Loading summary
Transcript96 lines
- [00:00]
A
Some of my smartest friends are saying OpenAI's Codex is the super app for AI. They're saying you pair that with GPT 5.5, images 2.0 and use Codex's interface and you are going to be unstoppable. So I called my friend Riley Brown to convince me that I should be using Codex with all these models. And in this episode we go through everything. How to use it from A to Z, how to create skills, how to connect it to notion, how to use Remotion to create beautiful, beautiful videos. If you've been meaning to get good at Codex, but you haven't had a second, well, this is your episode. You're in for a treat. And if you want to show some love to me and Riley, give this a like and comment and I'll see you in there. We are blessed to have Riley Brown back on the podcast. Riley, by the end of this episode, what are people going to get out of this?
- [01:02]
B
People are going to learn how to use the best interface to use AI agents, in my opinion. I think Codex by OpenAI. And that means if you have a ChatGPT subscription, you have access to Codex, which I think is the most powerful way to use AI agents. And we're going to learn the platform and we're going to talk about GPT 5.5, which was released yesterday. As of filming this, I don't know when this episode will be out. We'll talk a little bit about browser use. And you know, I've been a little skeptical about AI controlling browsers. I think it's kind of slow and a little boring. This is the first time that I see it, I'm like, oh, it's actually starting to be faster. And I could definitely see by the end of the year, these browser agents are going to be as fast as humans. So we're going to talk about that. We're going to talk about computer use. We're going to talk about just doing all of your knowledge work and coding work inside one single platform, which is Codex. How does that sound?
- [01:58]
A
This. Well, I just have to be honest with you. I got to say something. I am not on Codex today. In fact, I have never downloaded Codex. It's just I'm not in that world, okay? And this episode for me is a firsthand, basically tour hacks, tips and tricks. By the end of this episode, I want to be converted. I want to be converted to Codex in that ecosystem. I hope so. But I might be wrong. By the end of this episode, I might be like, you know, What? I'm just going to stick with my own stack. And you know, this is. There's just a lot of smoke here, so let's see.
- [02:38]
B
No, for sure. You know what? And I want to preface this by saying, like, I don't hate on anyone for using any stack. I think it's kind of dumb to do that. I think you should pick a stack and you should stick with it. I think moving from hot tool to hot tool is actually not the right strategy. If you're using Claude code inside cursor, Right. If you use cursor for all your knowledge work and that works for you, great. Just keep doing that if you want. I'm kind of permanently switching to codecs just because I work with seven engineers and all of them have switched to codecs. And we agree that it's pretty amazing and you can use Claude code inside of it, which is another bonus. But if you don't mind, I just kind of want to hop into kind of an excalidraw, kind of do a little precursor to codecs, if that's cool with you. Yeah, okay. So, yeah, so I like to start with the basics. I've been really trying to get people to understand the difference between ChatGPT and Codex. And so a good way to think about Codex is it is a state of the art AI agent on your computer, which is the Codex model or the GPT 5.5 model in a nice interface that allows you to build an app, create any type of document, control your computer and create automations. You can see that, right? This is the Codex interface. It looks a lot like ChatGPT, right? This is a new chat on Codex. And what I like to do whenever I create a new, like whenever I go to Codex and let's say I'm working on a new episode, maybe I'm working on the Startup Ideas podcast landing page. And then I'm also doing some market research. So we can actually create a new project and we can actually start from an existing folder. And let's just call this folder Greg or let's call it Startup Ideas Podcast and we can just create a new folder and this folder will be over here on the side. Can you see this? Maybe I'll zoom in a little bit. And anytime I hit new chat, this chat will show up in here. As you can see here, I have all my threads over here on the side and they're all kind of organized. Notice there wasn't anything in this one. And you can very easily just remove these from the side. So it's just like you have your folder and your chats within each folder, which is just a project. And so like research Startup Ideas podcast, look at his videos. And so I have a Skill which is YouTube researcher, and this allows me to immediately pull transcripts from YouTube channels. So I can say, like, take the transcripts from his last 10 videos and tell me only what he's doing wrong, Be negative, make a report. And so then you can fire this off. And what I like is it's very tailored to multitasking. So I can press command N. And this interface is a little similar to Manus. If you've ever used Manus before, it's very easy to see which. What your agents are doing, right? If it's done, you'll see a blue dot. As you can see here, these two are done. These