Claude Just Changed Completely: Here's How It Works (In 2026)
A full settings-to-scheduled-tasks walkthrough of Claude's post-update interface — Chat, Cowork, Projects, Connectors, Skills, and the model tiers that decide what any of it costs.
Posted
yesterday
Duration
Format
Tutorial
educational
Views
12.1K
750 likes
Big Idea
The argument in one line.
Claude's 2026 interface is organized around one decision — Chat for conversation versus Cowork for delegated work — and everything else, Projects, Connectors, Skills, and Scheduled Tasks, exists to make Cowork tasks repeatable and automatic.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
Someone who already has a Claude account but hasn't touched settings, projects, connectors, or scheduled tasks yet.
A solo operator who wants Claude to run recurring busywork, like inbox triage or CRM updates, without re-prompting every time.
Someone confused by how Chat, Cowork, Projects, and Skills relate to each other after Claude's recent interface changes.
SKIP IF…
You already run scheduled Cowork tasks and connectors daily — this is an onboarding-level walkthrough, not an advanced workflow guide.
You're only interested in Claude Code or Claude Design — both are explicitly out of scope in this video.
TL;DR
The full version, fast.
Claude's interface splits into two modes: Chat for conversation, Cowork for delegated tasks Claude completes on its own, both sharing one dashboard of recents, pins, and projects. Projects bundle custom instructions, persistent memory, and uploaded files so every chat or Cowork task inside inherits the same context. Connectors extend Cowork into real tools, reading Gmail and writing straight into a Notion CRM, and any workflow run once can be saved as a slash-command skill or promoted into a fully scheduled task that fires automatically. Usage is capped by a fast-resetting session limit and a weekly all-models limit; picking Sonnet as the default, Opus for harder tasks, Fable sparingly, is the biggest lever for not running out mid-week. Desktop Cowork can also read local files, but that access only reaches mobile while the computer and desktop app stay open.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Sign up, compare pricing tiers, then configure account-level settings: search & reference chats, generate memory from chat history, and a pasted custom-instructions template.
02:43 – 05:21
02 · Chat, Code, and Cowork
The core distinction: Chat for conversation, Cowork for delegated work, demonstrated with a sleep-optimization chat versus a Cowork-built HTML report on the same topic.
05:21 – 08:12
03 · Projects
Projects bundle instructions, project-level memory, and uploaded context files; existing chats and Cowork tasks can be pinned or moved into a project after the fact.
08:12 – 12:12
04 · Connectors, Skills, and Scheduled Tasks
Connectors link Claude to Gmail and Notion; a live demo has Cowork read new lead activity from Gmail and log it into a Notion CRM, then the workflow is saved as a skill and promoted into a daily 8 AM scheduled task.
12:12 – 14:07
05 · Usage Limits and Models
Session limits reset multiple times a day; the weekly all-models limit resets once a week. Model tiers run Haiku, Sonnet (recommended default), Opus, and Fable, in order of cost and power.
14:07 – 16:22
06 · Desktop, Cloud, and Local Files
The desktop app's Cowork mode can be pointed at a local folder and read/edit files on the computer directly; a cloud icon on a running task shows whether it's executing server-side or depends on the local machine.
16:22 – 18:11
07 · Claude on Mobile
Cloud-based Cowork tasks started on desktop remain viewable and continuable from the mobile app; tasks that touch local files require the originating computer and desktop app to stay on.
18:11 – 19:07
08 · Claude Design
A brief pointer to a separate video covering Claude Design, for prototypes, slides, documents, wireframes, and animations, plus the sign-off ask for a like and comment.
Atomic Insights
Lines worth screenshotting.
Claude's dashboard splits into two modes: Chat for back-and-forth conversation, and Cowork for delegating a task that runs and reports back on its own.
Turning on Generate Memory From Chat History lets Claude carry context across every future conversation automatically, not just within the current chat.
Projects bundle instructions, memory, and file context into one workspace, so every chat or Cowork task started inside it inherits the same rules.
Connectors give Claude read and write access to real tools like Gmail and Notion, letting it pull new leads from an inbox and log them straight into a CRM.
