WEBVTT

00:00:00.000 --> 00:00:36.380
We have massive news in the world of AI agents. Codex just added slash goal to their desktop app, which will allow agents to work on long running tasks for more than a day. Anthropic convinced Andre Karpathy to join the team, and Google is rolling out their own version of OpenClaw, and that's just the start. Welcome to the first ever AI super app update. Every week, I'm going to break down what is happening across the frontier AI agent companies and platform. The goal of this series is simple, to make sense of all of the updates that actually matter. Model updates, platform changes, new AI agent abilities

00:00:36.380 --> 00:01:16.095
so that we can stay on the frontier of using AI agents. My name is Riley Brown. I'll be doing this every week. Please hit like and subscribe. It helps me out a ton. Anyway, let's not waste any more time. Let's dive in. Alright, guys. We have a lot to cover today. I'm gonna try to move through it swiftly. This is gonna be a really fun one. Okay. First, we're gonna be talking about some updates from Claude. For the first time in a couple of weeks, I actually went back to the terminal a little bit. I I was actually vibe coding and I kind of missed the terminal. Anthropic released a brand new feature that's really fun. And so for the longest time, if you wanted to run Claude code, you would go to your terminal and you would run Claude and this would start up Claude and you kind of work in this linear chat that works

00:01:16.335 --> 00:01:17.055
down.

00:01:17.295 --> 00:01:19.935
Well, you can type in Claude

00:01:20.095 --> 00:01:28.710
agents and this is a new way to interact with AI agents in the terminal. And so this will allow you to kind of itask.

00:01:28.950 --> 00:01:34.390
Hey, can you please research the latest news on the Google IO conference, especially Spark?

00:01:35.590 --> 00:01:44.655
And so when I press enter, it fires it off and it shows up here in working. Hey, can you please research all the updates to Cursor. I want a very nice clean list of all their new updates.

00:01:46.335 --> 00:01:50.095
Can can you please look up all the different updates to Codex, please?

00:01:51.215 --> 00:01:57.240
Tell me all the latest updates of Claude code, do some research, figure out all the latest updates. I'm on inline links of all this stuff.

00:01:59.320 --> 00:02:03.400
Tell me other updates in the world of AI. Turn it into a little landing page for me.

00:02:04.440 --> 00:02:09.000
Right. So we just fired off five different tasks. So the way that you navigate

00:02:09.080 --> 00:02:16.385
is you can scroll to each one and then if you press right on the keyboard, it takes you to the chat. If you press left on the keyboard, you can go back.

00:02:16.705 --> 00:02:27.250
And so it allows me and so I can see that it's using the web research specialist, is a custom sub agent that I created a long time ago. This is kind of fun way to interact with AI agents

00:02:27.490 --> 00:02:29.810
in the terminal, which is pretty fun.

00:02:30.130 --> 00:02:35.490
And Anthropic just released this. And next, we have really big news. Andre Karpathy

00:02:35.490 --> 00:02:37.250
joined Anthropic.

00:02:37.250 --> 00:02:49.225
So Andre Karpathy had this to say on May 19. Personal update. I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I agree. That's why I started this series.

00:02:49.625 --> 00:02:56.650
I am very excited to join the team here and get back to r and d. I remain deeply passionate about the education

00:02:56.890 --> 00:02:59.530
and plan to resume my work on it in time.

00:02:59.850 --> 00:03:11.285
And so many people are speculating how big this deal was, how much is he getting paid. Some people are speculating that it's over a billion dollars. Many people are saying that this is kind of the tech equivalent

00:03:11.285 --> 00:03:31.700
of Ronaldo joining Manchester City. Some other people are saying that this will cause a migration back to Claude from Codex. And if you've been watching my videos, you know that I've actually switched over to Codex, at least that's why I'm doing all of my agentic tasks just because they have a more unified product. But if you look at all of the talent they have on their team,

00:03:32.020 --> 00:03:33.780
you know, super.com,

00:03:33.780 --> 00:03:37.835
their CTO left to join Anthropic. Nikki Palmer, a DEBT CTO,

00:03:37.995 --> 00:03:39.835
left to join Anthropic.

