Your Claude Code Agentic OS Sucks
CChase AI
Computing/SoftwareAdvertising/MarketingSmall Business/StartupsInternet Technology
Transcript
00:00:00Your cloud code agentic OS sucks and it's because you're focused on the wrong
00:00:05things.
00:00:05You're spending all your time on fancy dashboards and command centers like this
00:00:09one and this one instead of focusing on what actually drives value in a cloud
00:00:14code agentic OS.
00:00:15And that's this a skill and automation backbone that actually drives everything.
00:00:20The problem is creating something like this at a high level takes time,
00:00:25isn't flashy and can be kind of boring,
00:00:28especially when we compare it to these wild looking command centers that bring
00:00:33in a ton of views. But the truth is to get any value out of a cloud code agentic
00:00:37OS, especially when we're talking about the observability piece,
00:00:40the dashboard piece, the command center thing.
00:00:42It's only going to happen if this is locked in and that's because a strong
00:00:48agentic OS has three parts to it. The first is what you see right here.
00:00:52It's the skill and automation backbone.
00:00:54It's the idea that we are going to take cloud code and turn it into a system that
00:00:58can give us reliable outputs.
00:00:59We are going to take your daily or your teams or your clients workflows and tasks
00:01:05turn those into skills, turn those skills into automations where it makes sense.
00:01:09And in the process build out a cohesive system like you see here.
00:01:14So we can do the same thing over and over again at a high level and get
00:01:19consistent outputs. The second part of an agentic OS is the memory layer.
00:01:23How do we handle the idea of context engineering? Well,
00:01:27there's a number of ways we can do it.
00:01:28We can do something super fancy with full blown knowledge graphs and do something
00:01:32like light rag or we can keep it simple and just use something like obsidian,
00:01:36which is a 80% solution that's more than enough for the vast majority of people.
00:01:40And it's only once we've locked all that in does any sort of dashboard or command
00:01:45center for an OS make sense because the value of a dashboard really comes in two
00:01:51parts. First is the observability side.
00:01:53That's the idea that I can kind of cover some of the weaknesses of being in a
00:01:57terminal. Things like seeing my metrics for my social media channel,
00:02:00being able to quickly dive into different audience metrics,
00:02:03have all my research shown to me on one tab.
00:02:06The second half of that value comes from here, all these sort of buttons.
00:02:10And that's the idea that if I want to bring the power of cloud code to a team
00:02:14member or to a client who's never going to jump into the terminal,
00:02:17I can instead build out that skill architecture for them, assign it to these buttons,
00:02:22and they can essentially just execute them on command by just clicking them.
00:02:26And so today I'm going to show you how to properly set up this skill backbone.
00:02:30And then we're going to talk about the dashboard side of it because there is a lot
00:02:35you actually can do in this scenario. And there's really two paths.
00:02:37You can go down like you've been seeing. I've kind of been showing you two versions.
00:02:40There's the one you see here, which is a literally a part of obsidian itself,
00:02:44which is pretty cool because we also get an integrated terminal and there's this
00:02:47web app version, which is really built for distribution.
00:02:50If you're someone who's trying to bring in other team members or packages for
00:02:53clients, but before we jump into the nitty gritty of how to do it,
00:02:56a quick word from today's sponsor me. So as you know, inside of chase AI,
00:03:01plus I just released the Claude code masterclass,
00:03:03which is the number one way to go from zero to AI dev.
00:03:06But I have also just added an agentic OS masterclass inside as well.
00:03:11So everything you see in today's video, the prompts, the dashboards, the setups,
00:03:15all that can be found at a much deeper level inside of chase AI.
00:03:19Plus there's a link to that in the pin comment. Also today,
00:03:23I guess when this video comes out,
00:03:24I'll be running a free webinar of how to set up an agentic OS
00:03:28for yourself going through all three layers. So if you want to join,
00:03:32make sure to check out the pin comment as well. I'll have a link for both of those.
00:03:35So if this is where all the value lies, how do we set this up?
00:03:38And why is it set up like this? Why does it look like an org chart? Well,
00:03:42the whole org chart set up, like you see here, where we have stuff broken out,
00:03:46into different sections like productivity and research and content.
00:03:49This is just to help you visualize something that is ultimately invisible.
00:03:53This is just for your mental model.