are kind of like unread chats. So like it means it's done and I haven't clicked on it yet. While it's working, it kind of spins around. And so like what is the Startup Ideas podcast? And so you can see here it kind of just organizes chats into threads. And what we're seeing is this is kind of the dominant interface to interact with AI agents. You can see here, this is the new Claude code desktop app. Right here. You can see their code feature. It's very similar. You have these folders with chats underneath it. And if you've seen the new cursor interface, it's very similar. So all of these companies are deciding that this is kind of the correct interface to interact with AI agents. It just so happens that I think Codex is the best one. So let's go back to, to arc. And so the way that I see it is like in 2025, that was the year of the TUI, right? We had, we had this type of interface with Claude code. And it's kind of a. It's bittersweet because I missed these early days of using cloud code. I loved the terminal. It was really fun. And I think back to the early days when you and a few others were talking about vibe marketing. And I think it was super directionally correct using AI agents to do a lot of marketing tasks. I just think I'm curious to hear what your observation was, is this just wasn't a great interface for a lot of the people who wanted to, to get the benefits from it, but they didn't want to open a terminal or create skills like in these files. I think people in business just want an easier interface to do all of these agent workflows. And so what all these companies are deciding on is that the GUI is better. So we're moving from a terminal interface to a more clean, simple interface. You have your chats on the left, your agent in the middle, and then whatever the agent is working on on the right. Another way that I like to think about this is Claude code, or this Codex is like Claude Code and Cowork in a single product, but it's even better than that. And the reason why I like it more than Claude Desktop is Claude made a decision to split up Cowork and Claude code. I would definitely try Claude more if they didn't do this. I do not like them. They decided that people who are in business are going to use one product. People who want to code are going to use this other product. They have different limitations. Right? You can't just ask Cowork to build an app. You have to go to Claude Code to do that. Claude Cowork has different sets of permissions than Claude code. And you know me, I'm super hyped on Vibe coding. I think Vibe coding has gotten so easy that, like, 95% of the things that you would want to code, it's as easy as creating a presentation. So I think. And all a PowerPoint doc. Sorry if I'm kind of jumping around here, but within the Codex app, right? Codex app is. I can create this app right here. So this is an app. I posted this on Twitter yesterday. This is in one prompt. This is like a train simulator, and it's super good at physics. And this is me just testing the model, having fun. And I even added a crash counter so these trains can crash into each other. And it's presented by Allstate. But the point is, in the same interface, you can create an application and you can also create spreadsheets. As you can see here. It can create charts, spreadsheets. You can also create Word docs. You can edit and export these Word docs, and you can create PowerPoint presentations and you can export them with different apps. So you can export this to Canva, and this will just open it up directly in Canva. So I can make further edits. And they're going to add more exporting features as time goes on. Because me, as a startup founder, I'm constantly making all types of docs, and I'm making a lot of landing pages, lead magnets. I want to do it all in one interface. And so that's kind of the main reasons that I use Codex. Any thoughts so far?
- [10:08]
A
A million thoughts. But the biggest thought I have is what you're basically saying is Codex is the fastest way to do the most amount of things. Basically.
- [10:20]
B
Yes, it is. And it got all the primitives. Right. And okay. Actually we can dive into some more of the reasons that like these are why I like codecs over certain tools. There's a lot of reasons I like it that these things are also in Cursor and are in the Claude desktop app. But here are the things that only Codex does or Codex does better than all of them. Right. It is the only interface that does Vibe coding and knowledge work in one platform. If you go to Cursor and ask for a doc, it'll give you. It can create HTML file and run it in the browser, but it can't just create a doc and let you view it like a cloud artifact in one interface. You can do this on Claude desktop app, but again, you have to go to this Claude Cowork feature. I'm not a fan of Cowork at all, which may be a narrative violation on your channel. I don't know if you've been talking about Claude Cowork, but that's okay.
- [11:14]
A
I haven't, but I'm intrigued with Claude Cowork personally. I spend most of my time in Claude code, but I was thinking of doing a full on episode on Cowork because it's one of. One of the questions I get asked the most, which is should I be using Cowork and how to best use it.
- [11:34]
B
But wouldn't you wish that Claude code could just do what Cowork could do from the same interface?
- [11:42]
A
Yes.
