Latest / In the Loop AI / OpenClaw Overview: Which AI Agent Should You Use?

Transcript

OpenClaw Overview: Which AI Agent Should You Use? · In the Loop AI

0:00
Josh: You probably have six or seven apps on your phone that each do one thing.
0:03
Josh: One controls the lights, a thermostat, your security cameras.
0:06
Josh: Each has its own login, its own interface, and its own notifications.
0:09
Josh: It's horribly annoying. Now, imagine replacing that all with a single WhatsApp
0:13
Josh: thread or a single text message through your phone.
0:15
Josh: That is CLAW. And it's not just for home automation. People are running CLAWs
0:19
Josh: that monitor code bases and open pull requests when they sleep.
0:22
Josh: They even draft these investor reports by pulling live market data.
0:25
Josh: It can do anything that you can do, but digitally on your computer.
0:28
Josh: The word claw describes something different from the AI that most people know.
0:31
Josh: A lot of people are used to chatbots, which you type into, and it forgets everything
0:35
Josh: about you by the next session.
0:36
Josh: And then there's agents that are capable of actually going on the web and doing
0:39
Josh: things for you, acting on your behalf.
0:41
Josh: But then there's claws, things that run in the background that can orchestrate
0:44
Josh: complicated tasks that have persistent memory that never forget.
0:47
Josh: And that, that is what every major AI company in the world is building right now.
0:53
Josh: Anthropic, OpenAI, Google, Perplexity, everybody is building a claw.
0:57
Josh: And this episode, we're going to walk through all the options that you have
0:59
Josh: and which one is overall best for you.
1:01
Ejaaz: Yeah, the AI stack has changed quite aggressively over the last couple of months.
1:05
Ejaaz: If we think of like the LLM as like the brain, then claws are kind of like the
1:09
Ejaaz: arms and the limbs that allow you to actually do things.
1:12
Ejaaz: Now, everyone's familiar with ChatGBT and clawed. It's kind of like you ask
1:16
Ejaaz: it something and it gives you an answer in text, but it doesn't actually do anything for you.
1:20
Ejaaz: And then we had the progression to AI agents, which is basically strapping an
1:23
Ejaaz: LLM with a bunch of different tools, your email, Slack, whatever it might be.
1:27
Ejaaz: But it was still super clunky.
1:29
Ejaaz: It could like use the tool, but it didn't know when to use the tool.
1:33
Ejaaz: And that's when OpenClaw and Claw started becoming really popular because it
1:36
Ejaaz: tied in persistent memory and context about you with these different tools to
1:42
Ejaaz: create a personal agent that actually worked. And that's why it went super viral.
1:46
Ejaaz: Remember, this was a vibe-coded project, OpenClaw, formerly known as ClawedBot,
1:51
Ejaaz: over the weekend by Pete Stuy.
1:53
Ejaaz: And it went completely viral, where it has now tens of millions of users building
1:58
Ejaaz: their own versions of personal agents every single day.
2:01
Ejaaz: So if ChatGPT was the brain, you
2:03
Ejaaz: now have OpenClaw or all these different claws, which do stuff for you.
2:07
Josh: So now that we're familiar with what a claw is, there are kind of two tiers of these claws.
2:11
Josh: One being plug and play. these are companies like anthropic open ai meta through
2:15
Josh: their madness thing and perplexity these are the ones where you probably have
2:18
Josh: an account it's twenty dollars a month they give you access to all the tools
2:20
Josh: and then there is the fully custom ones this is what you are probably most familiar with
2:25
Josh: being OpenClaw and all of the open source alternatives to the OpenClaw infrastructure.
2:29
Josh: That's kind of for the hardcore tinkerers or people who really want to build
2:32
Josh: custom stacks for their claws.
2:34
Josh: But we should be starting probably with the foundation labs,
2:38
Josh: the labs that everyone's familiar with, because chances are you already use
2:40
Josh: them and have an account open with them already.
2:42
Ejaaz: Yeah. So OpenAI of all the foundation labs is kind of what took OpenClaw being
2:48
Ejaaz: a niche project to something more mainstream.
2:51
Ejaaz: They technically didn't acquire Pete Stye and the OpenCore project.
