Claude Cowork 101: How to automate your workday without touching code | JJ Englert (Tenex)

one of the things I love about having this project system. You're orchestrating now. Even though you might not be a developer, you're now an AI orchestrator. You can run many agents at once and have this top-level view to be able to just quickly see which agent needs your attention and just give them those permissions or to pop back in.

I don't know, this is like very sad. All my friends are agents. I'm a solo founder. It is like very hard to just me by myself ensure that all my ideas are great.

When you use sub-agents, It will spin up 3 different agents, and each of those agents can have their own persona with like a fresh context window, meaning like fresh perspective to go and look at your work in an objective way.

If you're a product manager, build it and like put your boss in there, your engineering partner, and your customer and say, every PRD review from these 3 points of view and give me feedback. If you're in marketing, your ICP, I think there's just a lot of places where this multi-view feedback mechanism is super useful.

100%. We could take this a step further and say, hey, create this sub-advisory skill. Maybe 3 of the agents, like one agent is literally your boss that you send an agent to go research who your boss is, their role, their perspective, and have them simulate the feedback that they might give you before you even go to them. The sky's the limit here. You just gotta tell Claude what to do and it's gonna go do it for you.

Welcome back to How I AI. I'm Claire Vo, product leader and AI obsessive here on a mission to help you build better with these new tools. When Claude Cowork first came out, I was pretty much a skeptic, but I have been hearing more and more that Cowork is the way that everybody I know, especially those who are a little less technical than us engineers, get their daily work done, which is why I'm so excited to have JJ Englert here from 10x, who is a Cowork power user, but is going to take a step back and show us how to zero to one on Anthropic's new get work done tool. Let's get to it. This episode is brought to you by Tines, the intelligent workflow platform powering the world's most important work. Business moves faster than the systems meant to support it. Teams are stuck with repetitive tasks, scattered tools, and hard-to-reach data. AI has huge promise, but struggles when everything underneath is fragmented. Tines fixes that. It unifies your tools, data, and processes in one secure, flexible platform. Blending agentic AI, automation, and human-led intervention, teams get their time back, workflows run smarter, and AI actually delivers real value. Customers now automate over 1.5 billion actions every week. Tines is trusted by companies like Canva, Coinbase, Databricks, GitLab, Mars, and Reddit. Try Tines at tines.com. /howiai. JJ, welcome to How I AI.

It is a pleasure to be here. Thank you for having me.

I'm excited about what you're gonna show us because when Claude Cowork, which is what we're gonna talk about, first came out, I wrote this article that kind of blew up on Twitter and a bunch of podcasts talked about it, which was, who is Cowork for anyway? Because when I first tried it, It seemed like a UI that had just been slapped on top of Claude Code. And I looked at it and I said, okay, if you're not, you know, in Claude Code all day writing in the terminal, but you're not super technical, like, what is this product that's in the middle? And, you know, shame on me because as it happens, product continues to progress week over week. The Anthropic team, the Claude team is cooking nonstop. And I just kept hearing more and more from my non-technical friends, oh my God, Cowork's amazing, or, oh my God, I use Cowork to do everything in my job now. And then, you know, I've given it a spin in the past couple weeks and the UI has really gotten there. What it can do really has gotten there. And so I'm so excited for you to show us the zero to one for Cowork. And I just have to ask you, what got you hooked on this particular tool?

Yeah, I think like many, we were just experimenting with it. And then quickly I saw that how easy it was to connect your business tooling to it. Whereas like, you know, I built a lot in cloud code, but if I wanted to just connect my Slack, my Gmail, my Notion, Cowork is just one click. And then all of a sudden you have an AI processing all of that information, helping you do your job better. And it's just so easy. And so like, that's where it started. But then I started to like really like go deep on it, I'm like, oh my God, this is really helping me my day-to-day. So I use Cowork for all of like my productivity stuff of like emailing, Slack, building projects and documents. And then Cloud Code I still build with.

I love that because everybody is saying, oh, everything's gonna be a CLI or an API or a TUI, all this stuff. And I'm like, people just love clicking a button. I don't know, they're just, people want a button. So part of what I hear for you is like, It gave you a button to click to get connected to all your business data. And then when you're working in Cowork, you're in sort of like business mindset, personal productivity mindset. And then when you're in Claude Code, you're in builder mindset.

Yeah, exactly. And, uh, you know, I'll pull up my screen in a second here, but you know, the Anthropic desktop app has 3 different modes: Chat, Cowork, and Claude. We've been in Chat for a while now with ChatGPT, et cetera, like I think most of the knowledge workers are like in chat mode too, but the Cowork switch over to that tab is like the first kind of switch where these knowledge workers, even if they're not technical, can start doing more with AI. And it enables you to just do so much with AI in a very similar chat interface, but where like it's actually doing stuff for you where like in chat it's like, cool, you told me what to do, but now I gotta go do it. Cowork is literally doing it for you.