00:03:39.995 --> 00:03:43.355
Brian McCann, another CTO from you.com,

00:03:43.355 --> 00:03:50.235
Workday CTO left. Bunn was acquired by Anthropic. People are genuinely leaving the most prestigious

00:03:50.235 --> 00:03:55.310
jobs in the world running massive companies to be an individual contributor

00:03:55.630 --> 00:04:03.950
at Anthropic. And as Anu put it earlier this week, it is the era of the polymathic individual contributor.

00:04:04.030 --> 00:04:06.925
We're seeing this across the board. Individuals

00:04:06.925 --> 00:04:12.125
with a ton of talent and areas of expertise, especially within AI and tech,

00:04:12.365 --> 00:04:13.565
are being recruited

00:04:13.805 --> 00:04:48.665
for hundreds of millions of dollars. And so this is just a very interesting trend. So the final update today regarding Claude is that they've just made some general improvements to their desktop app. You know, some people like the terminal interface for using Claude. I'm not one of those people anymore because they've created a desktop app. Claude has created one of the best super apps. A super app, in my opinion, is defined as an AI platform that you can chat with, that you can do knowledge work with, and that you can do coding tasks with, and it connects to all of your existing integrations.

00:04:49.230 --> 00:04:55.150
These AI agent platforms should connect to all of your existing tools and then you should be able to create automations

00:04:55.230 --> 00:04:57.550
and then whatever your AI agent does,

00:04:57.870 --> 00:05:07.885
it should be able to open up a little interface here on the right. And I'll be talking about super apps for the rest of time because I think this is the way it's going for enterprises,

00:05:07.965 --> 00:05:13.085
for business use cases. This is going to become a full browser. So that's what I define

00:05:13.245 --> 00:05:16.190
as an AI AI powered super app. Now I just think overall,

00:05:16.190 --> 00:05:24.910
the Claw desktop app has gotten a lot better over the past few weeks. It's a lot easier to use skills. I still don't like the fact that I can't use my

00:05:26.030 --> 00:05:35.805
skills. This is a skill that I created on Cowork called Riley Quality Content. If I go to Claude Code, I can't actually use that same skill. It doesn't exist.

00:05:36.205 --> 00:05:38.125
So that is why I'm using Codecs.

00:05:38.710 --> 00:05:40.950
I don't understand

00:05:40.950 --> 00:06:08.525
the relationship between Cowork and Cloud Code. Just give me one single app that can do anything, coding task or knowledge work. I think Cowork was one of their biggest mistakes. I will say when doing coding tasks on this app, the browser works a lot better. At the very beginning when they just released the new Claude code feature, the browser was a little spotty. It is getting a lot better, and I just think it's more responsive. So those are the updates today for Claude. Okay. So now let's move on to some updates coming from OpenAI

00:06:08.760 --> 00:06:14.280
and Codex. The same way that Claude has their desktop app, which is their super

00:06:14.520 --> 00:06:15.160
app.

00:06:15.640 --> 00:06:17.720
OpenAI has Codex,

00:06:17.720 --> 00:06:23.975
which is their version of a super app. One thing that they added to the Codex app officially

00:06:23.975 --> 00:06:24.695
is

00:06:24.855 --> 00:06:26.055
slash goal.

00:06:26.535 --> 00:06:30.855
On Codex, their super app, you can now type slash goal,

00:06:31.335 --> 00:06:36.935
enter, and you'll see it pop up right here. And now I can give it a super long task.

00:06:37.870 --> 00:06:47.870
And I just ask Codex, and you can think of goal in Codex as a way to tell Codex the objective for the current task so that it can stay oriented around the outcome,

00:06:48.110 --> 00:06:49.390
not just the command.