Any workflow run once inside Cowork can be saved as a reusable skill and invoked later with a single slash command.
A skill can be upgraded into a fully scheduled task that runs automatically at a set time every day, with no manual trigger required.
Session usage resets multiple times a day, but the weekly all-models limit resets only once a week and blocks all access if it fills up.
The four model tiers trade speed for power: Haiku is fastest and weakest, Sonnet is the recommended default, Opus handles complex tasks, and Fable is the most expensive and should be used sparingly.
On desktop, Cowork can read and write files in a local folder on your computer without any manual upload, once you grant folder access.
A Cowork task only stays reachable from your phone after closing your laptop if it never touched local files — any task that read from your computer's file system needs that computer and the desktop app to stay on.
Takeaway
One mode decision drives everything else in the new Claude
WHAT TO LEARN
Chat handles conversation and Cowork handles delegated work, and every other feature, projects, connectors, skills, scheduled tasks, exists to make Cowork repeatable and automatic.
01Account, Plans, and Settings
The free plan works for trying Claude, but real use benefits from the Pro plan, which adds Claude Code, Cowork, and Design on top of more usage.
Every paid tier layers on top of the one below it, so upgrading later doesn't mean redoing your setup.
02Chat, Code, and Cowork
Chat is for a back-and-forth conversation where the end goal is the conversation itself; Cowork is for handing off a task Claude completes and reports back on.
Mixing them up rarely breaks anything, but learning which mode fits which request is what actually improves the results.
03Projects
A project bundles instructions, a persistent project-level memory, and uploaded reference files so every chat or Cowork task started inside it inherits the same rules.
Project memory updates on its own as Claude learns more about the work done inside that specific project, separate from global account memory.
04Connectors, Skills, and Scheduled Tasks
Connectors give Claude read and, in some cases, write access to real third-party tools, not just the ability to talk about them.
In the demo, Cowork read new Gmail activity, matched it to leads in a Notion CRM, updated the records, and flagged the hottest deals on its own.
Any workflow proven once can be saved as a skill and re-triggered later with a single slash command instead of retyping the full prompt.
That same skill can be promoted one step further into a scheduled task that runs automatically at a set time every day, with no manual trigger.
05Usage Limits and Models
Session usage resets multiple times a day, but the weekly all-models limit resets only once a week and blocks all access if it fills up.
Sonnet is the recommended default for the best balance of intelligence and cost; Opus is for complex tasks; Fable is the most expensive and should be used sparingly.
06Desktop, Cloud, and Local Files
The desktop app can grant Cowork direct access to a folder on your computer, letting it read and edit local files without any manual upload.
A cloud icon on a running task confirms it's executing on Anthropic's servers rather than depending on your machine staying on.
07Claude on Mobile
A Cowork task started on desktop keeps running and stays viewable from the web or mobile app as long as it's fully cloud-based.
The moment a task touches local files on your computer, that computer and the desktop app both have to stay on for the task to remain reachable from your phone.
Glossary
Terms worth knowing.
Cowork
Claude's delegated-task mode, used when you want Claude to go do work and report back, as opposed to Chat's back-and-forth conversation.
Connector
A third-party integration that gives Claude read or write access to an external tool, such as Gmail, Google Drive, or Notion.
Skill
A saved, repeatable workflow that can be re-triggered later by typing a forward slash and the skill's name.
Scheduled task
A skill or workflow set to run automatically on a recurring interval, such as every morning at a fixed time.
Project
A dedicated workspace that bundles custom instructions, a persistent memory, and uploaded reference files around one ongoing topic.