00:03:54And it's the idea that you do a bunch of different things across a bunch of
00:03:58different domains in your day-to-day week to week flows and whether it's in your
00:04:01business or just in your personal life. For me,
00:04:04that is split up amongst things like my productivity. So things like Google,
00:04:09research, content, my community, my agency, my sales, on and on and on.
00:04:13And what we need to do for you is we need to take the giant
00:04:18morass of things you do in a day to day, right?
00:04:21All these different tasks and we need to break them out and we need to turn them
00:04:26into skills. Why do we need to turn them into skills? Well,
00:04:30chances are the way you work right now with cloud code,
00:04:32when you need it to do something,
00:04:34you just spin up cloud code in the terminal and you kind of tell it what to do.
00:04:37You're pretty much just using it as a slightly better chat GPT.
00:04:41And if you're doing this all the time,
00:04:44why are we not codifying this into a skill?
00:04:47Because when we codify it into a skill, there's a few things that gives us one.
00:04:51It's convenient. I'm taking that entire task.
00:04:54And instead of talking about it over the course of a paragraph,
00:04:56I just tell it to do skill, whatever it could be a single word and it does it.
00:05:00So the convenience is one piece. The second piece is that since we have codified it,
00:05:05we can also test it using something like the skill creator skill.
00:05:09We are able to actually benchmarks the benchmark, the skills we create.
00:05:14So we can see if a,
00:05:16does the skill even make sense because it will AB test it against us using the
00:05:20skill versus not having the skill at all. And over time, if this skill is good,
00:05:25we're going to start getting more deterministic outputs from a system that is
00:05:30inherently non-deterministic. Like when we talk about LLMs,
00:05:33there's a certain randomness to it, just inherent to how it works.
00:05:38Anytime we can make things less random, the better.
00:05:42And by codifying these things you do day to day and turning them into skill,
00:05:45that's one giant step forward in doing so.
00:05:47And while that makes sense to a lot of people, if you were to ask them,
00:05:50if they'd actually ever sat in front of their terminal, turn their mic on,
00:05:54opened up Claude and said, Hey, here's my daily plan. Here's what I do.
00:05:59Can you pull some skills out of that and then turn them into skills using the
00:06:04skill creator skill,
00:06:05you could probably count the percent like on one hand,
00:06:09which is wild because this is one of the easiest yet most powerful upgrades to how
00:06:14you use cloud code.
00:06:15And this visualization is kind of just there to help you think about it because
00:06:19we do a bunch of different things in a bunch of different domains.
00:06:22And oftentimes we can even combine a lot of the tasks we do into
00:06:28quote unquote like workflow skills or higher order skills that have it do a bunch
00:06:32of different things at once. For example,
00:06:33I have a skill called the content cascade skill.
00:06:37This skill for all intents and purposes is a content repurpose or when I create a
00:06:42YouTube video and I call on the content cascade skill does a number of things for
00:06:46me. It downloads the transcript. It creates a blog post.
00:06:50It creates a LinkedIn post. It creates a Twitter post. It spins up Playwright.
00:06:54It then posts those things for me.
00:06:57That's a bunch of different individual tasks all in one,
00:07:00but instead of breaking out into nine different skills, well,
00:07:03now it's just one skill.
00:07:04And that's something that can be a huge like productivity boost.
00:07:09But if you've done that with all the different things you do in your day to day,
00:07:12probably not.
00:07:13And it's this process of sort of walking through what you do step by step and
00:07:18codifying it. That's the power of an agentic OS.
00:07:21Everything we do outside of this, the memory layer, the dashboard,
00:07:24it's kind of just a nice little bow around it.
00:07:27And if you're someone who's not trying to work with team members,
00:07:30someone who's not trying to package these things and sell it,
00:07:32you could probably stop here and you're like, you know,
00:07:35the 80% solution and you're way ahead of the pack.
00:07:38And so to actually execute this process is pretty simple at its
00:07:43core. You're just going to do what I said, open the terminal,
00:07:47start a new session and just start talking. And at the end say, Hey,
00:07:51can we turn this into any sort of skills? Now,
00:07:54I have an entire prompt that breaks this down at a very detailed level of how to
00:07:58do this skill triage, but at its core, that's all we're doing.
00:08:01Here's what I do. Turn it into skills. Sweet. Okay. Let's test the skills.