00:06:49.630 --> 00:06:55.255
So I could, in theory press command n for a new task slash goal

00:06:55.495 --> 00:06:59.175
and we see the goal indicator right here. I want you to create

00:06:59.575 --> 00:07:03.735
by the end of this, I want you to create 30 iOS apps.

00:07:04.055 --> 00:07:13.310
Create the code for 30 different iOS apps. First, create a research step. I want you to research 30 iOS apps and I want these

00:07:13.630 --> 00:07:15.150
to be downloaded

00:07:15.310 --> 00:07:19.915
on my phone. And so I want you to send me build links and use the

00:07:20.235 --> 00:07:22.075
we could use the, uh,

00:07:22.475 --> 00:07:24.155
porous ios

00:07:24.155 --> 00:07:24.875
skill.

00:07:25.035 --> 00:07:28.235
Okay. So now it's running. It's sent as goal.

00:07:28.635 --> 00:07:32.475
And so now it's just a different mode and Codex knows

00:07:32.475 --> 00:07:40.050
that it needs to work for longer. It's not just focusing on the command, it's gonna stay aligned towards the goal. It'll do more planning

00:07:40.290 --> 00:07:56.715
and it's not going to stop until it reaches that goal. And I've seen people run this for you know, some people have shown that it runs for four hours and some I saw some guy post on Twitter that it ran for one day and fourteen hours for one of his deep coding tasks.

00:07:56.955 --> 00:08:06.410
The next big update is you can now share plugins on Codex. Let's take a look at this post from the OpenAI developers account on Twitter. You can now share plugins

00:08:06.490 --> 00:08:52.335
in Codex across teams. Today, we're launching plugin sharing in Codex. We're making it easy to share plugins with the people you work with. One example is the First Week Quest plugin. From the plugin directory, I can actually find this plugin that I built. I can go into it. I can see the default prompts available. Now I wanna send it to one of my teammates. I'm gonna click the share button and we have this modal here where I can specify what teammate I wanna share this to or I can also change this to anyone at the workspace. What this means is after I click save, I've now shared this plugin for anybody in the workspace. And I can copy this link and share it out with my team, and they're able to see this plugin in their own directory. So what do I mean by that? Plugin sharing also works the other way. In the shared with you tab, so if I go back to the plugins directory, I can see all the plugins that have been shared with me by my teammates. I can see an example here called finalize code plugin, which for example, this plugin my teammate wrote, which is very helpful before I finish my PRs. So on Codex, if you go to the plugins tab and you scroll down here, these are all of the different plugins

00:08:52.415 --> 00:09:00.000
that you can use within Codex. And it looks very similar inside Claude Code except they call them connections. If we were to go to the Vercel

00:09:00.000 --> 00:09:00.960
plugin,

00:09:01.360 --> 00:09:17.115
a plugin is basically an assortment of apps and skills and other types of files that are bundled together and you can create these plugins and then you can at mention them when you're using codecs. Right? I can at mention the Vercel plugin.

00:09:17.355 --> 00:09:38.315
But if you wanted to create your own plugin, you could just say, hey, can you please create a plugin that does x y or z? It should interact with this API and do this. I wanna create some skills within it. I'll update some documents and I want you to make those the skills within those plugins. Right? You can just ask Codex to create a plugin or you can type slash

00:09:38.555 --> 00:09:43.915
plugin creator and say, hey, can you please guide me through the steps of creating a plugin?

00:09:44.955 --> 00:10:00.470
And you can run this and Codex will actually guide you through the process of creating a plugin in Codex. And once you're done creating that plugin that ideally would help you solve a specific workflow, you can share it with the team. And they made it easier to share it with team members and have this page

00:10:00.710 --> 00:10:07.105
where all of the plugins that have been shared with you will show up, which is really cool. And the next next update,

00:10:07.345 --> 00:10:16.945
which is full design mode, is another update inside Codex that makes it closer to being a full vibe coding platform. You used to have to use Replit,

00:10:17.185 --> 00:10:22.870
Lovable, or Bolt to get these, like, design mode features directly in the app. Well, Codex,

00:10:22.950 --> 00:10:29.910
and maybe soon Claude Code, these tools are becoming full vibe coding platforms. Here's the video from the OpenAI developers account.