Fable
Claude's most powerful and most expensive model tier, positioned above Opus for the hardest tasks and meant to be used sparingly.
short definitional one-liner, works as a standalone explainer clip→ TikTok hook↗ Tweet quote
17:57
“The distinction here is, is it accessing files on your computer?”
resolves the whole desktop/mobile confusion in one sentence→ IG reel cold open↗ Tweet quote
13:28
“I suggest that most people start by using Sonnet as the default. It's a good balance between intelligence and cost.”
direct actionable recommendation viewers search for→ newsletter pull-quote↗ Tweet quote
18:16
“If you watch this entire video, you've already mastered 90% of what it can do.”
payoff line that validates the viewer for finishing→ TikTok hook↗ Tweet quote
04:13
“The other option is Claude Cowork. Cowork is what you wanna use when you want Claude to do work for you.”
the single sentence that defines the video's core distinction→ IG reel cold open↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphor
00:00After receiving a massive number of updates over the past few weeks, Claude is now more powerful than ever before. But opening it up for the first time may feel confusing if you're relying on any Claude tutorial that's more than a week old. So this video will be the most up to date guide you need to get started with the new Claude.
00:21I'll cover everything that matters and show you step by step how to use Claude in a way that anyone can understand. You can start by coming to claude.ai. If you don't have an account already, you can sign up for one here.
00:32Then you can take a look at the different pricing plans. The good news is that you can try out Claude for free. However, if you are going to use this seriously, I would suggest you at least sign up for their pro plan.
00:43That includes more usage than the free plan and includes Claude Code, Claude CoWork, and Claude Design. We'll cover all of that in this video. Once you have an account and you're logged in, your dashboard will look like this.
00:57Right here in the middle is the prompt box. This is where you can begin chatting with Claude. If you've ever used ChatGPT or Gemini in the past, this will work very similarly.
01:08You type in a prompt, send it along, and then you can chat back and forth with Claude. But before you start really using this, there are some settings that I suggest you change. Come down to the bottom left.
01:19You should see your name here, and you can click on that. Then click on settings. We'll start by coming over to capabilities.
01:26You wanna make sure that search and reference chats is toggled on. And if you see generate memory from chat history, you want that toggled on as well. That will allow Claude to learn about you as time goes on so that the responses it sends you in the future are more personalized for you.
01:45If you've already been using an AI provider in the past, you can click right here where it says start import, and that will walk you through the process of importing your memory from that other AI provider. Next, let's come on the left here where it says general.
02:00Down here, you'll see the instructions for Claude. This is where you can give Claude instructions and context that you wanted to keep in mind across all of your chats.
02:10What I've put together for you is this template where you can fill in just some basic information about who you are, what you're gonna use Claude for, how you want Claude to respond, and any other context you wanted to know about you. You could spend just a couple of minutes filling this out, copy it, and then paste it into instructions.
02:28Then Claude will know all of that about you and how you want it to act and respond when you're chatting with it. There are a lot of other settings in here, but the default settings for all those are more than fine.
02:41So we can leave those alone and then close the settings. Next, look on the top left here. You should notice that you can toggle between home and code.
02:50Code is Claude code. This is designed for coders and developers. We're not gonna be spending any time going over that in this video, but there are plenty other videos on YouTube if you are a coder and developer as to how to use Claude code.
03:04So we're gonna stick inside of home. Home is made up of two different parts. Inside the prompt box right here, we can see both of them.
03:13We have regular chat and we have co work. Chat is where you go when you just wanna have a back and forth conversation with Claude.
03:22For instance, here, I asked Claude chat, what is the best way to optimize my sleep? It sent me a response. I asked it a follow-up question.
03:30It sent me another response, and we had this back and forth conversation. So if the end goal is the chat itself, that is when you'd wanna use Claude chat.
03:40The other option is Claude co work. Co work is what you wanna use when you want Claude to do work for you. For example, right here, I told Claude Cowork to put together a presentation for me about how to optimize sleep.
03:54I told it to go out and find the most current research, look at some anecdotal reports, and put together an HTML report.
04:02And it not only gave me the information like it did in the chat, but also put together this presentation for me. There's definitely some overlap between what you can do with chat and what you can do with co work, but it is important to know when you should use which one. The good news though is that if you mess this up and you accidentally start a chat in co work or you have chat do some work for you, it's not gonna matter too much.
04:29But the better that you get at delegating the right task to the right mode, the better the results are you were going to get out of Claude. That will just come with some time and experience.