Why don't you get your screen started and you're going to show us if you were completely brand new to Cowork, if you had spent all your time in that chat tab and you were ready to click that middle tab, Cowork, in the Claude desktop app, what you would do. And I just want to call out for folks, Cowork is available in the Claude desktop app. So I think you have to be on desktop, not web. And it shows up again as one of these 3 tabs across the top. The middle one says Cowork and you have to click that and then you're gonna show us what to do once we're there.

So I have Cowork pulled up here. This is on the desktop app and they also have a mobile app as well. Mobile app does not work as well with Cowork. Uh, that's where kind of Claude Code continues to be really powerful. But Cowork, let me give you the lowdown here. So Cowork can actually access your desktop and a lot of people in the beginning days of like, hey Cowork, go and like organize folders for me, or go and like organize these receipts for me. And yes, Cowork can do that, but Cowork can do so much more as well. Cowork accesses your computer, it can access your browser, it can perform actions on your behalf. Uh, if you need to order or, you know, book reservations for dinner, it will go find, you know, the restaurants looking for you on your browser, you know, contact them, etc. Right now, when you get started with Cowork, you'll go to a new task And it's going to ask you to open up a project. Okay? Now, this is a very important concept that a lot of people kind of get stumbled on. It's like, what is a project? Well, in the simplest form, a project is a folder on your computer. This is where we are going to work out of and store our files for this project within that area. So if I look at my computer, I have a location called JJ, and within location JJ, I have projects, and this is where I build all of my projects. Now, some of these are Cloud Code projects and some of these are Cowork projects. Cowork is a really great, like, on-ramp to Cloud Code, teaching you some of the basic skills of, like, how to operate in Cloud Code without going full terminal mode, all that kind of stuff. So within this, I have a popular workspace that I use called Cloud Cowork Workspace that is, like, a generic high-level workspace. And I start off this workspace with what I call is my brain. Okay, so my brain goes into very detail of, like, What are my working preferences? Who are the team members that I work with? And if I were actually to open this up, I'm just going to hit spacebar. It's just an MD file, which is essentially like a doc file, a text file. This is a very specific series of instructions that we, you know, also known as a prompt, that Claude can read to get up to speed very quickly of who you are, how you like to work, who you work with, and all of your different preferences. This way, once I trigger a prompt within Cowork, it gets all of this information with it and just knows so much more about me. And sometimes we had like in, you know, ChatGPT, we had like that one thread of like, oh, this is my thread for like all my product marketing and this is my thread for like brainstorming about this or whatever. You had like that one thread with a lot of good context that was like very important to you, but you couldn't transfer it anywhere. If you set up the time with your folders, you'll be able to build those threads, but in a very transferable way that allows you to take that context and bring it with you every single time you want to do anything in Cowork. And that is the big unlock here. With Cowork, you're not starting over from the start. If you build up these projects, you're starting from a really good place where Cowork already knows about you, your preferences, all that kind of stuff. And then it just starts doing stuff with you. as a, a, as a partner here. All right.

Yeah. Before you show us how you'd create maybe a new project here, I just wanna call out some, some things for folks because, you know, if you're new to some of these, like I would say 201-level AI concepts going from chat to projects, you get really confused. I remembered when Anthropic came out with skills and people were like, what's a skill? I'm so— and I was like, skill is literally like a file that describes what you want to do. Like, let me just make it very simple for you. It's a file. File. And then what's a project? A project is just literally a folder with files in it. And so you don't have to be intimidated by these concepts. It's just I'm working on something, I'm going to spin up a folder, a drive for this, I'm going to put all my context and capture all my context in that one folder. So anytime I come back to work on that, I have the files that I'm working for. And this is how we work in business every day, right? You organize your G Drive or your Notion by folders. Exact same concept, except this time Claude can go in and actually do some work in, in that workspace alongside you.

Great call out. That's exactly it. And actually in this session today, we are going to be building like a daily operating system that I will make downloadable for you all to get at the end of this show. So definitely stick with us for that. I do just want to continue like overall explaining like what we're going to do, and then we're going to get hands on keyboard and start doing it together.

Great. Yeah.

Cool. So this is my general architecture here, and at the top level, I have this workspace map. And so this is essentially something I said, hey Claude, look through all the folders in this or files in this folder and generate a workspace map to help you navigate this better. So that way, whenever I'm in this workspace, Claude, and I send a prompt, Claude will look at that workspace map and say, oh hey, he needs help with a Twitter post, I'm going to go into his Twitter folder, use that skill that he's already prepared, and then write it that way. Without that workspace map, sometimes Claude has to ingest all of that stuff, which takes up more tokens, and it kind of makes it a little bit muddy because you're sending a lot of context to Claude where maybe it just needs to know how you'd like to write on Twitter and that's it. So the workspace map helps Claude navigate your folder structure in a better way and And again, it's as simple as connecting to Claude to any folder and just saying, hey Claude, create a workspace map for me. It's going to map everything out and create it for you.

Yeah, what I love about working with a tool like Claude, although other tools also do this in a desktop folder of files, is it is much faster at browsing files and figuring out what's going on in there than I am. And so you can actually, you know, I think this is a trick that anybody can use, which is whenever you're trying to orient yourself to something, some work, whether that's a repo, whether that's a folder that somebody, your colleague shared with you, you can open it in Claude and then just say, what's going on in here? And build, even build me a map to where, to where everything is, is. And that can be very helpful. But let's talk about how you actually kind of get started, cuz you're showing us kind of a little bit of the endpoint and I think people are saying, okay, this is great. I can, I can end up with this project that has my brain in it.