00:10:30.310 --> 00:10:38.095
If you go up to the top and you click what is the button that you click there? You click this comments button. You can leave annotations.

00:10:38.735 --> 00:10:44.975
And then when you leave annotations, you can press the settings button and this will allow you to edit the text color,

00:10:45.295 --> 00:10:48.095
font size, and you can edit it

00:10:47.940 --> 00:10:52.900
directly, which is really cool. Here you can change the font weight

00:10:52.980 --> 00:10:53.780
directly.

00:10:53.780 --> 00:10:55.060
You can see it here.

00:10:55.300 --> 00:10:56.580
It's being changed

00:10:57.060 --> 00:11:04.115
and this is becoming a platform where you have absolute control over any type of app or website

00:11:04.115 --> 00:11:14.915
or HTML file that you have open inside Codecs. And one thing I do wanna say is this isn't actually full design mode like on Cursor which directly actually edits the code.

00:11:15.235 --> 00:11:20.100
What happens here is, is, let's say I wanted to edit this right here

00:11:20.660 --> 00:11:38.115
and I would press the settings button. By default, I can just leave a comment. This annotation feature was originally designed to leave a comment. I can just say make this bigger. And when I enter that, it adds it as an annotation. Right? It sees the context, it sees that it says make it bigger and stores it as an annotation.

00:11:38.355 --> 00:11:41.155
So then all I could say is do these,

00:11:41.475 --> 00:11:47.550
I can run it and it will just do those annotations. It'll make those changes. And so all this feature is

00:11:47.790 --> 00:11:56.430
is let's say I wanted to make this text bigger and I clicked on this right here. I could either add it to annotation that says make it bigger or I could come in here directly

00:11:56.510 --> 00:12:00.955
and I could change the font size to say 25

00:12:01.355 --> 00:12:05.515
and then I could describe these, make all these headings

00:12:05.515 --> 00:12:06.955
this size

00:12:07.275 --> 00:12:10.635
and when I press enter, it adds this as an annotation.

00:12:10.795 --> 00:12:22.280
And so instead of editing the code directly, which is probably the ideal experience, instead it adds it as an annotation which will tell the AI to change it from 18 font to 26.

00:12:22.440 --> 00:12:34.785
I can just say go. And because I have steering on by default, it'll just keep making all of these changes as I say them. But notice here it went back to normal. This is not going to change until the AI makes these changes.

00:12:35.265 --> 00:12:37.505
And you can see here after refreshing,

00:12:37.825 --> 00:12:38.705
this is bigger,

00:12:39.250 --> 00:12:42.370
this is bigger, you know, I changed this to be smaller.

00:12:42.770 --> 00:12:45.490
But the point is you have two tools at your disposal.

00:12:45.650 --> 00:12:52.335
You have the annotation feature and the screenshot feature. Whatever's here on your screen, you can immediately take a screenshot

00:12:52.495 --> 00:12:59.855
which automatically adds it to your clipboard. You could upload a screenshot. You have that screenshot for and then now you have annotations.

00:13:00.095 --> 00:13:00.895
Annotations,

00:13:00.895 --> 00:13:02.335
I could say return

00:13:02.575 --> 00:13:04.495
this to previous,

00:13:04.975 --> 00:13:07.180
and I can fire off that annotation.

00:13:07.180 --> 00:13:08.940
The next update by OpenAI

00:13:08.940 --> 00:13:11.340
specifically on Codecs is called

00:13:11.500 --> 00:13:15.900
AppShots and I'm telling you this is one of the most fun features to use ever.

00:13:16.140 --> 00:13:16.700
So

00:13:17.020 --> 00:13:27.625
what I can do, let's just say I am using this app right here. What I can do is I can press both command keys at the same time. Watch this. It

00:13:29.305 --> 00:13:30.825
immediately adds

00:13:30.985 --> 00:13:40.540
the website that I'm on and it takes a screenshot and brings me to Codex. I can say, please tell me about these things,

00:13:40.860 --> 00:13:42.300
do research.