00:08:06Let's move on to the next domain in my business, in my team. And the thing is,
00:08:10this is going to be extremely customized and specific to you.
00:08:15I think we get kind of lost in the morass of like the 10 billion skills that are
00:08:19floating around. We go to these mega repos,
00:08:21like awesome Claude skills. And when you look through 10 million different skills,
00:08:25thinking this is what's going to change, you know,
00:08:27my day-to-day outcomes with Claude code.
00:08:31And it's like you're kind of looking for a diamond in the rough here when instead
00:08:34knowing that one of the most powerful parts of Claude code is how easy it is to
00:08:38customize it for you. Like,
00:08:39why aren't we leaning into that more in a systemized way,
00:08:43but outside of the custom stuff,
00:08:44I think there's a few things that almost everyone can get some value out of.
00:08:48I think on the productivity side, a big one is if you're in the Google ecosystem,
00:08:53I've kind of talked about it before using things like the GWS CLI to
00:08:58basically allow you to do anything inside of the Google ecosystem and turning
00:09:01those into skills, whether that's like email, triage, Google drive, work,
00:09:05or stuff on the calendar.
00:09:06But the truth is you can also just use the standard MCP connectors that come with
00:09:11Claude code. And I'm just talking about the basic Claude dot AI Gmail,
00:09:15Google calendar and drive.
00:09:17The only things you're really losing there is you're not going to be able to send
00:09:20emails, but you can still do drafts, which for a lot of people is good enough.
00:09:24Since they don't want it to actually send them off.
00:09:27And that takes 30 seconds to do. And like, it's such a productivity boost that,
00:09:30again, very few people actually do. Now,
00:09:33after you've gone through this skill creation process,
00:09:36next becomes the decision tree. When it comes to automations for each skill,
00:09:39it doesn't need to be on demand or is it something we can turn into routine inside
00:09:43of Claude code. Nevermind when we talk about routines and automations with cloud
00:09:47code, it's broken down into two different parts.
00:09:49That's going to be local automations versus automations running in the
00:09:55cloud. If you don't know which is, which just stick with local.
00:09:59That basically means it's going to run when your computer's on any of some
00:10:02version of caught up on the cloud.
00:10:04That means it's going to be run on anthropic servers and you're going to be
00:10:07limited to how many you can do because they're basically paying for it.
00:10:10And if you're on the cloud, Hey,
00:10:11it doesn't have access to your actual computer. It's not running on your computer.
00:10:15It doesn't have your CLIs or skills, your files.
00:10:17So most of the time it's just going to be a local automation if you're in doubt.
00:10:22And this is the process by which you create the backbone for a Claude code agentic
00:10:26OS. And I keep saying Claude code. The truth is Claude code is just the engine.
00:10:30And we'll talk about this a little bit, a little bit more.
00:10:32You could replace this with codex. You could replace this with really anything.
00:10:36You know, we're building the chassis for this.
00:10:39We can swap out the engine at any time.
00:10:42So everything I say here also applies to something like codex.
00:10:44Now let's talk about obsidian in memory very quickly before we dive into the
00:10:48command center observability dashboard piece,
00:10:50because I think a lot of people get confused about what obsidian is actually
00:10:54buying you and the point of it all.
00:10:55Remember the point of obsidian is simply an organization layer.
00:10:59Obsidian isn't doing anything special to all these markdown files.
00:11:04It's simply giving us the human being a way to kind of figure out what the heck is
00:11:09going on in our files and gives us a simple way of sort of connecting them.
00:11:13It isn't inherently changing the memory. This isn't rag.
00:11:17It's not embedding anything. There's no like vector database,
00:11:21despite, you know, these like cool graphics,
00:11:24like this isn't a true knowledge graph in that sense. That being said,
00:11:28being organized,
00:11:29especially when we talk about being organized at scale with thousands and thousands
00:11:32of documents is very important. And it's not important just to you,
00:11:36being able to figure out where stuff is.
00:11:37It eventually becomes important to Claude code at a certain scale in terms of
00:11:40token efficiency, refining things. That's why everyone brings up this, right?
00:11:45The Carpathi rag name, go through it very quickly.
00:11:47It's just the idea that we have a vault,
00:11:49which is where obsidian lives and some series of sub folders. Carpathi says, Hey,
00:11:53we have raw for like unstructured data. We have wikis, which kind of break the,
00:11:58take the unstructured data and turns it into like reports articles.