2:55
Ejaaz: It's technically still open source and they created a foundation,
2:58
Ejaaz: but OpenAI basically owns OpenCore at this point. Pete Stye works.
3:02
Ejaaz: He's a full-time employee at OpenAI and he's building out their version of OpenCore for OpenAI.
3:08
Ejaaz: And so in Sam Altman's announcement post, he says he's joining OpenAI to drive
3:12
Ejaaz: the generation of personal agents.
3:13
Ejaaz: And what's interesting about Pete Stye's take on this is he still wants to build
3:18
Ejaaz: OpenCore in the vision that he has for it.
3:20
Ejaaz: But something interesting that Pete Stein mentioned on his podcast with Lex
3:24
Ejaaz: Friedman is one of the reasons why Sam Olman was super excited about acquiring
3:29
Ejaaz: OpenClaw Project as a whole and bringing it so close to ChatGPT is Sam realized a weakness in ChatGPT.
3:35
Ejaaz: No matter how many apps or tools he created, no matter how much of an app store
3:39
Ejaaz: he built on top of ChatGPT,
3:42
Ejaaz: developers didn't really want to build on OpenAI because it didn't give them
3:46
Ejaaz: the customization capability to be able to build the types of apps that they wanted to do.
3:51
Ejaaz: And in their fairness, they can't envision exactly what AI apps look like right
3:55
Ejaaz: now. This is the Wild West at the moment.
3:57
Ejaaz: So he realized that it was probably best to build the number one or bring the
4:01
Ejaaz: number one open source ecosystem for building customized personal AI agents
4:04
Ejaaz: into OpenAI. And that's what he did with Pete Starr.
4:07
Ejaaz: Yeah, so since the acquisition, OpenAI has built out a suite of different features,
4:11
Ejaaz: which create their own version of OpenClaw for ChatGPT.
4:16
Ejaaz: A few come to mind. So they have computer use, which allows ChatGPT to take
4:20
Ejaaz: over your browser or desktop itself and do a bunch of things.
4:24
Ejaaz: Now, it's important to know that OpenAI to date has acquired a lot of memory and context about you.
4:29
Ejaaz: And now they're matching that with a personal agent, which is exactly what OpenClaw
4:32
Ejaaz: is. So that's great to see.
4:34
Ejaaz: They also have an AI browser called Atlas, which allows you to do any of the
4:38
Ejaaz: browser extension type of stuff.
4:40
Ejaaz: So if you want to do research or if you want to browse your email or do anything
4:42
Ejaaz: that is in your online digital world, it can now also do that.
4:46
Ejaaz: And the final thing is, of course, their coding agent, Codex itself,
4:49
Ejaaz: which they integrate with the former two tools that I just mentioned,
4:53
Ejaaz: which allow them to do the building and creation of new things that you want.
4:57
Ejaaz: So you have this personal agent that sits within ChatGPT right now that can
5:00
Ejaaz: do so much more than just be a chatbot LLM.
5:03
Josh: Now, perhaps you don't have an OpenAI subscription and you have something like Anthropic on Claude.
5:08
Josh: Well, you're in luck. Anthropic has probably the most robust alternative to OpenClaw.
5:13
Josh: They didn't acquire OpenClaw, so therefore they were forced to build their own version.
5:16
Josh: And they've deployed quite a few features and utilities that really make it
5:21
Josh: a compelling offering for the average person who doesn't want to fully build
5:25
Josh: a custom Claw stack. And there's this whole laundry list of things,
5:28
Josh: starting with Claude Cowork.
5:30
Josh: Funny enough, it only launched, what, less than three months ago?
5:33
Josh: It hasn't even been around for that long.
5:34
Josh: Two months ago. But it has been such a huge game changer because Cowork works
5:38
Josh: with your entire computer.
5:40
Josh: It allows it to access files on your desktop, execute those files,
5:44
Josh: and actually make changes on your home computer so it can take over whatever
5:47
Josh: you give it access to in a secure way.
5:49
Josh: Just recently, yesterday, they released Routines, which are basically cron jobs
5:53
Josh: for those who don't know. They're scheduled tasks that you can do with your computer off.