00:13:42.300 --> 00:13:45.020
Let's say I'm in my email. I use Superhuman,

00:13:45.020 --> 00:13:46.700
which is just an external

00:13:46.780 --> 00:13:52.175
desktop app to control my email. No matter what application you're on, can press command command.

00:13:53.695 --> 00:13:55.375
Please research,

00:13:56.095 --> 00:13:57.695
uh, this company

00:13:58.255 --> 00:14:00.655
and write a draft,

00:14:01.055 --> 00:14:02.255
uh, response.

00:14:02.575 --> 00:14:04.150
Right? I can just

00:14:04.390 --> 00:14:12.310
do that right then and there. And the final example I wanna show you is one that since I have the Google Docs,

00:14:12.310 --> 00:14:13.590
um, integration,

00:14:14.150 --> 00:14:22.815
I'm able to work in a Google Docs. So I can just say for a video outline, this is a random example I came up with. Codex, write your four paragraph essay

00:14:22.975 --> 00:14:24.415
here

00:14:24.735 --> 00:14:28.655
in this doc. So now we can press command command. Watch this.

00:14:29.695 --> 00:14:32.815
Go. Now what it's going to do is because

00:14:32.895 --> 00:14:44.050
there it knows exactly which document this is and since it has a Google Doc skill, it's actually just going to write its response directly in this Google Doc here.

00:14:44.930 --> 00:14:54.925
Oh my god. It's literally writing in the Google Doc with look, you can see that it's using computer use. If we go back to Codecs, you can see that it's actually typing it directly in the document.

00:14:55.165 --> 00:15:00.045
So you can very easily, no matter what app you're using, is you can press command command,

00:15:00.445 --> 00:15:02.045
give context to Codecs

00:15:02.370 --> 00:15:12.210
and then just say to do something and it can actually control your computer in the background and do the task that you need. This is the future of super apps. Alright. So the next

00:15:12.610 --> 00:15:16.315
area I wanna move to is Google and Gemini.

00:15:16.395 --> 00:15:19.275
And the first thing I wanna talk about with Google and Gemini

00:15:19.435 --> 00:15:21.355
is they had a

00:15:21.595 --> 00:15:31.780
conference this week, Google IO. It's their biggest conference for their entire company where they announce all these different things. In my opinion, from an AI perspective, it a nothing burger.

00:15:32.020 --> 00:15:34.420
Basically, nothing happened except

00:15:34.580 --> 00:15:44.485
for one specific thing that I can't stop thinking about which is Gemini Spark, which we'll get to in just a second. But the main overall focus from this conference,

00:15:44.725 --> 00:15:59.250
from this Google conference, they were just mentioning AI agents a ton. But my main problem is that I don't know what their super app is. I showed you that Codex and the Claw desktop app, those are definitive agent platforms.

00:15:59.330 --> 00:16:01.330
If you want to use AI

00:16:01.650 --> 00:16:07.890
and you wanna use it as a power user and get things done and be able to accomplish long tasks

00:16:08.050 --> 00:16:13.435
and connect it with all of your tools, know exactly which platform to go to if you're OpenAI

00:16:13.435 --> 00:16:15.035
or if you're Anthropic.

00:16:15.275 --> 00:16:23.195
But with Google, you don't know. If you wanna do agentic tasks, do you use anti gravity, which is their like kind of coding agent platform?

00:16:23.275 --> 00:16:25.115
Or do you use Gemini,

00:16:25.355 --> 00:16:30.860
which now has a new thing called Spark which is their version of Manus or Perplexity Computer?