Yep.

Um, but, but how do you get started?

Sure. So I'm going to go into projects here and I'm going to create a new folder. Okay. And this folder is going to be called Daily Operating System. Okay. And I'm just going to bring it back into my projects there. All right. This is the start of everything. And then it's an empty folder. Now I'm going to go back into Codeworks here. I'm going to open that up here and go to different folder, projects, Daily Operating System. There we go. We're going to allow Claude to operate within that. And it's saying, hey, do you want to use both of these or just this? And I'm going to uncheck this because I only want to use this folder right now. Okay? So I can say anything I want to this, but generally what I'm going to say is, hey Claude, I want to build a daily operating system. This operating system is going to help me manage my day-to-day and become a superhero within my workplace. So I want this operating system to help me in a variety of ways. First, I'm going to have it help me work through my emails and help me not only understand my inbox, but help me write better emails. I'm also going to use it to help me look through my Slack and figure out how I need to respond to some of my teammates and then help me write some of those Slack messages. And then I also want a thinking partner, someone that can act as like a mentor to me and help me through tough decisions. So I'm going to go through and ask you some things to help build some skills, but start to lay the foundation for what this project is together. So there's a lot there. And I'd say like, this is as much as you want it to be. Like, this is very a chat-focused experience. You could just start this off and it couldn't even look like any of that. It's just like you could just start in a task. It's however you want it to be. And as you grow more comfortable with it, Claude will adapt to that. But I think the general point is like, don't let that big prompt like scare you. There's a lot of different ways to enter this. All right. Now the first part is, okay, Claude looked at my prompt and it says, okay, I have some questions for you. Now it's going to start asking me some questions. So where do you want to store these files? I want to store it in my daily operation systems folder like it recommended. How do you want to interact with the daily operating system? I'm going to be doing it on demand at the moment, which essentially means like whenever I want to interact with it, I will go into Cowork and interact with it. But I will also show you how to automate some of this stuff too. For the thinking partner, what kind of support matters most to you? I really like the career coaching and decision frameworks. Same frameworks there. All right, so we have this thing going here right now, and let me explain Cowork Projects because this is a brand new feature and I absolutely love it. So right now we are in a chat here, okay? And we can start a new chat as well. And you could see that I have active tasks going on here. And so this view is kind of like a good orchestrator view of like seeing the agents that I have working for me. But Projects allows you to take that to the next level. So what I could do is I can go into a project here and I can create a new project. And I'm going to create a new project from an existing folder. That's the same folder that we've already created with the Daily Operating System. So I'm going to go in there, Projects, Daily Operating System, Open, and it's going to open this new folder. And what is the instructions for this project? I'm going to use this folder and operating system to help me navigate the day-to-day life of a busy professional. You're going to help me with some superpowers by applying some AI magic to it all. All right. And I'm just going to keep it there for now. So we are now in a project interface. All right. And if I ever wanted to get back to that, I click on projects, see all my different projects, and then go into the project interface. Now, what is the importance of like, why are we in a project rather than a task? Well, these tasks, they all kind of work, but projects allow you To start chaining tasks together, which means shared memory. Okay, so I can find this task right here, this, this chat that I just had, and I can go and I can move it into this project. Okay, now this, this project, this chat right here, and if any other chat that I have, they all share the same context, and they all— and Claude knows what you're working on and can bring that with all the new stuff together. Okay, so if you do anything new, it is very specifically knowing like what other chats you've had about this and picking up where you left off.

Yeah, so there are 2 things that people need to know about from an organization perspective, which is one, we're going to pick a folder on your computer that contains all the files and context and all that stuff for a specific area you're working on. And then 2, you're going to attach that folder to something called Projects, which is available in the left-hand nav, to make sure that every time you're working in that folder or in that topic, you're not only working with those files, but you're working with the memory of all the other things that you're either actively working on or have worked on in the past. And then you can give it, if you see in the top right, you can give it just general instructions, which is helpful for guiding any work you do in that specific project.

And this is specific project instructions. And so each project can vary for different kinds of instructions, right? Which is also really good context to have.

Yeah, and speaking of context, if you look along the right-hand side, there is a section that says context and there's, exactly what we talked about, which is the context of the folder and then the context of the memory that's been generated about this project.

Right. And so this is like the HQ for each project that you work on, right? And it's really become so helpful for me because as you start to have more tasks, all of the different agents for that project are listed here. And you could see like a high-level orchestrator view of like which agent is working, which agent needs your help. And all of those tasks and conversations are like shared together for that shared memory. Again, we want Cowork to do stuff for us, and we want it to do it well and give us good, consistent results. Sometimes when we operate at this high level, it's getting so much information about all the different things we're doing, and the results can get confusing or fuzzy. But if you're operating within a project, it's very specific context about this specific project and your requirements for it, and it helps you get results faster. Like, the prompts are literally faster because you're sending less context. You're saving tokens because you're more focused and organized, and you just get better results in a more consistent way. All right?