5:56
Josh: You'll find a lot of people have been hyping up purchasing mac minis because
5:59
Josh: they allow your computer to sit on your desk and stay always
6:02
Josh: on well these new routines offer a cron
6:05
Josh: job feature for your claude instance without your
6:08
Josh: computer needing to be on all the time that's very cool they recently rolled
6:11
Josh: out computer use which lets claude actually see your screen move the mouse around
6:16
Josh: it can click type navigate apps and it's all permission gated so it's very secure
6:20
Josh: you kind of choose what you want it to have access to and then it will go off
6:24
Josh: and do whatever tasks that you ask it to do And then finally,
6:27
Josh: there's this little thing you may have heard of called Clawed Code,
6:30
Josh: which is probably the most popular coding developer engine that currently exists,
6:35
Josh: all baked into one super app.
6:36
Josh: So the Clawed Desktop app is a super app that is functioning as a hybrid claw
6:41
Josh: that allows you to schedule these tasks, manage agents, actually engage with
6:44
Josh: your computer and do so whether it's turned on or off. It's a really powerful piece of software.
6:48
Josh: And personally, this is what I find myself using the most. I love the Anthropic
6:52
Josh: stack that they've built, how there's connectors and plugins to a lot of the
6:55
Josh: tools that I use that are third party.
6:56
Josh: And I found it to be just very easy to set up, very robust and very secure in
7:01
Josh: the sense that it only accesses the things that I explicitly give it permission to.
7:04
Ejaaz: Yeah, it's really funny. Two months ago, Anthropic was the most hated company
7:09
Ejaaz: from the OpenClaw community for one simple reason.
7:12
Ejaaz: They started just banning people using Anthropic claw tools for OpenClaw.
7:18
Ejaaz: And no one quite knew why. People thought that they fumbled by like,
7:21
Ejaaz: you know, this was a missed PR opportunity. You could have gotten so many new
7:24
Ejaaz: subscriptions and users.
7:25
Ejaaz: But Anthropic wouldn't let loose why they were doing this thing.
7:28
Ejaaz: Fast forward two months, and they've released, I believe it's 30 plus features
7:33
Ejaaz: specifically to rebuild OpenClaw within a sandboxed and closed environment.
7:39
Ejaaz: And the reason is pretty fair, which is OpenClaw is kind of the wild west.
7:43
Ejaaz: There's a lot of security flaws.
7:44
Ejaaz: It's kind of expensive. It's kind of hard to use. We create a packaged,
7:49
Ejaaz: curated, cheap version all for 20 bucks a month for you to use.
7:53
Ejaaz: And you don't have to worry about, you know, losing all your photos or your
7:56
Ejaaz: hard drive information. You could just quietly and confidently use this product and it just works.
8:00
Ejaaz: And that's what they shipped over the last two months.
8:03
Ejaaz: That's kind of a little bit of a misdirection because really it's over the last
8:07
Ejaaz: three weeks, they shipped the bulk of those features, just insane amount of delivery.
8:11
Ejaaz: It wouldn't surprise me if they're using some kind of closed open core implementation
8:15
Ejaaz: that they built within themselves to actually build and launch these features.
8:18
Ejaaz: A few more things that are pretty cool about their version of OpenClaw,
8:22
Ejaaz: which OpenClaw itself does not have.
8:25
Ejaaz: You can currently shut your laptop that you're watching this video on right
8:30
Ejaaz: now and still operate their version of OpenClaw because they have these things
8:34
Ejaaz: called dispatch and routines that you mentioned earlier, which allows you to do this.
8:38
Ejaaz: So they've really built a comprehensive suite that has rebuilt OpenClaw in a
8:42
Ejaaz: much more meaningful way.
8:43
Ejaaz: And it's important because Anthropic currently has the context and data on millions
8:48
Ejaaz: and millions and millions of new retail users, which is key.
8:52
Ejaaz: Like when it comes down to it, if everyone's building an open call,
8:55
Ejaaz: the key differentiator ends up becoming context and data.
8:58
Ejaaz: And Anthropic is one of two major companies right now, the other one being OpenAI
9:01
Ejaaz: that has this advantage.
