small command at the end of the prompt. So at number 12, we'll show you how we made it. Now those sub aents are just helpers. Okay, picture five researchers,

2:13each one taking one piece of the topic. In background tasks, you can see how many are running and how many have finished. And you can open any one of them to see what it's working on and which model it's using. And while they're working, you're free. So you can open another chat, answer your emails,

2:30whatever, and come back when it's done. And when you do come back, what's waiting is a real document in its own editor. You get tabs and fonts. You can make the page wider, drop in a table or an image, add a checklist or a quote block, and you edit it exactly like a Google doc. Sharing works the same way, too. You can keep it to yourself, open it to anyone with the link, or invite someone as a viewer or an editor. And if you need it somewhere else, it exports as a word file, a PDF or markdown, or go straight into Google Docs. So, think about what you'd use this for. a report

3:05on a competitor in your industry, a city you're thinking of moving to, a market you want to get into. For a lot of everyday documents, the whole job now starts and ends right here. So that's one line from you, five helpers researching at the same time. And you could peek at any of them while they worked. And what you get at the end is a proper report you could send to your manager today. And if you want to try this along with me, that exact prompt and every other one from this video, plus a checklist of all the commands are in my free WhatsApp community, links in

3:37the description, because once you start, you'll find yourself fixing things you never thought AI could touch, like the TV in your living room. Picture your Samsung TV's home screen full of ads. One person on Reddit got fed up with exactly that, so he used Claude code to build his own home screen with a big clock and only the apps and game consoles he actually uses. Samsung gives you no official way to swap that screen out. So, he also built a small helper that changes the TV's hidden settings and puts them back if anything goes

4:09wrong, and he shared all of it for free. Another person went the other way. Instead of fixing his TV, he got tired of checking whether Claude needed him. So, he gave Claude a physical status light, a USB lamp wired up with something called as hooks, which are little triggers that fire when Claude starts, stops, or needs you. So, the lamp tells him the moment Claude is waiting on him. And the third one goes back to the TV. Someone got so annoyed with how slowly the Plex app opened on his LG TV that he rebuilt the whole app with claude code. And he's put that out for free as well. So all three of them

4:44had a problem nobody else was going to fix for them. And all they did was describe what they wanted and Claude built it. And it doesn't matter whether you're an engineer or you've never written a line of code because Claude has become one of the easiest tools out there to use. So let's go through everything it can do one piece at a time. Starting with the part you already know and building up to the new ones with mods at number 12 and the biggest

5:09one right at the end. So this is the normal claude chat and even in here a lot has changed. Right in the chat box you pick the model and the effort level and a bigger model with more effort thinks longer and handles harder work. While for a quick question you can turn both down so the answer comes faster. Right next to that you choose how much it checks with you before it does something. And we'll go deeper into that when we get to the code tab. And if you don't feel like typing at all, you can dictate. So you talk and it types for you. Or you switch to voice

5:45mode and just have a conversation with it like you're on a call. A lot of people find it easier to explain a messy problem out loud than to type it out. Right now once you start using Claude for everything, you'll end up with a lot of chats. And that's what search is for. It looks through all your chats, artifacts, sessions, projects, and scheduled tasks in one go. So nothing you made with Claude gets lost. But search only helps you find things after the fact. And that brings us to number two, projects, which keep related work together in the first place. Like this one we use for our YouTube community

6:22resources where every chat inside shares the same background. That's what a project is, right? It holds the context for one whole piece of work. So when you open a new chat inside it, it already knows the instructions and the files. You can add instructions, add files for context, and even give it a folder to work from. Let's say you're working with a client for 3 months. You put their brief, their brand guide, and your notes in one project. And from then on, every chat you start in there already knows

6:52who the client is and what they want. So, you stop pasting the same background into every new chat, right? And making one takes a few seconds. You click new project, give it a name, let's say test, write what it's for, add context if you want, and hit create. And your project is ready. So projects hold the background. And the next question is where everything Claude makes for you ends up. And that's number three, artifacts. Okay? Every dock, deck, design, or dashboard lands here. And you can send any of it to your friends, your family, or your team with a link. So instead of a long answer sitting in a

7:29chat window, you get a proper file with its own page and you can come back to it next week and pick up where you left off. Between docs, slides, designs, and dashboards, it's slowly turning into its own little Google workspace right inside Claude. And the good part is this works the same way in the chat and in Claude code. So whatever you learn here carries straight over. Now everything so far