04:41What you'll also notice is that both chat and co work share the same dashboard. So if we come over here on the left under the recent section, you'll notice that right here is the chat that I just started.
04:51And we know that this is a chat because it has a chat bubble next to it. This is my coworker task, and we know that because it has this checklist next to it.
05:01This will show you all your most recent chats and tasks sorted so that the one that you most recently interacted with appears at the top. For any of these, you can click on these three dots on the right and pin them so that they get moved up here and don't get lost. You could also click on these three dots and rename them or add them to a project.
05:22We haven't talked about projects yet, but you can find those by clicking up here where it says projects. Projects are dedicated workspaces where you can organize chats, files, and instructions around one ongoing task or topic.
05:35For instance, right here, I have one called invoices. If we open this up, you'll notice that right here on the right, I have instructions that apply to any chat or co work task that I start from within this project. So since this is the project that I use for creating invoices, these instructions tell it to always output invoices in a particular style.
05:58There's also a section here for memory. This is a memory system that exists across the project. So as Claude learns more about the type of work I do inside my invoices project, it will update this memory so that's more helpful in the future.
06:15You can also add context, things like PDFs, documents, or other text that Clawd can use as a reference when you work inside this project.
06:25To set up a new project, you can click on the left where it says projects, click on new project, and give it a name.
06:32Then you can click on create project. And over here on the left, you can click on the plus icon, give it the project instructions, then click on the plus icon under context, and upload any files from your device, GitHub, or Google Drive.
06:46If you have some projects that you use quite frequently, you can return back to projects over here on the left and pin them so that they always appear on the left sidebar. So you can click on these three dots, click on pin.
06:59And now over here on the left under pinned, you'll not only see your pinned tasks and chats, but you'll also see your pinned projects. This makes it a lot easier to return back to them later.
07:10Now like I said before, any chat or co work task you start inside a project will stay inside this project. So over here, as this is working, we can return back to the sleep project, and we can see that chat right here.
07:25But you can also add other tasks and chats that you didn't start from within the project. Over here on the left is the task we started before. I can click on these three dots, hover over add to project, and choose the sleep one.
07:40So now that appears right here. And then here is the chat that we had about sleep, and I'll do the same thing. So what I wanna point out for you here is not only that you can move things into projects, but also that you can have both chat chats and co work tasks within projects.
07:59I'll say in general, pretty much everything we're setting up here in Claude works with both chat and co work. So all the settings we went over before, the instructions for Claude and the projects work with chat and co work. Both of them also use the same connectors and skills.
08:14Connectors are how you connect Claude with third party tools. If you click on the left here where it says customize, that will open up the settings.
08:23And then down here, you can click on connectors. These are all the third party connectors that you currently have added to your Cloud account. You can click on the top right here and click on browse connectors to see new ones that you can add.
08:38Some of the most popular ones are, of course, Gmail, Google Drive, and Calendar, but you can come in here and find any tool that you commonly use and see if it's available as a connector. And what's great about that is now Claude will have access to the information inside that app or tool you use and may even be able to write to the app or tool that you use.
09:01Here, for instance, if I close this out and return back to my connectors, you can see that Claude has access to my Gmail account, and it has access to my Notion account.
09:12So let's say you use Notion with your team to keep track of different sales leads. I put together this mock CRM right here with all the leads that we're following up on along with some comments that some of my teammates have left, their email, the last time we've contacted them, etcetera.
09:31What I'm going to ask Claude to do is to look through my Gmail for any new contact that we've had with these leads and to bring that information in to this CRM. For this type of task, I'm not going to use chat. I'm going to use co work because I actually want it to go out and do some work for me.
09:52I will tell it, go through my Gmail inbox. Find any new updates from my leads that I have in my sales lead CRM in Notion. Then update the database, give me a report on what's new.
10:05And you can see now it is logging the email activity into the records on Notion and flagging the hottest deals with a comment. It's doing all of that through the Notion connector.
10:17Now that it's done, it's giving me an entire report on everything that it found. And opening up the CRM, we can see all those changes in here. So that's how connectors work.