9:03
Josh: So if you've heard all that, and Claude isn't quite for you,
9:05
Josh: there's another competitor on the block that goes by the name of Meta.
9:08
Josh: And now Meta, they actually haven't built their claw harness themselves.
9:11
Josh: Instead, they acquired a company named Manus, if you remember,
9:14
Josh: for about $2 billion in December of last year.
9:17
Josh: And Manus is this harness that deploys fully autonomous agents to write code,
9:23
Josh: deploy apps, browse websites, do all the things that you would expect in a claw,
9:27
Josh: except doing so now with the meta operating system.
9:30
Josh: So if you'll remember, we recorded an episode last week on Meta's new local models.
9:34
Josh: They are looking to converge these two models and the agent harness into one
9:37
Josh: instance. that seems like it's on the right track to becoming a pretty serious
9:42
Josh: open claw competitor as well.
9:43
Ejaaz: But not for the reason that you might think. So Manus and Meta have differentiated
9:48
Ejaaz: their open claw offering in a very specific way.
9:51
Ejaaz: They haven't offered it necessarily. Their hit product hasn't necessarily been
9:54
Ejaaz: for retail users. It's been for advertisers.
9:57
Ejaaz: Now, that's where Meta makes majority of their money through Facebook,
10:01
Ejaaz: Instagram, and the likes.
10:02
Ejaaz: They basically have a crazy advertising network. In fact, I read an article
10:05
Ejaaz: earlier this week, I think two days ago, where their advertising rev is now
10:09
Ejaaz: becoming extremely competitive with Google, which is like the monopoly on the
10:13
Ejaaz: sector for God knows how long.
10:15
Ejaaz: Manus specifically is being used to help advertisers not only create their advertising
10:21
Ejaaz: campaign, but select product imagery,
10:24
Ejaaz: control the tools that they use to kind of manufacture and articulate specific
10:29
Ejaaz: advertisements to a target audience.
10:31
Ejaaz: It does all the research for them in terms of how much time they should expose
10:34
Ejaaz: an ad to a particular user. All these finer details is now managed by their
10:38
Ejaaz: version of OpenClaw Manus.
10:40
Ejaaz: And I think that's such an interesting niche because this is the sort of direction
10:43
Ejaaz: that Meta has started to take with their LLMs as well.
10:46
Ejaaz: Last week, they launched their own foundational model, the first that they've
10:49
Ejaaz: launched in over a year called MuseSpark.
10:51
Ejaaz: And it's not amazing as a general LLM, but it's great for specific tasks.
10:55
Ejaaz: And they're doing the same with OpenCore. So if you're an advertiser watching
10:57
Ejaaz: this, or if you're someone that wants to market their product on any of Meta's
11:00
Ejaaz: apps, definitely go check out this product because it's extremely useful for
11:03
Ejaaz: you right now and no one else is doing it.
11:05
Josh: And the final claw in a box service that we're going to be talking about today is Perplexity,
11:09
Josh: who just recently released perplexity computer and perplexity
11:13
Josh: computer is interesting and slightly different from the rest of
11:16
Josh: these options because perplexity is model agnostic it
11:19
Josh: actually uses i believe 19 separate models that you can kind
11:21
Josh: of choose from and it'll pick depending on what the task
11:24
Josh: that you're using is and that model agnostic allows it
11:27
Josh: to do things that are kind of unique to the perplexity platform it
11:30
Josh: does everything that you expect it to it can do the research it can
11:33
Josh: design code deploy projects whatever kind of miscellaneous tasks you want to
11:37
Josh: do but it also has the benefit of being able to do so i mean you could do it
11:41
Josh: fully in the cloud on the servers but they're also going to be rolling out a
11:43
Josh: version that runs locally on your machine similar to what we were talking about
11:46
Josh: with claude and open ai so perplexity is also in there it seems like the most
11:50
Josh: accessible one i think a lot of people.