00:16:31.020 --> 00:16:35.900
It's a little bit confusing, I'll get to it in a second. Or do you use Google AI Studio

00:16:35.980 --> 00:16:42.825
or is this only for building apps? My thesis around this whole space is that all of these types of tasks,

00:16:43.065 --> 00:17:00.690
agent tasks, app building tasks, marketing based tasks, automation tasks, people are gonna want to do it from one single platform. They want all of their skills in one platform, their integrations in one platform, and then the web browser is what will give you that interface. Right? Because remember, these platforms, if we look at Codex,

00:17:00.850 --> 00:17:15.865
right, it comes with a browser. And this browser is a full browser. So, like, depending on what task you're working on, you can have any interface pop up here or you can search the web or have AI use browser use. Right? You can use browser use to

00:17:16.985 --> 00:17:19.945
control the browser. The AI can just control the browser.

00:17:20.105 --> 00:17:27.660
And so that is the paradigm I think we're moving towards. And with Google, you just don't know what their super app is. But I do believe

00:17:27.820 --> 00:17:30.540
that the Spark announcement

00:17:30.540 --> 00:17:38.220
is actually interesting. So this right here is a leaked video of Gemini Spark, but it's actually going to be a part of Gemini.

00:17:38.540 --> 00:17:42.695
And so it looks a lot like that kind of distinction between chat

00:17:43.015 --> 00:17:43.655
and

00:17:43.815 --> 00:17:50.135
co work, you know, in the Claw desktop app except now you have chat and Spark. This app between chat

00:17:50.135 --> 00:17:52.120
and Spark has has the potential

00:17:52.280 --> 00:17:59.720
of being a super app because here you can add your folders so it can also control your the folders on your computer,

00:17:59.960 --> 00:18:07.555
but it also can spin up virtual computers so it can do agentic agentic tasks in the cloud. It also has all of the different connections.

00:18:08.035 --> 00:18:10.275
It is very agentic.

00:18:10.435 --> 00:18:19.600
It shows the, like, the trajectories that the agents go on so they can think for a really long time and it just connects to all of your apps. It connects to your Google apps.

00:18:19.760 --> 00:18:24.480
Here, you can connect your tools. It has Canvas, deep research, guided learning.

00:18:24.720 --> 00:18:26.960
You can create music, video, images.

00:18:27.360 --> 00:18:33.695
You can use Notebook LM. You can have it search through your Google Drive and photos. This has so much potential.

00:18:33.695 --> 00:18:36.175
But where I lose interest,

00:18:36.495 --> 00:18:49.970
I just don't trust them to focus enough because they're also making updates to anti gravity, which is like the AI developer platform at Google. They're also making updates to AI Studio, which is their vibe coding platform, their version of Lovable.

00:18:50.290 --> 00:18:55.985
And so my thought is it's not gonna be as good of an agentic app platform as Codex.

00:18:56.145 --> 00:19:05.185
It's not gonna be as good at vibe coding as Lovable or Replit. It's not gonna be as good at doing like true engineering work as Cursor,

00:19:05.265 --> 00:19:11.280
the anti gravity platform. Right? The anti gravity platform is not the best at anything.

00:19:11.440 --> 00:19:13.360
And right now, that is the story

00:19:13.520 --> 00:19:14.480
of Google

00:19:14.640 --> 00:19:21.280
and Google DeepMind, the pioneers of AI. They are not the best at anything. Their video model's worse than SeedDance.

00:19:21.280 --> 00:19:39.630
Nano Banana's worse than ChatGPT Image two. Their LLMs are nowhere close to frontier level. Their vibe coding tool's not the best. Their chat experience is not nearly as good as the normal chat version on Claude. Claude, I think, is the best chat app. Nothing that they've created is state of the art.

00:19:40.030 --> 00:19:45.790
You're gonna have to really go deep into their bag. Right? They have video analysis is one of the modalities

00:19:45.790 --> 00:20:09.720
in their Gemini app. You can upload a video, you can analyze that. That's like the one thing it's best at. But other than that, I hope they just announce a code red and focus all their attention on one of the apps. And maybe it's just the Gemini app and they focus on the chat and the Spark. But who knows? Right? We'll see. And just speaking about not being the best, the final update for Google is just their new model, and the new model is Gemini 3.5

00:20:09.720 --> 00:20:10.600
Flash.