10:27They quite literally connect Claude with third party tools. Now this workflow that we just went through is great, and I wanna be able to do this same workflow again in the future. Now what I could do is start a new task in the future and just give it the exact same prompt that I gave it just a couple of minutes ago.
10:49But instead, I'm going to turn this into a skill. A skill is just a repeatable workflow, and they're so easy to set up because I can just tell Claude here, now turn this into a skill so every time I ask you to update my lead CRM, you go through this workflow.
11:06Once you send it along, it will then create the skill for you. You can click right here to save the skill. And then when you click on new on the top left, you can invoke this skill from either chat or from within Cowork by typing forward slash and typing in the name of it, which was CRM, I believe.
11:24Yes. Update lead CRM. So if I click on that and send it along, it will then execute that exact same workflow, and I never have to repeat myself ever again.
11:33But let's take this one step further because I want this exact workflow to run for me every morning automatically at 8AM. So I will return back to the task where we first executed this workflow.
11:48And this time, I will tell it, run this for me every morning at 8AM. And sending that along, it will then set up the scheduled task, which you can then approve.
11:59That will run every morning for me at 8AM. If I wanna make any changes to it, I can click on the left here where it says scheduled. This will show me all my scheduled tasks, and I can either pause them, edit them, or delete them.
12:11Before I show you how Claude works on the desktop and mobile app, I wanna quickly talk about usage limits and the different models. If you click on your name on the bottom left and open up the settings, over here on the left, you can click on usage.
12:26In your account, you may see two or three of these different usage meters here. Everyone should see this one, which shows your current session. This resets multiple times per day.
12:38If it reaches a 100%, you're not gonna be able to use Claude until it resets. You should also have a section with your weekly limits, and you probably have one called all models. If this one fills up, you're not gonna be able to use Claude until it resets for the week.
12:53All of these limits are currently shared between chat and co work, but that may change in the future. There's also a limit meter down here for Fable. We'll talk about that in a moment because that may or may not be available by the time you're watching this video.
13:07Closing this out, you can then come to the prompt box to see all of the models that are available to you. The one at the bottom, which is Haiku, is the weakest model, so it will go through your usage slower.
13:19The one at the top, which will be either Opus or Fable, is the most powerful model that you have access to, which will go through your usage faster. As things currently stand, I suggest that most people start by using Sonnet as the default.
13:33It's a good balance between intelligence and cost. For more complex tasks, you can then jump up to Opus. And if you find yourself barely getting through your usage limits every day and every week, you can make Opus your default model.
13:47Fable is a bit of an unknown right now because as of the recording of this video, it is included with your subscription. But Anthropic has said that in the future, it won't be included with your subscription. Rumor has it that may change.
14:02Either way, it's the most expensive model, so you wanna use it very sparingly. Let's now jump over to the Claw desktop app. This is it right here.
14:11And the first thing you'll notice is that this looks almost identical to the web app. But under the surface, there are some special things going on here with Cowork that unlock some capabilities that you do not get on the web. Now chat is going to work exactly the same as it does on the web.
14:28But if you switch over to Cowork, you'll notice that it now says here, do you wanna work inside a project or a folder? That's because the desktop app has access to local files on your computer. With this selected, I can now click where it says choose a different folder.
14:45Now I can select a folder on my computer. Here, for instance, is a folder I have called budget. Inside of this, I have all of these files, these Excel files going over my budget, my expenses, my financial goals, all of that.
15:00So selecting that, Claude Cowork will now have access to those files that are inside that folder. I didn't need to attach them.
15:10I didn't need to upload them. It sees what is on my computer if I give it permission for that. And using that, I can now ask Claude Cowork to create reports for me, documents, spreadsheets, presentations using the information in those files.
15:27This can allow Cowork to look at the files, edit them, or create new files directly inside my folder on my computer. As this is working, let me show you something else.
15:39Anytime you start a new task using Claude Cowork on the desktop app, you always wanna check up here for this symbol. If you see a cloud, that means it is running in the cloud. So if we return back to the web app here we go.
15:54We're back in the web app. Over here on the left, you can see that under recents, I have that task that I just started on my desktop app.