11:52
Josh: Use perplexity kind of as a google extension and
11:56
Josh: the way they've integrated computer is very easy and simple to
11:58
Josh: understand in fact one of the new things that i found was cool
12:01
Josh: is they deployed a tax expert where it
12:04
Josh: will actually go through your filings and your w-2s and
12:07
Josh: it'll give you the best tax advice that it can
12:10
Josh: being a non-cpa but what i
12:12
Josh: find interesting about this is just how accessible it is
12:15
Josh: i find that a lot of the other things that we've talked about they're
12:19
Josh: fairly open-ended it's on you to figure out your use cases for
12:22
Josh: what you want to use this claw architecture for with perplexity
12:25
Josh: they really do a good job of kind of honing in on specific use
12:27
Josh: cases for you like the taxes like recently they interray with your bank accounts
12:32
Josh: where it could actually see your specific finances and and give you these kind
12:35
Josh: of unique but specific use cases and i think for a lot of people that may be
12:39
Josh: interesting so i would say perplexity probably is the one that holds your hand
12:42
Josh: the most and if you want something simple model agnostic that would be the one to choose.
12:46
Ejaaz: It's hilarious to me because every time I think perplexity doesn't have a moat
12:50
Ejaaz: and they're going to die, they somehow prove me wrong.
12:54
Ejaaz: Like you mentioned, it's model agnostic. They sit on all of these other LLM
12:58
Ejaaz: providers. Their moat isn't really the fact that they can access all these different models.
13:02
Ejaaz: It's because they've created something called an agent harness,
13:05
Ejaaz: which is, I guess, something that sits on top of all the agents that you would create.
13:09
Ejaaz: It helps provide the right context at the right time and using the right tools
13:13
Ejaaz: at the right time, which sounds like a claw, but they just do it better.
13:16
Ejaaz: Across a variety of different models where Anthropik is just focused on Claude
13:20
Ejaaz: and OpenAI is just focused on ChatGPT.
13:23
Ejaaz: And the advantage that they have is not only is it cheaper, but it's more contextually
13:28
Ejaaz: oriented around a specific task that you're doing.
13:30
Ejaaz: So you mentioned two, you mentioned taxes and you mentioned the financials,
13:34
Ejaaz: specifically like you can act like a CFO right now.
13:36
Ejaaz: That's a good example of like how it can go deep into a specific vertical for
13:40
Ejaaz: something that you don't want to do, whether it's taxes or sorting out your
13:44
Ejaaz: accounting or finances.
13:44
Ejaaz: And it can like figure that out in a couple of seconds.
13:47
Ejaaz: And all of it for accessible, I think it's like for like a $20 a month package. So very cool.
13:53
Josh: Yeah, that's the price of all these kind of all-in-one boxes.
13:55
Josh: Is there about $20 a month to use the software that gives you access to all
13:59
Josh: of these claw functionality tools?
14:01
Josh: Now, that is the conclusion of the closed source all in a box,
14:05
Josh: press one button and have everything done for you.
14:07
Josh: Next is the open source movement. This is where OpenClaw lives and breathes.
14:12
Josh: This is the most popular form of claw currently, just because OpenClaw went so viral so quickly.
14:17
Josh: And OpenClaw is, like everyone knows, the open source version that started it all.
14:21
Josh: It is the orchestration platform for agents that gets things done and is fully customizable.
14:26
Josh: So if you find yourself as someone who is a tinkerer, someone who likes to experiment,
14:31
Josh: someone who has resilience to fix a lot of bug issues that you may run into
14:35
Josh: this is the platform for you because of how customizable it is due to the open
14:39
Josh: source nature you have full platform control of everything that it does
14:43
Josh: from the md files which suggest how it should act and behave to the actual code
14:48
Josh: base that that you that it uses every day to actually make these decisions and
14:51
Josh: enact on your behalf if you're an open claw user you're actively using open
14:55
Josh: claw or you've at least already tried it it's more for the people who want the
14:58
Josh: full customization who want to do anything that they can imagine with it and
15:01
Josh: And there's also some interesting examples, Ides, you were telling me,
15:03
Josh: of things that people have done that you wouldn't really necessarily think is
15:07
Josh: something an OpenClaw instance would solve.
15:09
Ejaaz: Okay. What's a problem that you might have five years from now?
15:14
Ejaaz: Let me explain. Let me paint this picture for you. You own a house.
15:18
Ejaaz: But it's missing one quintessential component. You don't have a pool in your backyard,
15:23
Josh: Josh.