- [11:43]
B
Yeah. And so to me, I don't see any good reason why you wouldn't allow Claude code to do the stuff that Cowork could do. And so that's just my beef with Claude Cowork and that's why I've been annoyed by it and I haven't made any videos on it. It's because every time I try and use it, I run into some sort of limitation because the way that they set up the environment where the agents can do stuff, there's way more restrictions on Cowork, which makes it super annoying to use, in my opinion. Anyway, I think codecs models are better at really complex tasks. And we actually, we know this for sure. We've tested this extensively as a team. We're highly incentivized to write the best code and write it fastest in the quickest amount of time. And Codex just does a better job with infrastructure with really hard tasks. I'VE one shotted the other day I one shotted an app like Replit, a Vibe coding tool that spun up sandboxes and you could Vibe code web apps on your phone in one shot. It took an hour and 20 minutes to generate the code. And this was on 5.4. It's pretty impressive. So if you remember last year OpenAI started working on Atlas. I thought it was kind of, I was just like, damn, they're kind of losing focus. They're building this Atlas feature, you know, it's the standalone browser. How many people are going to use it? Well, recently they had like this big red alert at their company and they decided to kind of push all of their efforts into one single app, which was Codex. And now what they're doing is they're just putting Atlas inside. They're basically putting Atlas inside like this is the Atlas browser. And I know this by interacting with people at OpenAI, this is going to become a full web browser. So you're going to be able to log in, it'll remember your login info. And so instead of having like a web browser with all of your tabs open and you have to decide what task you need to use, instead you'll start a chat with an AI and the browser will be there right on the right and you're, you'll have different windows open for the specific task. And I think that's just a better way to do knowledge work, you know, and you can just open up tabs directly in the app that you're creating, whatever it is that you want to create. And you'll be able to open up Notion and you can download the Notion plugin and so you can have your AI edit Notion and you'll have Notion open directly in the app in the web browser, which I find to be pretty cool. And so that's another thing. They also have a built in Remotion plugin and because their web browser is the best, you know, I created this video the other day. Oh boy. I've been doing a lot on here
- [14:22]
A
and Remotion for those people who don't
- [14:24]
B
know is what so Remotion. And by the way, in three hours I'll be launching a video on how to use codecs for remotion. Remotion turns code into videos so you can create motion graphic videos with code. Well, guess what? Luckily for you, AI will write all that code. So you can just create, you can just type out what motion graphic scene you want and it will create it. And Claude, as many people in your audience know, Claude has went on this massive run of launches every single day. And by the way, it's reconnecting. I think Codex is getting a lot of load right now. There's just a lot of people using
- [15:08]
A
it because you're talking about it, you're out there saying it's a super app, it's the best, it's where you should be spending your time. You want to bet on OpenAI and then all of a sudden it's reconnecting.
- [15:20]
B
Yeah, dude. I mean, what can I say? Well, I'll try and be quiet before we shoot our next video. I'll just pin this. Ideally it'll just work in just a second. We don't need to show it for now. But the point is you can generate really high quality motion graphic videos. And Anthropic did this when they first started doing their product launches. They were just launching them with Remotion videos. And it's really good for software products where you like have something happening on a user interface. It can generate the user interface, it can generate the motion. And the thing that's really cool about Codex is you can just search the Remotion plugin and once you enable the Remotion plugin, any chat that you use, you can say emotion and you can say emotion, please create a video. And then I created another skill which is called Internet Image Puller. It doesn't just pull images, it pulls all of the assets for any given company from the Internet. So it might work for 10 minutes. It'll pull all of a brand's assets, put it in an HTML file, and then Remotion or Codecs can have all of that information for Remotion. So it can use your logo, it can use your colors, it can use your fonts in the video. You don't have to like manually tell it what to do. It'll just pull all of that information. And so that's a really fun way. And this is how we do a lot of our launch videos and some have gotten over a million views. So this is a really useful thing. It was just really annoying to use with all of the other tools. Codecs is the only one where you can just at Remotion and it opens directly in the app.
- [16:59]
A
Yeah. And you know, for people who aren't super familiar with Remotion, it's not just that it takes concepts and turns the videos into code or creates videos that are code. These videos are so high quality, it is actually insane. Like, actually like I've. I know multiple people who've gone from zero to a hundred thousand Instagram followers by doing Remotion videos. Now, they're high quality Remotion videos and they go back and forth and back and forth. I'm not saying you one prompt it and you get the best possible, you know, videos, but these are. Is that Remotion right here?
- [17:41]
B
Yeah. So this is 7, 7, like almost 800,000 views. This is all Remotion. You can't hear it, but like you can create motion graphic videos.