Great.

So let me continue to build on this so we can eventually have something that you can download at the end of this show to start using for your own daily operating system. Sound cool?

Yeah, let's do it.

Cool. So with a couple of things that I want to call out here, connectors, so important. So to add a connector, I'm going to right-click here and I'm going to go into connectors. And I could see that I've already enabled some of my connectors, but I'm just going to click here so you can see this view. So connectors allow you to connect your systems to Cowork. Okay. So right now I have connected Gmail, Google Calendar, Google Drive, Notion, and Slack. And if I wanted to add more connectors, I can go to browse connectors and I can see a ton of connectors that I could do. It's normally like a one-click, you authenticate and it comes in and you're connected. That essentially means Cowork has access to that application. and a lot of those actions within that application. If I wanted to go into like my Gmail connection here, I could see that Cowork now has access to all of these tools and can actually do these for me. And if I wanted to get very specific about the permissions, I can go in here and say, hey, Cowork, I'm not going to allow you to create an email draft or to do something, but I am going to allow you to do this. Or sometimes I want you to do this, but always ask me for permission. So with each connector, you can go in and modify those permissions. But generally, this gives you a connection to your tools that you can use. Now I can go back into Cowork here and I'm going to go back into my projects, Daily Operating System. And the first thing I'm going to do is say, hey, analyze my Gmail, specifically the emails that I have sent in the last 30 days and use this to create a writing skill specifically for emails that I write, using, uh, my emails as an example of my writing structure. So whenever I want you to write an email in the future, you know exactly how I write my emails, and you'll use this skill. So this is going to start building an email skill for me, and it's going to do it by connecting to my Gmail, looking at my outbox of all the emails that I've sent in the last 30 days, analyzing that writing skill, and just perfectly matching the tone that I use in my writing. You know, AI is used for writing in a lot of different ways, and of course there's a lot of AI slop, but when you can feed AI like a series of like 100 messages, it could really get close to your exact writing style. And that's what this is doing here.

So just to, to kind of take a step back, connectors are really important because this is not only the, the tools that are gonna do work on your behalf, right? I, I think this is something that people might miss. Which is, yes, we can do connectors so you can draft emails to Gmail or write docs in Notion. But there's also this ingest path that can make your use of these AI tools a lot better. And I haven't seen this use case before, so I want to call it out for people, which is have AI go read your emails and come back and build a skill, which I think is really important. And skills, remember, for folks are just these like containerized little sets of instructions that can be called at any time. Go build a skill that writes emails in my tone. And, you know, I think as somebody who's developing a daily operating system, I call this your like anti-to-do list, is you come up with these lists of things that you just like never wanna do again. Like, I never wanna have to first draft my emails again. I never wanna have to like show up to a meeting late because there's a conflict on my calendar or they've been set back to back. I just want these things I never have to worry about.

Mm-hmm.

And I think as somebody who's coming in and trying to think about how could I use Cowork or something for my daily operating system, this idea of like burning down your anti-to-do list and then building skills on all those is, is really sharp. And what I think you're showing is you can use connectors plus the brains of Cowork to kind of come together and not only functionally achieve that, but achieve that in a way that's high quality for you personally.

Yeah. And the results have been really good with it too.

Yeah. Are you ha— I mean, are you really getting emails that like sound like, sound like you? Exactly. Okay.

Exactly. And normally what I'll do is I'll go into my instructions here. Mm-hmm. And, uh, I haven't done this step yet, but I'll say, never send emails on my behalf. Only write them as drafts for me to review. Right? And now this is a very clear, uh, instructions of like, hey, it's actually never gonna send anything for me 'cause I don't trust that yet. But everything that I ask it to send, it's gonna write that draft and just prepare it in my outbox. or my drafts so I can approve it and then send it as well.

By the way, you're being very chill about all your voice entry stuff. What is your voice tool of choice?

WhisperFlow.

WhisperFlow.

It's just, it's captured my heart.

I'm a monologue girl, so, you know, we all pick our favorites.

You know what? It all works. It all works.

It all works. This episode is brought to you by Cursor. If you all have been watching How I AI, you already know this. Cursor is my favorite way code with AI. Whether I'm using plan mode to build out an ambitious feature, reviewing AI-generated diffs right in my editor, or kicking off cloud agents to multi-thread our roadmap, I reach for Cursor as my favorite multi-model coding platform. Even better than building myself in Cursor, I love collaborating with BugBot to fix PRs for code security and quality and have begun relying on Cursor's automated agents to keep our codebase It's not just me. The most ambitious teams love Cursor too, including engineers at Stripe, OpenAI, and Figma. Ready to build more? We're giving $50 in Cursor credit to How I AI listeners. Claim your credits at chatprd.ai/howiai. That's $50 in Cursor credits by going to chatprd.ai/howiai. Okay, so you've started— I see a popup here. So tell me what's going on.