15:24
Ejaaz: Well, what if I could help you with that?
15:27
Ejaaz: What if not me could help you with that, but what if my OpenClaw or my Claw
15:31
Ejaaz: agent could help you with that?
15:32
Josh: Sounds lovely.
15:33
Ejaaz: Someone created an OpenClaw agent that finds 500,000 to $1.2 million homes without a pool,
15:41
Ejaaz: gets the images from Zillow or any other public website, renders a pool or an
15:46
Ejaaz: image of a pool into their backyard, and then sends it to them with a cold DM
15:50
Ejaaz: or cold email, basically saying, hey, doesn't this pool look really nice on you?
15:55
Ejaaz: You should like consider leasing us or using us to create and set up your pool for them.
15:59
Ejaaz: And the inbound has just been insane for it. So it scans satellite imagery for
16:02
Ejaaz: mid-market homes, filters lot by size, pulls homeowner direct from public records,
16:07
Ejaaz: and then like gets in contact with them with the new rendered image.
16:09
Ejaaz: And the conversion rate, supposedly, from the claim, this is Twitter,
16:13
Ejaaz: this is X after all, there are some crazy claims, has apparently been pretty
16:16
Ejaaz: damn good. And if you look at the video over here on the right,
16:19
Ejaaz: it looks pretty photorealistic.
16:21
Ejaaz: It has like a video version of it as well. So you can see it kind of like a
16:24
Ejaaz: pool floaty going around. Just a cool use of open core.
16:27
Ejaaz: And then this second example, which I found like super interesting was this
16:33
Ejaaz: mother uses or currently runs 11 open claw agents to help her raise her kids.
16:39
Ejaaz: Now, that sounds insanely dystopian, but the framework is ironically,
16:44
Ejaaz: in my opinion, one of the most useful examples of using open claw agents ever.
16:48
Ejaaz: So one example is she has a job to maintain.
16:52
Ejaaz: So one version of using all these open claw agents is she only uses voice space.
16:56
Ejaaz: She doesn't type anything. Now, Josh, I know you're a massive fan of using voice
16:59
Ejaaz: notes or speaking to your AI alarms.
17:02
Ejaaz: So am I. It is just a much more efficient way to use it. So she doesn't code anymore.
17:06
Ejaaz: She just speaks to agents about the specific type of thing that she wants to
17:10
Ejaaz: build in her particular job. And it goes away and it does that.
17:13
Ejaaz: But then she also needs to educate her kids.
17:15
Ejaaz: Now, she homeschools her kids. And what she does is she takes her phone.
17:19
Ejaaz: She takes a picture of a syllabus that she wants to teach her kid for the day.
17:22
Ejaaz: And it generates a personalized lesson plan for each of her different kids tailored
17:27
Ejaaz: to their specific preferences and she walks them through it.
17:30
Ejaaz: Now, remember, it's homeschooled. So you don't necessarily have all the apparatus in your home.
17:34
Ejaaz: She provides that context to all her open core agents and it knows which tool to use.
17:40
Ejaaz: So like if there's a toy that she can use to demonstrate the specific lesson,
17:44
Ejaaz: prompts her to go grab that toy and do all of this. So you might be wondering,
17:48
Ejaaz: well, this is completely insane. Why would you use AI to teach your kids,
17:51
Ejaaz: blah, blah, blah, blah, blah.
17:52
Ejaaz: The net net effect is it's freed her up to be more of a parent and spend time with her kids.
17:58
Ejaaz: So all of that to tie a loop around and said she has more time on her hands
18:02
Ejaaz: to spend time with the kids. So I thought that was a super cool use.
18:05
Ejaaz: Now, going back to your point around open source, open claw agents in particular.
18:11
Ejaaz: The customization is like a massive win for people, but it's also how cheap
18:16
Ejaaz: it is for you to be able to run it on a local device. Maybe it's on your phone.
18:20
Ejaaz: And also if you want to run like completely crazy projects like raising your
18:24
Ejaaz: kids, using 12 open claw agents, you probably don't want to be spending a bunch
18:28
Ejaaz: of different money on it.
18:29
Ejaaz: And so she specifically uses PicoClaw, which I'm referencing right here.