- [17:50]
A
Yeah.
- [17:51]
B
And it's really easy to just like create pretty simple videos. And I think when you're doing a launch video, if it's a good idea, you should try and keep it as simple as possible. Never have multiple things happening at once. Oh, okay. It just popped up. It's testing if it's running. But like this is what would run on the side of your screen and it will in just a second. But it literally creates this like cap cut, like timeline. And so this is. And so these are all of your compositions. And so here we have this demo that I made yesterday. And this was in one shot. And the reason why I was able to generate this in one shot is I actually have this thing called Brand Assets. This contains all of the assets that I might want to use. And so once you've generated a video with all of your brand assets, you can just reference that to your AI and you can say, hey, please use this background or use this component or use this logo. And you notice here, I just labeled all of these logos here. And so I can always say, like, please look at the brand assets. Use the X logo, the YouTube logo, the TikTok logo, and the cursor logo and, and put it in this video, in this scene. And so if you just get everything that you might want to use in here, it's easy to tell your AI to use it later, if that makes sense. Anyway, you can create videos and you can open it right here in the browser. As you can see, that took a little bit because Codex is running a little slow. But this is kind of like an everything app, which is pretty cool. And then to finalize why I use it over other ones is the computer use is better. It is faster than any computer use that I've used before. And this means you download computer use as part of Codex. It can literally control your computer. Not only can it control your computer, it can control multiple of your apps at the same time, which is a mind blowing experience. And you can see the mouse moving around on your device. I had it controlling Canva yesterday. It was moving around pretty fast, controlling Canva. It exported the right document and it fed it back into Codex. It was a really fun workflow and that's just a really cool experience. I highly recommend trying this just to kind of see where AIs at in terms of computer use. And then as I talked about earlier, it can control the browser that's directly inside Codex. So pretty soon we'll live in a world where an AI will be able to go to any site, open the browser, you're going to be logged in. So it'll have all of your login information. It'll be a full browser like Atlas, and it will be able to take any action. So there will be basically no limitation to what it can do. And then they released this new Chronicle feature so that it kind of like remembers everything. So it'll kind of. And this one might have some privacy stuff that you might want to look into if you're super big on privacy. This is my work computer and I like testing stuff too much, so I had to enable it. But it basically watches your screen for context so you don't have to explain yourself again. It just kind of knows what you're working on and it stores these as like little Codex memories and it allows you to stay in flow so you can move chats around. You don't really need to be that organized. The AI will just kind of have context over what you're doing. And these are kind of the main reasons I really like Codex.
- [21:17]
A
I didn't know about Chronicle. First time I've ever heard of that. Is it worth. I know there's privacy issues and security issues. I know the comment section is gonna not love it, but from a productivity standpoint, are you seeing meaningful uptick?
- [21:34]
B
So I. They released it two days ago, so I'm not gonna lie, there's been way too many things to test. But if you go to personalization and you scroll down and go to Chronicle, I recommend learning more. This is not an endorsement. I'm using a work laptop. I test everything. I go unlimited permissions. I don't care about the privacy stuff. I'm testing everything. That's what I do. I love it too much. So by no means am I telling you to do this and that it's worth the risk. I have no idea what the risk is, but you should look into it because it is an interesting technology. At the very least, having an AI, assuming it's secure and everything, having an AI that understands what you're working on, so. So that when you type Something it knows exactly what you're referencing that is inherently useful. Whether or not it works really well, it's up for debate. I haven't tested it long enough. You know, it came out two days ago.
- [22:27]
A
So what else do I need to know? Like, because I'm downloading Codex for the first time, so what are things in terms of setup, remotion? I'm going to add projects. I understand plugins.