7:53still waits for you to type something. But number four doesn't. Look at this morning brief. It showed up at 8 a.m. without anyone typing anything. Pulled today's calendar and the important unread emails and put what needs attention first at the top. That's a scheduled task. And this is basically your AI scheduleuler. Okay. You give it a time and a job and it runs that job for you on repeat. So you're setting up an automation without touching any automation tool. And setting one up is just a short message, right? In scheduled, we clicked new task, then create with Claude and typed, [music] "Set up a scheduled task that gives me a

8:30morning brief each day at 8:00 a.m. Keep it short and scannable." Then it asked us to connect Gmail and calendar because it can't read what it can't reach. Asked a couple of questions about what we wanted in the brief, and that was the whole setup. From then on, it sits in your schedule list where you can run it right now to test it, pause it when you're on holiday, edit it, or delete it whenever you want. If you're wondering what to schedule, there are templates for a daily briefing, inbox triage, meeting prep, a weekly review, content ideas, or keeping an eye on a topic, or you can set one up by hand. Give it a

9:05name, write the instructions, and pick the time. Let's say you run a small business. Every Monday morning, it could go through last week's customer emails and tell you who's still waiting on a reply. Or every evening, it could pull the news in your industry, so you start the next day already caught up. But the easiest way is still create with Claude where you just tell it what you want and you can even ramble in Hindi or whatever language you think in and it understands you. Just know that scheduled tasks need

9:33a paid plan and they run in the cloud. So the work happens even when your laptop is shut. So you set it once and the work starts before you've even opened your laptop. But there's still one thing that wastes a lot of your time and that's explaining the same job to Claude over and over. And that's exactly what number five fixes skills. This is discover. And when we recorded this, there were 57 categories of ready-made skills in here. Think about it. Every time you start a new chat, you explain the same job again. How you want it formatted and what steps to follow, right? A skill is a saved way of doing

10:11one job. Okay? You explain it once and after that you just call it by name and Claude follows the same steps every single time. The simplest place to start is under customize where a few skills come builtin and you just switch them on. Let's say we turn on engineering, then finance, then design, and they're ready for the next chat. Now, let's go back to discover the page we opened this one on. And look at the number next to each category. Engineering alone had

10:39over 1,200 skills. Code had about 1,100. Design and media had 445. And there's sales, there's even games. So, whatever you're working on, search for it here first and try it with one prompt for your own use case because someone has probably already written down the steps you were about to explain. And if you've got your own way of doing something, you click add, and you can upload a skill, create one with Claude, or record your

11:07screen. That last one is bloody clever. On a Mac, you record yourself doing a boring task once, let's say, updating the same Google sheet every morning, and Claude watches that recording and turns it into a skill, so you never do it by hand again. So, think about the jobs you repeat every week, the way you write a client proposal, the format of your weekly report, the checklist you run before sending an invoice. Each of those can be a skill, right? And number six is one more way to make Claude do things your way. Look at this five slide deck

11:40on Claude's top five/comands. Every slide is in the same colors, fonts, and style, and it came out of a tab you might not have found yet. That tab is called design. It used to be its own separate app, and it only recently moved inside the Claude desktop app. So, if you don't see it in your sidebar, click more, then edit sidebar, drag it up, and click done. And while we're on that, everything we're doing in this video is in the Claude desktop app. So, if you've only used Claude in the

12:10browser, that's the one to download. Inside design, you can make slides, you can design things, and you can set up a design system, which is basically your brand written down once. Okay? You can give it your own colors, typography, and spacing and make it yours. Or you can just describe the feel you want and ask it to make one for you. Let's say you make carousels or graphics for your business every week. This is what keeps all of them looking like they came from the same brand even when five different people are making them. We let it make ours and it came out playful and digital. And look how much it covers,

12:46right? There are principles. There's how the brand should sound. Every color has a job for surfaces, borders, text, and accents. And then there's type, spacing, corner radius, components, accessibility, and even the icons. That's the stuff a designer would usually hand you in a brand book. And here it's sitting in one place ready for Claude to use every time it makes something for you. And that's how we got the deck. We asked it for a five slide presentation on the top five claude/comands and it built it using

13:18that brand on Opus 5.5 at medium effort. And once it's built, the whole deck stays editable. You click on any text and change it. Change the colors or the typography. Make something bold or italic or underlined or leave a comment on a slide and let Claude fix it for you so you're never stuck with what it gave you the first time. There's one more option in design called motion for motion videos, but right now it's in beta for team and enterprise plans, so most of you won't see it yet. So design takes care of how things look. And number seven is getting clawed into the apps your work already lives in. And