16:05It is available in the cloud, and it's still accessing my local files even though I'm viewing and interacting with it here on the web. That will work as long as your computer is on and the Claude desktop app is open.
16:20That is particularly helpful when you use the Claude mobile app. Opening it up, you can use it with regular Claude chat here. And opening up the side menu, you can see all your recent Claude chats here.
16:33But if you click on co work, that will show you all the co work tasks that are running in the cloud. Here's the one that we just started. Opening this up, you can see the response that it sent when it was finished, and here's the report.
16:47So this means that you can start a task on the desktop app and see the results and finish and continue the conversation on your phone, again, as long as your computer is on and the Claude desktop app is open. One final point I wanna clarify because I don't want there to be any confusion here.
17:06When you start a new co work task using the Claude desktop app, you will be able to access that task from any other device like your phone if the task that you start is completely cloud based.
17:20In other words, if I just come in here and tell it to go through my Gmail and then to pull the information from Gmail into Notion, all of that is cloud based. So I'll be able to access it on my phone even if I close my laptop screen before it's finished completing that task.
17:39However, if I run a task like this one where I had it accessing local files on my computer, I would have to start the task and leave the computer on with the Clawd desktop app open in order to be able to access that task from my phone.
17:57The distinction here is, is it accessing files on your computer? If it is, then it needs that connection, and your computer needs to be on in order for you to be able to communicate to it with your phone or from the web app.
18:11So that is a new Claude. If you watch this entire video, you've already mastered 90% of what it can do.
18:17Now many creators will lock a course like this behind a paywall, but I put all my content for free on YouTube because I want everyone to learn how to master these AI tools. All I ask in return is for you to give this video a thumbs up and leave a comment letting me know that you appreciate videos like this. Honestly, I read every single comment and it's what motivates me to keep making more content.
18:39The final piece of Claude that we haven't covered yet is Claude design, and that's because it's an entirely different beast. And since I already made an entire video about it, I'll just link that for you right here.
18:52You can use Claude Design to create prototypes, slides, documents, wireframes, or animations. You can see an example of one of those right here.
19:00So if you wanna learn how you can do that yourself, click on this video, and I'll see you there in one second. Bye for now.
The Hook
The bait, then the rug-pull.
Claude shipped a wave of interface changes in quick succession, and if the tutorial you learned from is more than a week old it's probably already wrong. This walkthrough resets the mental model from scratch: what Chat and Cowork actually are, and everything built around them to make Cowork repeatable.
Frameworks
Named ideas worth stealing.
03:07concept
Chat vs. Cowork
Chat: back-and-forth conversation, use when the end goal is the conversation itself
Cowork: delegated task, use when you want Claude to go do work and report back
The single mode decision that determines how every other Claude feature should be used.
Steal fordeciding whether to open a chat or hand off a task in any AI tool with both modes
13:15list
Model tier ladder
Haiku 4.5 — fastest, weakest, burns usage slowest
Sonnet 5 — recommended default, best balance of intelligence and cost
Opus 4.8 — for complex tasks
Fable 5 — most powerful and most expensive, use sparingly
A cost-ordered ladder for picking a model based on task difficulty instead of defaulting to the top tier.
Steal forany usage-capped AI subscription with multiple model tiers
10:53model
Skill to scheduled task pipeline
Run the workflow once manually
Tell Claude to turn it into a skill
Invoke the skill with a slash command
Tell Claude to run that same task on a recurring schedule
A four-step path from a one-off prompt to a fully automated recurring task.
Steal forany repeatable business workflow, e.g. lead follow-up, reporting, inbox triage
CTA Breakdown
How they asked for the click.
VERBAL ASK
18:16subscribe
“give this video a thumbs up and leave a comment letting me know that you appreciate videos like this”
Frames the ask around free-content generosity ("I put all my content for free on YouTube") before making it, then states he reads every comment — soft rather than hard sell.
A 28-minute weekly roundup that stress-tests Fable 5 on real work, unpacks Apple's biggest AI leap yet, and explains why NotebookLM just became a research agent.