- [22:42]
B
Yeah. So, yeah, let's talk about plugins. So plugins, you know, there's a lot of words right now and people talk about mcps, they talk about plugins, they talk about skills, they talk about integrations, connectors. Right. You have all of these words and if I'm being honest, the difference between them is very blurry. Right. Just kind of. I clump all of those words together and I just call them skills. It just makes sense in my brain. Right. You have an AI agent by default that's really smart. It can search the web, it can create apps, but if you don't give it any skills, then it's only going to be a generic model. What you need to do is you need to give your agent access to the tools that you already use to make it useful for your workflows. And so here they call them plugins. And again, skills and plugins, you can create your own skill. You actually can't build your own plugin unless you get it approved by OpenAI. So that's kind of the difference. Plugins are more official. And so one of the plugins that I use is Slack, so it's connected to our Slack. I can ask any question. You know, we have like 30 channels as a company and I can just ask AI to summarize everything that pertains to me. And because Codex keeps memories of me and I've given it a document that I've told what I think is important. It can find the things that are important, it'll surface them to me and you can also connect it to your email. So it can send and receive emails or you can like tell you who emailed you if something's really important and you can connect it up to notion, you can connect it to sheets, and, you know, pretty soon you're going to be able to do it for all of them. The Expo one is pretty useful. It allows you to kind of view a mobile app preview directly inside codecs, which is really cool. And then obviously remotion and Canva, you can just connect it to your existing tools. And if you don't find the tool that you want to use in plugins, you can create your own skill and you do that simply by creating a new chat and you say, I want to create a new skill that does whatever, blank. And then you can also say use the skill. Wait, I think it's skill. Oh, maybe they don't have it anymore. To reference a skill, you use a slash. To reference plugins, you use the ention, which I think is weird. I think they should combine all of them. Right. So I could say use computer, use browser, use. To use a skill, you do a slash and you can see all the slashes here. I don't see. They used to have a skill builder skill, but they don't have it. I think it's just smart enough to just say create a skill and it'll just create it. After you create a skill, it'll show up here in this tab. And here are all the skills that I've created. I've created.
- [25:39]
A
It's so confusing, right? Because like, you create a skill and then it puts it into the plugins tab.
- [25:46]
B
Well, no, no, a skill is in the skills tab. Oh, yeah, yeah, yeah, you're right. It's in the plugins tab, has plugins and skills. And so I think this is super messy. Right? That's why I think they should just combine it all into skills or pick one and just stick with it because it is super confusing. And the fact that they have the different, like you have an at for plugins and then you have a slash for skills I find to be, like, kind of annoying. But I will say I've tried all of these tools. This is the simplest version that exists yet.
- [26:22]
A
Okay. And in the simplest way possible. What is the difference between a skill and a plugin?
- [26:30]
B
Honestly, I think plugins are just like a deeper integration. I would be lying to you if I knew the technical thing. I just know that, like, if you're a company and you want to create an official plugin, you go to OpenAI and you create it with them. It's similar to, you remember the apps SDK on ChatGPT? It's more of an official process. Right. There's a reason why, you know, we can't just. We couldn't spend the rest of this video making the Greg Eisenberg plugin and then get it approved. Right. After this video, you'd have to reach out to OpenAI. I'm sure they have thousands of companies trying to create plugins and you can only imagine the value of showing up here on the plugins tab. Right. So this is Kind of their more official integration. And then skills are like what you create. And these are just again, all of these skills are mostly just like a folder that is stored somewhere in here and then there's a skill MD file in that folder that has specific instructions. And AI is really good at creating skills. So all you have to do is just say, hey, any repetitive task that you have in codecs, say hey, I want to create a skill so we can do this faster next time. And then the AI will be like, okay, cool. It may ask you some follow up questions and then it'll just create the skill and then it'll show up here.
- [27:45]
A
Yeah. One thing you might want to do is, you know, if you're downloading Codex for the first time, maybe you connect it to your Google Calendar and you basically say like, here's what I do on a given week and you know, help me create a set of skills that can help me do my job better.
- [28:04]
B
Absolutely. Yeah. I think one of the best ways to do it is just to like literally speak into your phone. I just, I tell everyone to do this. Just yap about everything that you do on a day to day basis. Like go through every task, don't skip any details, take 20 minutes to just speak about everything, get it in a document and then just look at it and be like, okay, what parts could I automate with something? Anytime there's an automatable process and you can convert it into instructions that could be completed on a computer, that is a skill that you could create on codex. One really cool thing is. Right. So we just created this report using my YouTube researcher skill and it looked at the startup ideas podcast, negative only
- [28:50]
A
content report and cheer slowly goes down my cheek. I love this. Let's, let's.