13:56back under customize, that's what connectors do. Okay, let's say Notion or Gmail. You click on Gmail, click add. It opens your browser. You log in and pick your account and it sends you straight back into Claude connected. It's literally one click and a login. And once it's connected, you can ask things like, "Find the last three emails from this client and tell me what they're waiting on." And it goes into your inbox and finds them instead of you copying emails into the chat. And if connectors bring your apps in, plugins bring in whole new abilities. They're one level bigger than skills. A bundle of skills and commands that come together. Okay,

14:34you can browse the ones from Anthropic, add more from a marketplace, or straight from a link, upload one, or create your own. The easiest way to add one is straight from GitHub, which is basically a big public library where people share these for free. Let's take a design plugin called Impeccable, which stops the things Claude designs for you from looking generic. This is its page on

14:56GitHub. And all we do is copy its link. We don't download anything, right? Then we come back to Claude, go to customize, then plugins, and click create with Claude, which opens a normal chat. We paste that link in, send it, and Claude reads that page and gets the plug-in ready for our account. Claude then asked if we wanted our own copy of it to change, a brand new plug-in built the

15:19same way, or just the plug-in installed. We picked install. It asked whether we wanted it in the Claude app, in Claude code, or both. We picked both. And that was the whole install. And we did all of that from the normal chat using something straight off GitHub without opening a terminal or reading a single setup guide. So, the normal chat already

15:41does a lot more than answer questions. And that's the version most of us use every day. But the part that changes how you work is one tab over. And that's also where the two biggest new updates live. And that's number eight, the code tab, which is claude code living inside the same app. The name scares a lot of people off, but there's no coding in anything we do here. And you'll see it's almost the same box you've been typing into. If you ask me, it's the most underrated part of Claude right now because it's where Claude stops just answering and starts working inside your files on your computer for as long as

16:22the job takes. And this is why we said forget Co-work at the start. If you use co-work, this will feel familiar because co-work is built on the same engine as Claude code. The difference is that the newest things in this video like mods and running six sessions at once live here in the code tab. So, this is where you want to be. On the left, you've got your earlier sessions. These are some of our previous builds. And in the middle, it's just a chat box, same as [snorts] before. Quick note on plans before we go further. Docs, slides, and design work on the free plan. And the code tab needs

16:56pro or above. Now, the first thing you choose in here is where it works. Local means it works on your own computer with your own files. And a cloud session runs on Anthropics machines, so it keeps going even if you close the app. Then you pick a folder, one you've worked in recently or a brand new one. And that folder is its desk, right? It works on whatever is inside it. So give it one folder per job and it stays focused on that job. And once it's working inside your files, the next question is how much it can do without asking you. And you set that right here in the

17:30permission menu, which is where people slip up. Manual is the careful one. It asks you before it edits a file, runs a command, or goes online. And accept edits lets it create and change files inside your folder without asking. Plan is where you should start because it looks around, writes you a plan, and

17:49changes nothing until you approve it. And auto lets it work without the routine questions, while a separate model checks every action and blocks anything that goes beyond what you asked for. And by default, it still blocks things like permanently deleting files that were already in your folder before the session started. Right? Let's say you ask it to clean up a messy downloads folder. In plan mode, it first tells you these 40 files go into an invoices folder. These go into photos and these old installers get deleted. And then you get to say yes or change the plan before a single file moves. And then in settings, there's a bypass option. Turn

18:28that on and it won't ask you anything at all. It can change or delete files and just keeps moving. And Anthropic's own docs say to use that only on a separate test machine with nothing important on it. So keep it off. With the settings sorted, the box itself is simple. A lot of people use it like a best friend or a brain dump, right? You explain what you want in plain English or in your own language and it gets the job done. The trick that makes it work better is giving it a goal instead of a task. And there's a command for exactly that, which we'll get to. Search works exactly

19:02like it does in the chat. And so does artifacts, the same home for your docs, decks, and designs. Okay. And from here, artifacts can pull in the brand we built in design. And that's how we made our next deck. Here it is. A nine slide deck, cover to closing slide, all of it in our brand's colors and fonts. And the whole thing came from one line. Let's rewind. We opened artifacts in the code tab and kicked into one. And right there it lets us choose a design system. So we picked the one we made in design. Then we typed create a presentation on claude co-work versus claude code. And that's

19:39the payoff of setting up your brand once because every deck after that starts looking like you. You pick the model and the effort right in the box. Same as in chat. And there's also fast mode which is quicker but costs about double and comes out of extra usage credits. So leave it off unless you're in a hurry. Then it starts building and you can