00:12:02And then we have outputs for like deliverables. So Hey,
00:12:05I did some research on AI agents, which, which went to raw.
00:12:09That research got turned into an article about AI agents in my AI agent wiki.
00:12:13Hey, I turned that into a slide deck. That's sort of the idea.
00:12:16The truth is you don't have to do that at all.
00:12:19All you need to do is you need to figure out something that makes sense to you.
00:12:24And it needs to be created in a way that you and Claude code could snake your way
00:12:29through the folder system. If there was a hundred thousand files in there,
00:12:33a base line like this is a good start, especially because there are things called
00:12:37master index files and index files all over the place.
00:12:40These index files are essentially at every level of obsidian.
00:12:45And remember obsidian is just a folder.
00:12:47So we're talking about every sub folder we go down.
00:12:49There's some sort of folder that's acting like a table of contents.
00:12:52So if I'm in the vault and I click on the wiki folder inside,
00:12:57the wiki folder is a table of context called an index file, which tells me, Oh,
00:13:02inside here we have agents rag systems and content creation wikis.
00:13:06Cool. I know where to go. I go inside the AI agent folder. What's inside there.
00:13:11There's another index. There is another table of contents saying, Hey,
00:13:16inside the AI agents folder,
00:13:18we have this document and this document that's the biggest thing I would take out
00:13:23of Carpathi is the idea of indexes and indices and the idea that for every layer
00:13:27I go down in obsidian and my file structure,
00:13:30there's some sort of master document that points me in the right direction.
00:13:33If you don't have that in the beginning,
00:13:34have fun figuring that out when you're 5,000 documents deep. For me,
00:13:38in my scenario, I have several folders. I have an archive content notes,
00:13:42dashboard, inbox, ops, project systems, wiki makes sense for me.
00:13:47I have an index. I understand what's going on.
00:13:49You like all these things need to customize it. So it makes sense for you.
00:13:53And speaking of customizations, now let's go into the dashboard piece.
00:13:57These command centers for these agentic operating systems.
00:14:01We talked a little bit already about the value play there, right?
00:14:03It's the idea that there's visibility and I can actually see things that I
00:14:07couldn't see in the terminal.
00:14:08And we have sort of like these skill panels that anyone could use.
00:14:11The next question becomes why the heck are there two of them?
00:14:14Why do you have this one inside of obsidian itself?
00:14:17Cause I'm inside of obsidian here.
00:14:19And why do you have this one as a streamlit app on a local host?
00:14:22That's essentially a web app. What's the difference between these two,
00:14:25which makes sense for what, well, I think the value play for the streamlit
00:14:28applications are really any sort of web app.
00:14:31That's your dashboard layer for agenda. Go ask is for distribution.
00:14:35If I want to bring this to a team or really if I want to package this for
00:14:38clients, having it set up like this is super easy.
00:14:41I can have the template inside of a GitHub and I can clearly or very
00:14:46quickly distribute that to anyone anywhere.
00:14:48Setting this up takes literally seconds.
00:14:50And if this is meant for a non-technical team member or a non-technical client,
00:14:54keeping it as simple as possible like this and just having clear buttons that are
00:14:57mapped to skills and it executes them. That's great. That's all they want.
00:15:01The obsidian for dashboard is a little bit different.
00:15:04You're trading distribution for really ergonomics at this point.
00:15:08And I would argue a little bit more power because it's super easy.
00:15:11As you can see here to also have an integrated terminal inside
00:15:16of your obsidian command center,
00:15:19which basically means I now have the best of both worlds,
00:15:22not to mention because it's inside of obsidian all my stuff is right here for me
00:15:26to play around with. And obsidian is infinitely customizable like over here,
00:15:30right? You know, I have my full calendar, but this isn't like a calendar plugin.
00:15:34This is literally me just having the Google calendar webpage
00:15:38open and put here on the right hand side on the overview of a very clear idea of
00:15:43what's going on that day, what my tasks are,
00:15:45what's going on with the activity feed and like where I'm at across different
00:15:48communities. I want to dive deeper into audience stuff.
00:15:51I have a tab for that. I want to dive deeper into research.
00:15:54I have a tab for that that shows like trending, GitHub repo stuff going on,
00:15:58hacker news, as well as some of my briefs, which are also tied to skills,
00:16:02things like headlines, things going on X and YouTube and like content opportunities.