- [28:57]
B
All right. The show is drifting away from. I meant to do it as a joke. I'm not actually trying to nitpick your channel, but I will say that as uncomfortable as it is to get AI to critique you, it is a useful activity. I've done it before, you know, like sometimes you can't turn a blind eye, you know, and it can. And AI will actually find. It can search other channels and see what they're doing right and then compare it to what you're doing wrong. It's like this channel's doing this right. That's why they're getting more views than you. Right. And these exercises are good. You should spend more time on what you're doing wrong than what you're doing right. In my opinion. And just focus on Improving it every time. Now let's say that this was the most useful report you've ever seen, right? Let's say we go through it and we're like oh my God, like we need to do this more often. One thing that you can do is you'll notice here we have this automations tab. I want you to do this every single Friday actually. Every. Yeah, let's do every single Friday for the videos of that week. I want to know what I'm doing wrong. Please create an automation. Oops, I don't know why I didn't pick up on that. Oh, it's because I have my AirPods
- [30:15]
A
in
- [30:19]
B
create automation 9am on Fridays. And so it really is that easy to create an automation. You just type in that you want this to happen. This is how I normally set up automations. People sometimes want to create an automation before they do it. It's just like do it once and then tell the agent to do that every time on this day. And that's a kind of a good workflow. And you can see here that it shows created weekly Startup ideas podcast Fridays at 9am if you click on it, it opens up right here and it shows you the status, which is active, which is good. The next run is on May 1, 2026. The last run is nothing because it hasn't run before. And there you go, it tells you what project and then you'll notice Here there's now a 2 on automations and it shows up right here. And if you want to test it, all you have to do is come up here and hit run now and it will run that automation and that's how you can kind of test to see how good it is. And then if you notice that something's wrong, you can press edit and you can edit the automation and I recommend doing that a few times. Like whenever you're setting up an automation you want to make sure it works and that's how you set up these recurring tasks and some recurring tasks that I'm seeing a lot of people have in the AI space or just people in business is they just have AI analyze their Slack and their email like twice a day and it sends them an email summary. So they just, they get it very concise and they all have links to relevant chats or emails. So it's really easy to navigate. I found that to be a really useful one for me as well.
- [31:57]
A
And you know you one shot at that negative report. So if I let's say I'm not happy with the output for Whatever reason to refine that. How do I do that?
- [32:08]
B
Yeah, I mean say like, hey, by the way, the report you gave me was too long. I want more concise put and make it a slash. And by the way, this document is a skill by default that is called word docs. So I just say make it a word docs instead of markdown and then, and then you can see, I'll use the doc skill here. I'll do two things. Convert the markdown into concise docs and update the Friday automation. You see, I was gonna have it recreate the doc and then I would say, okay, now update the automation. But it decided that it was just gonna do it for me, which is cool. Which is actually not how I would wanna do it. And one thing is in skills you can always include references with skills, so or in, in your automation. So you could actually include the document that it creates inside the automation. You can say, hey, it should look like this document, right? And you can give it an example. I always tell people that really good instructions are good, but giving it a really good example is great, right? You give an AI one good example, it's amazing. You give it five good examples, it's just going to do a great job every time. Time. That's kind of how I think about it.
- [33:36]
A
So is there the concept of an eval on output, meaning an eval, basically valuation, short for evaluation. Basically this is a 9 on 10 output, 10 on 10 output and it keeps going until it gets good output. Is there that concept within codecs?
- [33:55]
B
I mean you could set it up yourself. This is GPT 5.5. It'll run for as long as you want it to run. And so you could just tell it say like, hey, I want you to evaluate it. The problem is, I think in that case it would probably be useful to have examples of really good outputs so it knows what you want. You know the AI, they struggle with these subjective tasks, right? Coding is very, you know what a good output is because the app looks good and it fully works. So it's easy to verify whether it's good or not. But a report, you know my startup ideas podcast. So this is another problem that Codex has. Oops, that's kind of annoying is like the bullets are all weird. Like so this is one thing, this is another example of something you can fix. Like fix the bullets, their weird formatting. What was I saying before the weird formatting?
- [34:59]
A
Before the bullets completely threw you off, you were saying those bullets are. I don't blame you. Those bullets, the bullets Will throw anyone off.
- [35:10]
B
Yeah, yeah. Honestly. But yeah, you can just create these, these. I don't remember what I was saying.
- [35:19]
A
You were saying. I don't, I don't remember. The bullets really took us for a left turn.
- [35:27]
B
We're talking about skill.
- [35:30]
A
I think you're talking about the right outputs. Examples.