19:59slide this panel away while it works. Under environment it says this computer and we pointed it at one folder, a new one in downloads called claude co-work versus code. So everything for this job stays in one place and you know exactly where to find it later. And while that deck is building, look at number nine. It's that same session open on a phone

20:20and it's still working on the laptop. It's called remote control. We type / remote-control on the laptop. Open the clawed app on the phone and there it is. Right. So, we can check on it or give it the next instruction while we're stuck in Bengaluru traffic. That means you can start a long job before you leave the office. Check how far it's got on the

20:40way home and tell it what to do next. All from your phone while your laptop does the heavy lifting. Okay. it needs a paid plan. Now, back on the laptop, number 10 is the set of panels around the chat box. This one is its memory for the session, something called the context window, and you can think of it like a whiteboard Claude writes on while it works. On the model it uses by

21:03default, that's up to a million tokens. So, it can hold a really long job in its head without losing the start. Okay? And when the whiteboard fills up, it squeezes the older stuff into a short summary by itself. So, the job keeps going. Next to it, the connectors you set up in the chat are here, too. And you can connect or reconnect them from this panel. There's also a terminal in here that works on the same files as Claude. And if your folder is connected to Git, there's a view that shows you exactly what Claude changed file by file. You won't need either of those for

21:36anything in this video, so don't worry about them. The one you will use is the built-in browser. It uses a clean profile separate from your own Chrome. So none of your login or history are in there. And you can open any website right here. Let's say apple.com. Claude uses it too to open what it's built and check that it works before it tells you it's done. And the first time it wants to click or type on any website, it asks you first. And the files pane shows every file it's made in that folder. and you can open and edit them right here without hunting through Finder or File

22:11Explorer. Now remember the deck is still building this whole time. And that's where number 11 comes in. Say we have a question but we don't want to stop it, right? So we open a side chat with command and semicolon on a Mac or control and semicolon on Windows or you just type /btw and we ask what it's covering. It answers in a few seconds from what's already in the chat. Slides one to four are done. 5 to nine are still planned and the main job never stops. And the side chat doesn't add anything back into the main conversation. So you can ask a silly question without messing up the job. You

22:48can open as many side chats as you want. Just remember they're gone once you close the app. And a couple of minutes later, there's the full deck. Nine slides. Every one of them in our design style. Now, normal Claude Chat can make a deck like this, too. and slides are even on the free plan. What the code tab gives you is that side window where you ask a question halfway through a job and

23:10the [music] job keeps going. And number 12 is mods. The new one we've been building up to since the start. Remember that progress bar from the research report at the start and that little command at the end of the prompt? That bar came from us and we made it just by asking for it. When the report was running, this bar stayed on screen and showed us how far along it was. A third done, then more than half, and what the job would have cost if we were paying per use instead of on a plan. Mod is Anthropic's own word for it. Okay, they only released mods at the start of

23:47October. And a mod is a change to Claude's own screen that you make just by describing it. So, let's open a new session and ask for one. Create a mod that shows a progress bar and the estimated cost while a task runs. Trigger it with / aagent-info. Claude writes the mod and reloads the app. And from then on, whenever we add / aent-info to a prompt, the mod starts when the prompt starts, the bar sits there while it works, and it goes away by itself when the job is done. And you

24:18can make one for whatever bugs you. Let's say you want a timer that shows how long each job took, or an alert when a long task finishes, or a little panel that keeps your to-do list next to the chat. You open a new chat, describe the change you want, and it's there. One warning, though, a mod runs with your permissions, so only use ones you made yourself or trust. And mods live only in the code tab. So, the app bends to how you work. If something's missing, you ask for it and it's there the next time you open Claude Code. And just like the chat, the code tab has projects, too,

24:53except here they start from a folder. You pick the folder, start a chat, and every chat you add inside it shares the same background. So, a new chat starts with the same folder and files as the ones before it. So, it's one project per big piece of work. Let's say your YouTube channel, a client or your own business, and everything for it stays together, the chats, the files, and the folder. And if you started something in a normal chat and want to keep it going, you make a new section, let's say crypto, move that chat into it, and then keep adding chats under it, like this

25:27one, asking what Bitcoin is. Schedule tasks and customize work here exactly like they did in the chat. So, there's nothing new to learn there. So you start a job at your desk, you steer it from your phone, and when you come back next week, the project still has the background that job needs, but that's still one job at a time. And the newest update of the lot, number 14, changes that. Number 13 starts with the skills from earlier. Watch what happens when we type a slash and the word YouTube. Our own editing skill pops up, ready to use, because any skill you've added shows up here the moment you type a slash. That