So this is one of the things I love about having this project system. You're orchestrating now. Even though you might not be a developer, you're now an AI orchestrator. You can run many agents at once and have this top-level view to be able to just quickly see which agent needs your attention and just give them those permissions or to pop back in. So right now we do see that, hey, this agent needs our attention. We can go in and review that, and it's actually just asking us if we want to allow this. Okay? So without even having to go in there, I can just allow that. I could go into that connector and change those privileges just to say always allow, so I don't have to do that. But in this case, that was set to ask for my permission. Okay?

This is one of my problems with Cowork, which is this is just like such a lovely non-technical experience until it shoves like a bash command in your face and is like, hey, do you wanna allow this or not?

Yeah.

And I just, for all, all the non-technical folks out there that just were like, hey, hold on, I just saw a big line of code that said, are you okay with running this? You can also like copy and paste that and ask a different Claude chat, what does this mean?

You know?

And I think what's kind of interesting about these hybrid tools that are kind of user-facing, non-technical folks, but have this like Claude code back, uh, backend is it's a nice way to actually step your way into the Claude Code experience bit by bit. But again, don't be afraid. I think you can be pretty unafraid of some of these alerts that may look technical, cuz you can always go ask, what does this mean? Um, should I do this? And I will also say that I think Cowork in particular has been tuned to be pretty user-friendly. Like there isn't a dangerously skip permissions on Cowork quite yet. And so Anything that Claude's asking you to do is probably in the realm of safe. Um, although it's important to read whenever you hit that approve button.

I was on the fence about calling this out, but people ask me Cowork versus OpenClaw, and I know you're really big in OpenClaw.

Snatch it.

And, and I, and I am too. I really, really love it. But the way that I think about this is OpenClaw is for work and business where you can trust it with your, yeah, more significant connections. And OpenClaw is incredible, but you maybe don't trust as much with your business tooling, especially in the first. It's a great personal assistant, it's a great autonomous agent, and could do so much. And so I use both together, but Cowork is like my daily business driver, and OpenClaw is more like my personal assistant, can like just help me in a lot of things with limited access to a lot of its own accounts.

Yeah, I'm full Claw, full press Claw world. Um, but again, I mean, again, I cannot go throughout, through my day without somebody texting me like, Cowork is doing my job for me. I'm hap— happy as can be. Um, so you've shown a, shown an example that I think is generally applicable to people, which is write, write my emails. What else? I mean, just kind of top hits of workflows that you think something like Cowork could do for you that you think belongs in, in your operating system. Yeah.

Yeah, for sure. So I just looking at this agent here, it analyzed 46 emails and it went in and created an email style guide for me and a voice profile and saved it to my memory as well. So I'm going to go back over here. The next thing that I really like to use is Claude as a thinking partner. Thinking partner can come in many shapes and sizes. It could be like, how do I respond to this Slack message? How do I respond to my colleague and give them good feedback? How do I think about my career and what I should do next? Now you can have a skill or a thinking partner for each one of those, or you can have a generic one. Right now we are going to just create like a thinking partner, like a slash mentor that you can just trigger and just like help you think through important decisions. And it will kind of like talk back to you in a way that you want to be supported. And so I'm going to launch another agent here and I'm going to say, help me create a new skill that is going to be our thinking partner. This thinking partner skill, when invoked, is going to allow me to think about how to respond to something, how to make a decision, how to critique a colleague that I'm working with in a good but effective and helpful way, how to think about my career at large, and just generally going to help me think through the bigger items of whatever task is at hand. And you can invoke this by saying, you know, help me think through this. Or navigate my thinking partner or anything like that. And then one last thing I do want to add is, please ask me any questions if you're unclear. I love the asking any questions if you're unclear because generally a lot of times it is, and it just forces it to really get the more context that it needs to make sure that it's building something specific to you. Now, we've talked about skills a lot, and we've actually even started creating skills. Skills are really powerful. You could think about it as like a very detailed prompt that is reusable in many different ways. And so rather than saying, hey, go off and create this newsletter for me and then tell it everything that you want your newsletter, you can do that once and have a very clear pattern for like everything in your newsletter. And then you could just say, hey, write the newsletter and it will invoke that skill and automatically use that instructions every single time.

Yeah. And for folks, I do actually have literally that skill in my, my core ChatPRD repo. I use it through Cloud Code. So I just do a /newsletter every Monday and it goes and looks through my changelog. It writes the newsletter, it does it in my voice, it formats it for the newsletter platform, it pushes it via API, and then I just go in and do a couple edits and move on with my life. So that's, that's a really great, great use case.

Let me have one more that we get going here.

Yeah.