00:20:10.600 --> 00:20:20.015
So everyone was hoping for this model to be released, Gemini 3.5 Pro, which was their Frontier model, but by the looks of it, it's not going to actually be frontier.

00:20:20.015 --> 00:20:25.215
They seem to be well behind OpenAI and Anthropic in terms of creating a really good

00:20:25.935 --> 00:20:29.295
agentic model like Opus or 5.5

00:20:29.295 --> 00:20:30.415
from OpenAI.

00:20:30.815 --> 00:21:01.485
And if you wanna watch Theo's video, you'll you can dive deeper into his yet a lot of negative things to say about Google. I'm not gonna agree or disagree. I just talked to enough people that said it's not worth testing. It's really fast but expensive. It's not as good as Composer 2.5 which is Cursor's new model, which we'll get to in a second. And so, yeah, pretty disappointing on the AI front for Google. Right? They released kind of this new video model called Gemini Omni. It's just not as good at generating video as Seed Dance by ByteDance,

00:21:01.485 --> 00:21:13.690
which is a Chinese company. So again, they're just not the best at anything. Then finally, we're moving to Cursor. Their first big update is they released a brand new model. If you don't know already, x AI or SpaceX

00:21:13.690 --> 00:21:17.465
agreed to buy Cursor for $60,000,000,000

00:21:17.465 --> 00:21:18.825
with a $10,000,000,000

00:21:18.825 --> 00:21:20.105
opt out clause.

00:21:20.265 --> 00:21:23.385
So Cursor now has access to a lot more compute

00:21:23.545 --> 00:21:26.985
and they were able to train Cursor 2.5.

00:21:26.985 --> 00:21:35.790
They're training a bigger model which will be Composer three, I believe, which is supposed to be way better and I'm telling you, this model is really good. I

00:21:36.030 --> 00:21:42.190
really am enjoying using Cursor and their in app browser is arguably better than Codex.

00:21:42.190 --> 00:21:46.995
If you haven't used Cursor in a while, it looks a lot like Codex

00:21:47.235 --> 00:21:51.635
in the sense that you can you have your workspace here on the left, you can add a new project,

00:21:52.195 --> 00:21:56.435
and I can, uh, open a folder here, call this test test,

00:21:57.300 --> 00:22:00.340
and we can open this up. So we're in a new project

00:22:00.580 --> 00:22:03.860
in test test, say, please create

00:22:04.180 --> 00:22:05.540
a cool

00:22:05.700 --> 00:22:09.940
landing page that looks like link

00:22:09.625 --> 00:22:13.945
tree. And so we are using their brand new Composer 2.5 model,

00:22:14.185 --> 00:22:15.865
which hardly uses

00:22:16.425 --> 00:22:18.905
any tokens. It it is

00:22:18.905 --> 00:22:33.770
very, very cheap to use. Like you could use it for like an hour and I don't think it would cost you a dollar. And so it is really fast and fun to use and they do have an in app browser just like Codex. As you can see here, we can search,

00:22:34.010 --> 00:22:36.890
you know, this is a full browser just like

00:22:37.130 --> 00:22:39.875
Codex. We can go to any website that we want.

00:22:40.915 --> 00:22:42.675
And here, it's already done,

00:22:43.315 --> 00:22:44.195
I believe.

00:22:44.515 --> 00:22:49.795
We can open up this landing page. Look at that. That I did not skip the video at all there.

00:22:50.115 --> 00:22:54.400
It's already done and it knew who I was, Riley Brown. I'm gonna say please

00:22:54.400 --> 00:22:55.200
make

00:22:56.080 --> 00:22:59.120
this in the style of cursor.

00:22:59.440 --> 00:23:05.235
It is much faster. This would not go nearly as fast if you use Codex or even Claude.