00:16:06Again, having this,
00:16:08if I'm in a pure terminal setup is just a little bit clunky.
00:16:12It's a little more difficult. The problem though,
00:16:14with the obsidian setup and I kind of alluded to it is the idea of distribution.
00:16:18How could I distribute something like this to a team or to a client?
00:16:23You can kind of do it because this whole dashboard command center is essentially
00:16:28just a custom plugin that Claude code created, but it's a little more, again,
00:16:32clunky and awkward to set this up for somebody else. It's not just like, Oh,
00:16:37clone it. You're good to go. It's like, okay, clone it. Now go into obsidian.
00:16:41Now enable these plugins. Now move this here, move this there,
00:16:44do all this stuff. So there's a certain awkwardness to it.
00:16:48So if you're someone who's like a solo operator and you're just like, Hey,
00:16:52I want an agentic OS with Claude code.
00:16:54I want all these cool customizable buttons, whatever they may be.
00:16:58And I also want the terminal like clearly available all on the same pane.
00:17:02The obsidian Ford route is perfect. If on the other hand, you're someone who's like,
00:17:07I'm just trying to package this for teams and clients and turn this into actual
00:17:10product. The web app is the way to go,
00:17:12but understand these systems are only as powerful as the skill architecture it's
00:17:16built upon. It's just a nice layer on top of Claude code,
00:17:19because if you don't have that,
00:17:21this is just some fancy nonsense. That's all it is, right?
00:17:26You need some actual meat to this. So don't forget where you make your money.
00:17:30So I'm going to wrap it up there.
00:17:31I hope I was able to make it a little bit clear as to where I think the value in
00:17:36these agentic OS systems are at.
00:17:37I see a certain contingent of people who really rail on these and say they're
00:17:41worthless. I don't think that's a fair assessment at all. When they do,
00:17:45it's usually kind of purely targeted on the dashboard side of it,
00:17:48which makes sense if you're arguing against the dashboard or the command center
00:17:52in the vacuum, but that's not real. The power really is right.
00:17:56The dashboard and all this is somewhat of a facade,
00:17:59like what's going on is behind it. And that's where it's sort of the focus.
00:18:02I think should be. And if we focus on that and the idea of skills and everything,
00:18:06it's like,
00:18:07are we then arguing that you shouldn't have a system of skills that are codified
00:18:11that are based on what you do in your day-to-day life?
00:18:13I think you have a hard time arguing against that. Oh, one last thing,
00:18:17something other people brought up the idea of costs, which is an important one,
00:18:20especially if you've been paying attention lately.
00:18:22And the idea that the dash P command doing headless,
00:18:26Claude code runs is something that apparently anthropic doesn't like anymore.
00:18:31And by doesn't like, I mean,
00:18:31they're throwing you $200 to use exclusively on that, but it's on API costs.
00:18:35Is there an issue with that in this whole setup? Because as you can imagine,
00:18:40all of this is running headless, Claude code under the hood. Yes and no.
00:18:45For 200 bucks a month, you would have to be kind of like spamming these.
00:18:49They get to that point. And so I think in reality,
00:18:55it's probably not going to be issue if it was an issue and you felt like you were
00:18:59hitting usage issues or clients were hitting usage issues.
00:19:01I think the simple solution is you just move this all over to something like
00:19:04codec CLI because codex is great and they don't have these issues as well.
00:19:09And you get more, you get more bang for your buck.
00:19:12And switching everything under the hood here for, for codex, very simple.
00:19:16I mean, you could use college code to do it.
00:19:18You would just point it at the code and just be like, all right,
00:19:21we'll switch it. So now it calls the codec CLI instead of Claude.
00:19:26So this is something you could essentially like refactor and a matter of minutes.
00:19:30And you can even put like a button on the dashboard, which I might do.
00:19:33It's like, all right, let's go to the codex version.
00:19:35So just something to be aware of in reality, I think for 99.99% of people,
00:19:40it has no effect. So that's what I'm going to leave you again,
00:19:43everything you saw here,
00:19:45if you want the actual like my exact setup for the subsidy and command center
00:19:50and everything else, you can find that inside of chase AI plus
00:19:53and make sure to check out the webinar that's going on, you know,
00:19:57and I don't know 20 hours from this video being posted.
00:20:01So besides that, I'll see you around.