So this is already building. Hey Claude, I want you to build another skill, and this is going to be a review agent. For this skill, you're going to launch a series of sub-agents that have each their own persona, and this can differ per the task. But an example is if we're launching a newsletter, we want to have sub-agents that represent our ICP. Maybe for me it would be AI builders, AI executives, and AI interested people. And the goal of this skill is to allow sub-agents to be able to review your work and give you honest feedback and to see from 3 different perspectives where you need to focus at your next iteration to improve the quality of your work. Let me know if you have any questions. So this is another skill that I really love and I use like across so much of my work. It's like I use it in my newsletter, I use it in my podcast, I use it in my social media writing. It's a sub-advisory skill is the general pattern. And what it allows you to do is Claude, you know, it's working with you in this context and it generates an output. When you use sub-agents, it will spin up 3 different agents and each of those agents can have their own persona. So like, hey, agent 1, you are an AI builder. Agent 2, you are someone that is so worried about security. And agent 3, maybe you're just a mom worried about AI usage or a dad worried about AI usage or just, That's an arbitrary example, but my point of this is you can define 3 individual personas with like a fresh context window, meaning like fresh perspective to go and look at your work in an objective way. That then brings back that feedback and it helps you kind of figure out like, how is this going to be received before I even send it, right? And if you include your ICP personas in there, it also helps with this a lot too. And I found really good results with this kind of framework. for creating your skills.

One thing, the one thing I think is really important to call out, as, you know, people think about where this is applicable in their work, is so many of us in the last 5, 6 years have moved to remote-first working, where getting collaborative feedback from your teammates is actually highly expensive. You have to call a meeting, everybody's busy, synchronous work is quite hard. And so being able to spin up pre-feedback feedback from AI, which will give you sort of a roundtable view of, your work allows you to come to those very expensive meetings, honestly, with much higher quality, pre-thought-out work. Um, and, and also, I don't know, this is like very sad. All my friends are agents, which is like, it's really hard to work solo. Like, I'm a solo founder. It is like very hard to just me by myself ensure that all my ideas are great. And so I haven't heard of anybody giving this advice of like very specific use of sub-agents, which, as you said, Is just imagine like there's a chat that's chatting with other chats that have some but not all the context and their own point of view. Maybe that's how you think about subagents for folks that aren't super technical, and then aggregate that back up and give you some feedback. I think is a really useful just hack that everybody should have for their work. So if you're a product manager, build it and like put your boss in there, your engineering partner, and your customer, and say every PRD. Review from the, these 3 points of view and give me feedback. If you're in marketing, again, like your ICP, give feedback. I think there's just a lot of places where this sort of multi-view feedback mechanism is, is super useful.

100%. We could take this a step further and say, hey, create this sub-advisory skill. And each, maybe 3 of the agents, like one agent is literally your boss that you send an agent to go research who your boss is, their role, their perspective, and like really define like the people that you work with and have like them simulate the feedback that they might give you before you even go to them. Like, the sky's the limit here, and you just got to tell Claude what to do, and it's going to go do it for you. So there's a lot of tools that you have at your disposal now, uh, to, you know, automate a lot of this. Like, one example with this is I also run a newsletter, and so the skill that we have for the newsletter is almost like a 10-step skill of like, before we even start writing this newsletter, I want you to interview me. what this newsletter should be written about. Then once you're interviewing me, I also want you to run a sub-agent to do internet research on what is good in the AI world this week. What should we talk about in this? And once my interview is complete and once the research is complete, we now have an idea of like the content that we can do. And then it's going to start working through a series of sub-skills, which are like, what should our subject line be? And what are really good subject lines? And it's going to have one set of instructions for that. And then my newsletter has like very specific body segments, and each body segment has a specific skill within it saying like, here's what makes a really good AI native section or what makes a really good AI future section, right? And at the end of all those steps, it has an evaluation part of the skill where it's going to go through a series of checklists and also launch that sub-advisory board. And so it's going to go through all of this, and all I need to do is say, UltraThink Skill, and then it's going to start this series that it works through and just helps me. And by the end of it, it's done so much work for me, and I've just had to answer some questions, and then it's outputted a draft that I think is really good. And I also feed it my past results of my newsletter of like, hey, we've written this together, this performed okay, and I give it the results. And it has a series of feedback and like, you know, review of like how these things are performing to see what's working and what's not. Which is a very important step because like, if you don't tell AI what success is to you, AI doesn't know what success is for you. So I always love including good examples and bad examples. This is very clear for like writing. If you're writing and you really like something, include it in a good example. But if it comes back with something that you don't like, put it in a bad example of like, hey, don't do that again. That doesn't look good. And this way you're showing Claude what is good for you and what is bad for you, because that's subjective and different for everyone. And Claude needs to know it. All right. So we have this multi-agent perspective here, and we can see that we've done a couple of things. This one, the thinking partner skill needs my review. So I'm going to go in there and I'm just going to review it. And it's looking for some information here, and it looks like it had opened up. So they have this new interface here, which allows you to review your thinking skill or whatever skill you created and see like, and like test it in real time. To see how it would come back and evaluate it like that, right? So that's very helpful if you want to use that. But what I want to do is I want to go back to my project here, and I want to create something called a scheduled task with you. A scheduled task allows you to create something that runs on a frequency. So it could run every hour, day, weekdays, weekly, whatever. And this scheduled task is going to be a morning debrief. I want you every single morning to look at my email, my Slack, and my calendar and check it to help organize a plan of action for the day. So that way, when I get into the office, I am aware of any kind of messages that need my attention. And you can start preparing me for my day by looking through my calendar and looking through my events. And if I need any kind of meeting prep, helping me do that. But generally, the output is going to be like a plan for the day. helping me get started on the best foot every single day. So that's going to be the prompt, and I'm going to run that daily, and I want it to run at like 7:30 AM. And I could choose what model I want this to run with or where these outputs should go. In this sense, it's going to be, you know, the daily operations. That's going to be good. And I'm going to just be able to save that.