18:33
Ejaaz: And I think in some instances, IronClaw, when she's using personal data and
18:36
Ejaaz: she wants to use like a Rust-based architecture.
18:38
Ejaaz: All of that to say is all of these are free.
18:41
Ejaaz: You can download them, access it yourself, and you can customize it in whichever way you want.
18:45
Ejaaz: So if you're listening to this and you come up with a crazy idea,
18:47
Ejaaz: please do it and let us know what you built because we're trying to track all
18:50
Ejaaz: of these use cases whenever we can.
18:51
Josh: Yeah. And a testament to the claw architecture being so resilient and robust.
18:56
Josh: The real limitation, as it's always been, is how you can most creatively implement
19:00
Josh: these tools to be most effective for you.
19:02
Josh: And the ceiling is still like largely untapped. We don't know where this clocks
19:06
Josh: out because people are doing really cool and interesting things with it every single day.
19:10
Josh: There's so many great examples of it and great enhancements too.
19:13
Josh: Now what sits on top of this layer even are the markdown files.
19:17
Josh: So if anyone who's familiar with OpenClaw, there is like a soul.md
19:21
Josh: file and there there are a series of these markdown files
19:24
Josh: that kind of give the model intention they tell it
19:27
Josh: what to do they tell it how to act and there's an entire kind
19:30
Josh: of pseudo industry that's been built on top of this orchestrating of the agents
19:34
Josh: of telling them how to act by embedding interesting markdown files with unique
19:38
Josh: sets of directions now gary tan founder of y combinator he appears to have built
19:43
Josh: one that has kind of become the most popular one perhaps which is G-Stack.
19:48
Josh: He says, can you explain what G-Stack is and why it's this kind of valuable
19:52
Josh: second layer to be built on top of your claws?
19:54
Ejaaz: Yeah, I've seen conflicting opinions on both sides about what Karytown has built
19:59
Ejaaz: here because he hyped it up.
20:00
Ejaaz: Basically he said, I have made OpenClaw into this insane personal assistant
20:05
Ejaaz: that actually does productive work for you and mine's the best implementation of that.
20:09
Ejaaz: A bunch of engineers then evaluated what he had actually built and realized
20:13
Ejaaz: that he had just wrote a very extensively And in his fairness,
20:17
Ejaaz: very well articulated, meticulously detailed document, which describes what he wants his agent to do.
20:24
Ejaaz: Gary Tan, if any of you have watched any of his interviews, is an extremely verbose kind of guy.
20:29
Ejaaz: So it makes sense that, you know, he's worked with a bunch of these companies.
20:31
Ejaaz: He's had a bit of experience.
20:32
Ejaaz: He kind of knows what to ask for. And so the point that he proves with G-Stack
20:36
Ejaaz: is, number one, here's an easy and simple way that you can level up your open
20:40
Ejaaz: core agent in a matter of minutes.
20:42
Ejaaz: Use his framework and architecture and just fill in the blanks,
20:45
Ejaaz: like add in the details for your specific project.
20:47
Ejaaz: And you now have an amazing blueprint to build an amazing open core agent.
20:51
Ejaaz: But number two, how easy is it that it just requires two documents that you
20:55
Ejaaz: can spin up on your MacBook or on your laptop right now that can change your agent for the better.
21:00
Ejaaz: So the criticism for a bunch of industry folk was like, Gary,
21:04
Ejaaz: this is amazing, but this isn't anything necessarily novel.
21:07
Ejaaz: It's a skill issue. You need to be able to articulate.
21:09
Ejaaz: And that's basically something that is very apparent in the AI age.
21:14
Ejaaz: It's moved from being able to know how to code to be able to know how to communicate and explain.
21:19
Ejaaz: And that's the trend that we see with OpenClaw and a bunch of these other AILLMs.
21:22
Josh: And thankfully, all of this explaining happens in plain English.
21:25
Josh: And I think that's the most exciting and empowering part is it doesn't require
21:28
Josh: you to be technical or to be a developer in order to improve these things.
21:31
Josh: You just need to speak to it and kind of coerce it into doing the things that
21:35
Josh: you want to do, whether it be through Anthropic and Claude or through the OpenClaw
21:39
Josh: instance that is fully open source.