00:23:05.315 --> 00:23:15.715
It is significantly faster and it's even searching the Internet. This model is pretty close to Frontier, especially for front end work. And now it's generating all the code, making some changes,

00:23:17.315 --> 00:23:20.240
and boom. I think if we hit it with a refresh,

00:23:20.400 --> 00:23:23.120
there we go. It went to Cursor,

00:23:23.120 --> 00:23:24.320
grabbed the brand.

00:23:24.720 --> 00:23:28.960
This is looking pretty good. Yeah. Cursor has a brand new model. It's much faster

00:23:29.360 --> 00:23:39.695
and I have yet to make a video on Cursor. I may make a complete guide on Cursor talking about all the new updates, but it is a lot like Codecs. You just can't create documents,

00:23:39.855 --> 00:23:41.055
PowerPoints,

00:23:41.215 --> 00:23:44.175
and spreadsheets the same way you can in Codecs

00:23:44.175 --> 00:23:56.960
or Claude Cowork. They don't open up naturally here on the side, but my guess is that's coming very soon because Cursor, in the SpaceX announcement where they announced that they were going to acquire

00:23:57.120 --> 00:23:59.920
Cursor, they said we are trying to build

00:24:00.240 --> 00:24:01.040
the

00:24:01.040 --> 00:24:24.610
next generation platform for coding and knowledge work. So I know they want to get in this market of using AI agents in business. This is a full super app contender and I'll be covering it very closely going forward. They are becoming a full super app which means they're gonna have feature parity with Codex and Cloud Code. If you come up here, not only do they have a marketplace which has all of the, uh, integrations,

00:24:24.690 --> 00:24:26.290
same thing as Codex

00:24:26.290 --> 00:24:32.770
and Cloud Code, they now have automations. So that's what we covered today on this week in super apps.

00:24:33.265 --> 00:25:17.780
The point of all of this is that we have these companies that are going to be worth trillions of dollars all moving to try and capture the enterprise market, and it is going to be a bloodbath. They're gonna battle to the death to try and capture all of this value and in the process they're going to create platforms that are so good and it's going to move so fast because they're competing. It gives me a reason to go through and like study all of the stuff that's going on with these companies and if there are more contenders out there, I will follow those companies as well. And by the way, if you guys wanna support me and take a look at the project that my team and I have been working on, we've found a way to add quad code and codex to iMessage. So we bought the domain choruschorus.com.

00:25:17.780 --> 00:25:28.020
This is my AI agent James Bond. Hey James, what are you up to? And so I'm talking to Claude Code. Very soon, you can talk to Codex and so you can actually sign in with your ChetGBT subscription

00:25:28.100 --> 00:25:33.915
and your AI agent will respond and it can do anything. Like, I can say, like, build a landing page

00:25:34.075 --> 00:25:36.475
for me, Riley Brown,

00:25:39.595 --> 00:25:48.540
you know, and it will do it. It already created a little landing page and this is running in the cloud. You can see here that it's running. It has like a little sandbox sandbox domain.

00:25:48.620 --> 00:25:55.900
I can do this from my phone. I can do this from iMessage on my computer and it all lives on chorus.com.

00:25:56.060 --> 00:25:59.260
You can name your agent. Mine is named James Bond.

00:25:59.580 --> 00:26:04.085
And just like Claude, uh, just like Codex or Claude Code, um, you have files,

00:26:04.085 --> 00:26:10.325
skills. You can very easily give it a phone number. Like, you can just immediately start texting it. You can add WhatsApp.

00:26:10.485 --> 00:26:44.390
Slack is coming next week. All of the connections in the world. And soon, it's gonna become a lot more proactive. We wanna make one that's a lot like easier to get started, and we're just gonna have it start doing tasks for you, and you can create cron jobs. These are my four automations that I have it do. We're gonna simplify the product a lot, and we're making it mobile first. But this is kind of our cloud based mini super app. Our team has been working really hard on it. We've been having a lot of fun building it, trying to figure out its personality and being able to swap in Claude code or codex or any of the open source models. Give this a try. You should get, I think, $10 of free credits. Thank you guys so much for watching. I'll see you here for the next video.