- [35:35]
B
Yeah, yeah, yeah. The number one thing that you can do in a company right now, and we're talking to a lot of companies who are trying to implement AI, is just collect as many good examples as possible. This allows your agents to have examples that they can reference, store them in a knowledge base that's easy to. To find if you store them in like a notion database, for example. AI is really good at going to a notion database, scouring the notion database, finding the right document to use as the example. And that way you can just kind of give it all of your documents. As long as you keep it organized, it's going to be able to use it pretty well. And so collecting examples is a really useful thing. And I've noticed a lot more companies are even recording their screen a lot more in order to capture workflows. Because I've heard rumors that the future versions of OpenAI, their newest model and version of codecs, you're going to be able to upload videos to it, and so it can watch the video, see exactly how you do it, and then it will be able to control the computer just like you. And that's going to be a whole new world. And so I'm kind of preparing for that world because most of my workflows as a creator or like I'm in marketing, I. A lot of my workflows require using an interface or using a tool like CapCut, like Canva, and AI will pretty soon get to a point where I can do that.
- [37:00]
A
And just a side note, you know, if that's the case, that, you know, these super apps are going to allow recorded videos to be uploaded to them. You got to think that the value of recorded content is going to go up in the AI age. So just from a startup idea perspective, there's a lot there.
- [37:23]
B
Yeah. I mean, and why do you think you hear rumors of Meta and Microsoft employees now required to record their screen and do certain tasks? And it's kind of like it's become a meme in that a lot of people who are anti AI where like, oh, when they start recording your screen, it's over, they're trying to replace your workflows. And. And yeah, it is kind of scary in a corporate setting but, like, if you run a small business and you want to accomplish more with like a small group of people, like, you want to be able to offload some tasks that you wouldn't normally do to an AI that can just kind of handle these tasks. Let me see here. Did it fix it? There you go. So, yeah, you can create any sort of document in here. Recording your screen will be big.
- [38:17]
A
So the number four negative, if you go back to that doc. One of the things that I get wrong in this podcast, apparently, according to Codex, is my interviews are way too soft. So it says my guests, get a friendly Runway. There's not enough pressure on failure modes, cost security, churns customer acquisition, who should not use the product. So I'd like to take this opportunity to take the Canadian out of me and not be so friendly and ask you the question with respect to Codex, you know, who should use the product? Cost securities, churn customer acquisition. Like, you know, what are the hard questions? I'm not asking that. I should be asking about Codex.
- [39:01]
B
You know, what I respect about you is you got feedback and you are applying that feedback 10 seconds later, which I admire. That's great. Who should use the product? I think that if you're constantly. I think content creators should use this platform. I'm a content creator first. That's how my brain thinks. I'm constantly creating lead magnets which require vibe coding. And if. And yeah, I find that to be super useful. And I'm also creating documents. So if you're creating websites and documents, you might as well do it in the same platform. Because ultimately, like, the websites that I create have to do with the documents that I create. I don't want them to be in separate platforms. I want to be able to create a document, do research, outline my thinking, and then convert it into a landing page seamlessly. And that's what you can do inside Codex.
- [39:56]
A
So
- [39:58]
B
as much as I want to help you give good feedback, I think there's a reason why these companies are racing to create the super app is because they think that everyone in Knowledge Work will use these platforms. And I truly believe that that's where we're going. I'm not saying you need to use Codex. I have no vested interest in Codex. I'm not an investor in OpenAI. I've never been paid by them. I just like their platform and I think it's the best right now. And I see a clear path to it being the best over time. So that's why I'm trying to prepare now, because it's only 10% as good as it will be maybe once these AI gets really good at controlling a browser. That's what I'm really excited about. So for all those reasons you should use it. But again don't use it if you don't want to.
- [40:45]
A
Anything else on this platform that's just a must know in terms of getting the most out of it.
- [40:49]
B
Honestly, you know what, I maybe should have started this but, but so yesterday in my video I posted this video. I wanted to test browser use and so I had it make this chess game. And then I said, I said ok, use at browser use and play yourself in chess one move at a time like earlier. So what you're going to see here, hopefully if Codex operates properly. And by the way, we're using 5.5 on medium intelligence which I will talk about that right after. But let's see if this AI plays itself using the browser. And I think this is just like a good way to kind of get an idea of like how the browser use works. I recommend just having fun with it at first. I think people try to do like oh see, you can see the mouse. You see that? So it's literally playing itself in chess as both teams. And look at how fast that's like working, right? Like you know, a few months ago this was way slower than this. I don't know if you tested any of the browser use stuff, but this is like pretty seamlessly playing.
- [42:09]
A
Yeah, it felt like with Manus for example, it was really cool that browser use was even possible but it really felt like it was on a dial up modem. And this feels like it's broadband 100%.