21:41
Josh: There is so much customization to be had and
21:44
Josh: i think it's kind of on you to decide where you fall
21:47
Josh: on the spectrum of plug and play where you just kind of want a
21:50
Josh: claw on rails that something like anthropic can provide to you or if you're
21:53
Josh: the tinkering type who really wants to build custom stacks like that mother
21:57
Josh: who is using it to help her parents or a company that is built being built around
22:00
Josh: creating pools there is no limit to the use cases you can find it's just a matter
22:06
Josh: of your preferences and how you
22:06
Josh: want to engage with this but it very much seems like this This is the new,
22:10
Josh: architecture that's going to stick around. This is kind of the foundation of
22:14
Josh: what an AI first operating system could look like.
22:16
Josh: And it's very clear that all these companies are converging on the same conclusion that this is true.
22:21
Josh: And this is a very important part of the stack to own. And I think the one thing
22:24
Josh: that is for sure is we are going to see this continued improvement and progress
22:28
Josh: on this at a very rapid rate.
22:30
Josh: I mean, Anthropic, they've been on fire, man. They've been releasing new features
22:34
Josh: for this every single day at this point.
22:36
Josh: So I encourage anyone with an account anywhere, really, just to try it out and
22:39
Josh: start flexing that muscle to see where it can improve these whatever parts of your life.
22:44
Ejaaz: Yeah, I wonder where this is going over the next couple of years.
22:48
Ejaaz: You have all these companies spending hundreds of millions of dollars to either
22:51
Ejaaz: acquire the thing or build the thing.
22:53
Ejaaz: They're all kind of building similar products. I think what it's ultimately
22:57
Ejaaz: going to come down to is the data and context that you can use to train these things.
23:02
Ejaaz: After all, that's what like made it a differentiated product.
23:04
Ejaaz: And that's what makes it an exceptionally good AI operating system.
23:09
Ejaaz: Two companies that I've noticed haven't really got a version of this,
23:13
Ejaaz: but probably should, is Google. Well, Google actually does.
23:16
Ejaaz: It's called Mariner, right? But no one's ever heard of that, right?
23:20
Ejaaz: No, but no one seems to use that. And they seem to have kind of like left that
23:23
Ejaaz: by the wayside to focus on their main LLM.
23:26
Ejaaz: And the other is Apple, who I'm actually more optimistic on because I know that
23:30
Ejaaz: they've signed the deal with Google Gemini,
23:32
Ejaaz: and they're going to be launching AI Siri hopefully for the fifth time on a
23:37
Ejaaz: failed attempt later this year which should end up becoming a really useful version of
23:42
Ejaaz: Their open claw agent. So I'm excited for these companies to launch their things.
23:46
Ejaaz: But I believe that is all we have for this episode.
23:49
Ejaaz: As Josh mentioned earlier, if you are listening to this, and if you are inspired
23:52
Ejaaz: by either the closed sandbox version from Anthropic,
23:56
Ejaaz: or the open source version from a PicoClaw or an IronClaw, please,
23:59
Ejaaz: we encourage you to go out and use these things because there is a draught of
24:04
Ejaaz: viable use cases for a few of these different AI things. Everyone has ideas,
24:07
Ejaaz: but no one's actually building the thing.
24:08
Ejaaz: It's never been easier to vibe code a custom AI agent. And we want to hear what
24:12
Ejaaz: you guys are interested in and are building.
24:14
Ejaaz: If you are a listener on this show and you aren't subscribed,
24:17
Ejaaz: please do so. Or if you're listening to this on Spotify or Apple Podcasts, please give us a rating.
24:22
Ejaaz: Leave us a comment. We always love hearing feedback from you guys.
24:24
Josh: And we'll be back tomorrow for the weekly roundup and possibly a surprise.
24:28
Josh: There's a rumor in the mill that Opus 4.7 is coming.
24:31
Josh: So we may have some big news dropping later this week. We'll see.
24:34
Josh: But we will be here to monitor the situation as always.
24:36
Josh: And as always, Also, thank you so much for listening, for watching wherever
24:40
Josh: you are, and we'll see you guys in the next one.