It needs a name.

Ah, there it is. Thank you. I was like, why is it morning?

I wanna call out for folks, we've seen a couple of these like morning debrief, uh, examples on How I AI. We've seen, um, Hillary has shown us she uses Obsidian hooked up to her Claude code, hooked up to her calendar. She writes plan my day. We've seen a couple of these scheduled sort of like news I need to know, uh, tasks in ChatGPT. I think what I want to call out for folks, especially those who are again new to using AI beyond asking a question in ChatGPT, which I know honestly many of you are, which is I've used ChatGPT or Claude to ask a one-off question, get some info, you know, like write a song for my grandchild or answer a question about my medical health. Or write a document to my HOA, whatever. We, you know, people have used that use case. But what I think you're showing here is something that wasn't accessible to folks, which is you can actually build a little software-powered system for yourself to make your day go easier. And it could be writing emails for you. It could be, you know, every morning send you a message about what your schedule looks like. Let's say you're not even— you don't even have a busy schedule. Like I'm thinking about you know, my dad's retired, like news that you're— you might be interested in for the day. I have a daily, you know, in the technical world we call it a cron. And when you're on the— when you're on the OpenClaw side, but I have a daily schedule and in the morning I get pushed something kind of fun and interesting in the AI space. And at night I get pushed an article that will like make me feel good, you know? And so you just think about If the internet could come to me, as opposed to me having to go to the internet, what would I want the internet to come to me with? And that's what I think you focus your morning deep brief on. So it could be like highly functional from a productivity perspective. It could be, you know, go fill your water glass, don't drink coffee on an empty stomach, go for a walk and say hello to the sun. It can be any of those things, but it can be scheduled and it can be pushed to you.

Right. And right now, since we've scheduled this within a project, it has all the context of your project as well, meaning the skills, your connectors, and everything. And that's why having it scheduled within a project, like building up that context, it gets very powerful over time and you're going to get better results. So this could be a project debrief where like if you're a project manager, every single day it's going to look at this project and just help write a plan of action for all of your team members. To— and send as a Slack message. So like once they enter, that project plan for the day is done and very specific to each project, right? So there's just so many ways you can use it. And this also connects to all of my tools. So it's going to read my email and my calendar and my Slack and just prepare a daily plan with all that context in mind in a safe way. That's like the unlock for me is like just being able to read all these tools and use an AI on top of that to like help me do my job better.

Yeah, and I'll just call out for folks, 'cause again, I think this is sort of like a Claude Cowork 101, folks that are maybe newer to using AI in this particular way. And the things that you're gonna hear is like, do I want AI reading my email? Do I want AI reading, you know, my calendar? And the recommendation that I give to everybody is, again, just think about this as progressive trust, which is at first you just may want it to draft emails for you. Then you may want it to read your emails. Then you may want it to read your calendar or your Slack or any of those things. Then you may want to have like a shadow JJ that actually just operates completely independent of you and you're in the Bahamas hanging out and, you know, Agent JJ is actually doing your job for you. Just think about this on like a trust spectrum. You'll get, you know, you'll get to Clairvaux level at some point where you're just like floating, floating through the tokens. approving API access. But I do think, you know, people get nervous about this. It's a trade of information for productivity. And it's like we're all on one side or the other, which is you give the AI a little bit of information, it can help you be very, very productive. And you've got to figure out where you feel comfortable on that spectrum in your business, in your personal life. I think you and I are just showing if you get to this far further edge, You're going to get a lot of value out of it that you couldn't before.

Yeah. And especially with the skills, like I know everyone says skills, skills, skills, but like I have a skill for everything. I have a skill for writing on Twitter. I have a skill for writing on LinkedIn. I have a skill for doing my YouTube, my podcast, my newsletter. Like just having that built-in infrastructure that whenever I need to do any of that stuff, it knows exactly the way I like to work. That is a very big unlock for me. Another thing that I've been doing is I'm working with Anthropic through my company 10x right now. And one of the things that we're doing is we're hosting a workshop for Anthropic. And so I threw in all of the documents that Anthropic gave us of like, we're a program guidelines and workshop guidelines and everything. And so anytime I ask a question about this project, it has all of that context in mind. It knows exactly what Anthropic wants it to do. It wants all that kind of stuff, and it's very clear and easy. And so whenever I start a new project, The first step is like, all right, I'm going to create a folder. I'm going to put all my files in that folder. I'm going to open up that folder in Claude, call work, create a project for it. And now that project can access all that information and be very specific about helping me do better with that context in mind. And it's just a big unlock for me. And it's a very easy way to get started doing more with AI that will teach you a lot of skills that you can transfer over to actually building stuff with AI, which is Claude Code. which is also incredible. And the rate at which Cowork is improving is really, really awesome. And so it's just one of those things where it's like, if you're ready to do more with AI, which hopefully you are because that's where we're going, and those that can are rising up really quickly. So hopefully you are. And if you are, Cowork is a really good next step for you to blossom a little bit, spread your wings, and just see what you can do with AI doing stuff with you.