26:05same slash menu is where the commands live, too. There are a lot of them, and some are pretty technical, so here are seven worth knowing, going from the one you'll use least to the one worth learning today. The first is /advisor, which lets Claude check with a second model at the key moments of a task. A bit like asking a senior for a second opinion before a big decision. So, you can keep a faster, cheaper model doing the routine work and only bring in the

26:32stronger one at the moments that matter. And Anthropic's own docs say that usually costs less than running the stronger one the whole time. Next is slashcle, which wipes the slate and starts a fresh session with empty memory. And you want that when Claude starts mixing up two different jobs, right? New job, new slate. Then there's /usage or /cost, which shows you what this session has cost and how much of your plan is gone. So check it before a big job and you're never surprised halfway through. After that, /d doctor runs a checkup on your setup when something's broken, tells you what's

27:08wrong, and fixes it once you say yes. And it can also go through the instructions you've given Claude and tell you what to clean up. So, if something stops working, type /docctor before you type anything else. Fifth is /deep research. You follow it with a topic, let's say Bitcoin, and it starts a research job on it straight away. The same kind of work you saw at the start

27:30of this video. Sixth is /btw. The side chat you saw with the deck. Just faster to open and you'll use it more than you think. And the last one is slashgoal. Instead of a task, you give it a finish line. Let's say all 10 product descriptions written and each one under a 100 words. And Claude keeps going turn after turn until a second

27:52model checks that the goal is met. Because there's a clear finish line, there's something to check the work against. So it doesn't stop halfway and wait for you to type continue. If you learn one command from this video, learn slashgo because you stop checking in on it and just hand over the whole job. And

28:09it's freaking useful. And that brings us to number 14, which only came out in the middle of October, and it's the one I promised at the start. This is six clawed sessions on one screen. Each one working on a different part of the same YouTube video at the same time. We didn't open these six by hand, right? We typed one simple prompt into the main session asking it to spin up a thread for each part of the video, like the thumbnail, the title, the outline, the visuals, and the editing. And the main session started a nested session for each part. And they all got to work at the same time. Think

28:49of the main one as the producer on a shoot. You only talk to the producer and the producer hands out the work. So the thumbnail session works only on thumbnail ideas, the title session only on titles, the outline session only on the structure and so on. And each one gets on with its part without waiting for the others. And because every nested session gets its own memory, the main chat doesn't fill up with everybody's work. And as a result, nothing gets mixed up. Now you might be thinking we

29:17already saw five helpers at the start. So what's new here? Those five were helpers inside one session all working on one report. Here every part gets a whole session of its own with its own memory. The main session coordinates all of them and each of those sessions can still send out its own helpers. Right? And that's also how it's different from the projects we saw earlier. Those give all their chats the same background for a job that runs for weeks. While nested sessions split one big job into parts that each work on their own. One thing to know though, this can look a little different depending on your plan. On Pro

29:54and Max, Anthropic is building this same idea into a new version of projects where one main chat runs threads for you in the cloud and they're letting people in through a wait list. So, if you don't see it yet, that's why. To watch them all, we drag a session across or click the three dots and choose open in split view and keep adding them. 1 2 3 4 5 6 all running side by side. And this is the best way to multitask with Claude because you can watch every part moving at once instead of switching between tabs. And when they're done, each one reports back and the main session hands

30:30us one plan with every part in it. This is where the way most of us use chat stops making sense for big jobs. That way, you're one person typing six prompts one after another and stitching the answers together yourself. Here you're managing six and you only ever talk to one. So here's how to start tomorrow. Take the first thing you were going to ask chat GPT or Claude and type it into the code tab in plan mode instead. Or if you're on the free plan, start with a dock or a deck in chat because the habit is the same. Then this week, set up one scheduled task, turn

31:05one job you repeat into a skill, and give one big job a/goal. Do those three and you'll stop using Claude like a search box. And you don't have to remember any of this because every prompt from today, including the one for the mod and the one that [music] splits a job into six sessions, is in my free WhatsApp community. The links in the description. So, go grab them and try the mod first because that's the one that makes all of this click. And if you want to hear from the other side of this race, we went to the US and sat down with the people building it. We asked

31:36OpenAI's head of product, the person who runs chat GPT and Codex, what happens to our jobs? And at Meta, we tested whether this is the end of the smartphone. Both are on your screen right now. So start with the OpenAI one and watch the Meta one right after. And if you haven't subscribed yet, do it now because Claude keeps adding new things every few weeks like mods and these six sessions. And every time it does, we'll show you

32:02exactly how to use it in plain English. I'll see you in the next one.