Well, and again, I wanna go back to folks that are like, okay, we got these 2 like podcast hosts telling us how they're planning their like YouTube release is using Cowork. I wanna say that again, these projects do not have to be work-related, right?

Sure.

They don't.

Right.

You can make a project that's house maintenance.

Right.

Let me tell you a project that you can make, house maintenance. Every couple months you need to change your air filters.

Right.

Every, you know, like, you know, in the winter you wanna make sure to plan for XYZ. You've got a big kind of like remodel project going on. Put that in a Claude Cowork project.

Yeah.

Generate the things that you need to generate. Keep it in, keep it in a workspace. And so you can be really creative about what the idea of a project means. I've seen other people who are doing hiring and they're saying, here's my recruiting engineers project, and it's all the work they need to do around job descriptions, sourcing, onboarding, interviews, all in one project. So again, it doesn't have to be personal productivity. It doesn't have to be workshops with Anthropic. It doesn't have to be your YouTube. And I love— I'm glad you're showing this screen, which is Cowork has this list of ideas of kind of things you can do with it. I think like, good job, um, Cowork product team. Because again, people just— this is part of the point of this podcast, which is people hear that AI is awesome, they want to use it, they're ready to go, and then they sit in front of a chat box and they're like, what? Do I do with this? Like, what can I possibly do?

And is AI gonna steal all of my data?

And is AI gonna steal? Those are the 2 questions. Like, is it gonna steal everything and run off with my spouse? And like, what do I actually do? And so I think like pod, you know, this podcast is a resource. We actually put, we're now at almost 200 individual workflows documented out of this podcast.

Wow.

Um, on the ChatPRD How I AI blog where we've just like given you step by step how to copy this stuff. We'll do this episode as well. But I think the point is just like find something you're working on and give it a go. JJ, I know that we got to get you back to your Cowork projects and your podcast, your Cloud Coding and 10x and all that stuff. So let's just do really quick lightning round, super, super fast answers. One, outside of what we just saw, what is your number one favorite Cowork use case that you've discovered?

Content creation is huge. Doing this Anthropic workshop right now, like it's so intense. But so this project has been really helping me because they gave us like 30-page document of like what to do and what not to do. And at any time I could just go into like, what do they say here? What do they do here? And it literally helps me. So navigating a lot of files with Cowork is just so helpful.

Amazing. I've heard a lot of event planning with Cowork and a lot of event planning with Codex, which is OpenAI's kind of version of this. Very good. I'm waiting. I call— how I request, if you are planning your wedding, With Cowork. I want to hear that.

Great use case. Great use case. It can go find the vendors for you. It can write the emails for you. It could— oh my God, great use case.

I know, I know. This is gonna be my next business. I'm gonna do AI-driven wedding planning. Okay, my second question to you, outside of Cowork, what, what's, what's a tool that you're loving?

I do Remotion a lot. I'm doing programmatic video creation for like so much of my stuff right now. That's really working well. I really love pencil.dev too. It's like the cursor but for design. And so you tell it what you want and it'll design for you. So those are some of my 2 favorites that I've been digging into outside of Anthropic, uh, Claude code stuff.

Yeah, man. Okay, if I haven't done a Remotion video by the time this comes out, yell at me in the comments. I will do it. People have been asking this, um, from me for months and months and months. I love Remotion too. Again, uh, programmatic video generation, super, super cool. And you can do it with, with, with Claude Code. And then my last question, which is you're— this is maybe tricky because you seem to be a Whisper Flow sort of talk to the AI with your human voice, which I think keeps us a little bit more polite. But when AI is not listening, what do you do? What's your prompting strategy?

I'm generally polite because I, I, I'm generally polite, but I will, I will put on the caps locks and I will let it know sometimes when I'm talking, when I mean business. I don't normally swear at it though. Usually when it's not behaving, it just doesn't know what I'm asking. And that's a good chance to better communicate of like, here's what I'm asking, here's what a successful output gets me, help me figure that out. Ask me any questions to help us get back on track. Launch a sub-agent to figure out what's going on here. That kind of stuff helps me navigate it back to being like, on track. And then also saying like, hey, what we just experienced, that was not good. Add that like scenario to like my negative outputs so we don't do that again.

I, I love it. I found myself— I don't yell, I do sometimes all caps. I found myself being like, this is truly insane. What you just did is cuckoo. Fix it. Um, not, not super polite, but, but effective. Okay, JJ, where can we find you and how can we be helpful?

I think I'm pretty much everywhere. I run a YouTube channel called 10X Builder. I'm on X for @JJEnglert, LinkedIn @JJEnglert as well. I work at 10X. I lead our community enablement here. We're doing a ton of stuff. So if you need help bringing AI transformation and enablement to your org, let us know. But otherwise, it is a pleasure being here. We love your work. I love your work. Thank you for inviting me. It's been a joy.

Thanks for joining us. Thanks so much for watching. If you enjoyed this show, please like and subscribe here on YouTube, or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time.