- [42:25]
B
Yeah, it just checkmated black. So white just checkmated black. White wins by checkmate. I played it through the in app browser one move at a time, alternating white and black. Final result, white wins by checkmate. And here we can see the full sequence. In my video it did this in one prompt. I said build a chess game, then use browseruse to test the game and play yourself until one team wins. And that's what it did in one shot, which I mean that's pretty insane. We're desensitized to it now. But that's insane. If you think about where we were even a year ago and then you think about where we will be a year from now. You know, it may just create its own canva a year from now and then use that. It might be like oh, it'd be way faster for me to just create my own tool because Canva doesn't have this little feature and it can't add it. Like, it might as well just make its own feature and then add it and then control it.
- [43:20]
A
So are you not using Claude code at all anymore? This is like your daily driver.
- [43:25]
B
Okay, two things I forgot to cover. That's one of them. Command J opens the terminal, type Claude. I have it set up to automatically run in dangerously skip permissions mode anytime. Like I can say, redesign the chess app to make it the anthropic vibe. Boom. Congratulations. You're using Claude code inside Codex and you're using it with the Claude subscription. So you are getting, you're benefiting from that same, the same savings. Another reason why you want to use these models in these platforms directly, like in Claude or in Codex, is you just save so much money because they're subsidizing. Like you get thousands of dollars of tokens for your $100 per month or $200 per month plan because you're using their platforms. So like, for me, there's no really good argument on not using these platforms, if I'm being honest, because you get so many tokens as a, as a user and you can also just use Claude code directly inside Codex and it'll work. So that's one thing. Also, another thing is, you know, Theo and some other people have been talking about how they like to use 5.5 on low and medium effort for most tasks, right? If you ask it to make a simple change and you use extra high, it'll often like really try and put a ton of effort into it and it might go off and like do unintended things. And so that's one thing to look out for. I'm not very good at discipline enough at doing that. I end up just on extra high, just running it over and over again. But that's what I've been hearing other people say. And you can also, if it feels like it's going too slow, you can turn it to fast. Just know that this will be more expensive. You'll run out of credits faster.
- [45:17]
A
I also think it's worth mentioning that 5.5 is an expensive model, right?
- [45:22]
B
Yes. If you use it via their API. Let's see here. Look at that. So COD code just changed the chessboard, right? And then you can just throw it away. The only thing is, if you use, if you use codecs in the app, you get that like spinny animation and you can see when it's done. That does not happen when you use it in the browser. They're kind of like separate things, so just keep that in mind. But, like, look at how easy that was. Now people like, oh, but Claude's way better at design. So it's like, just make the whole app and then tell Claude to design it. Like, that's, that's what I say. It's that easy. What did you just ask right before this?
- [45:59]
A
I was just saying that it should be noted that 5.5 is a more expensive model.
- [46:07]
B
Yeah. If you use it via the API, it's twice as expensive as the recent model, which is GPT 5.4. It's twice as expensive. But what you'll find is that the model is actually more efficient. So it'll get the job done and get the job done better with less tokens. And I think that's where I think the benchmark thing will be a year from now. It's not going to be how token efficient is it? It's going to be how much money and how much time does it cost to do a specific task? You know, the same way you would hire a contractor or a human employee. You don't think about it in terms of token efficiency. You just think about how expensive is it to get a task done. And so if you were. What this model does is it really thinks about your intent more and so it'll go more directly towards what it is that you want. And so you may find that it actually gets things done better and faster. But, yeah, it definitely can cost more money when using it on the API. I haven't noticed my credits going down twice as much when using it in the app, so that's hard to say. And then it's also 20% more expensive than Opus 4.7, so that's one thing to consider. So it is pretty expensive.
- [47:26]
A
You said that there were two things that you wanted to cover. One was terminal.
- [47:30]
B
One was the models, and the second was using Claude code in the terminal. And again, this is a full terminal. So if you're technical, like, you can do anything in here. And I think, yeah, I mean, you can create any type of document. I'm still learning this right now. So, for example, you can leave comments on the browser. This is similar to Cursor has this feature as well. So you can just comment directly on this and say, like, this should say Greg Chess. Chess. And what you'll see once you press enter is you see that one annotation enters the chat. Let me close out of this. And then you can add another one. You can say, like, the pieces should be bigger and you can just say bigger and so you can add three annotations and then just say make these changes. That's kind of how I do it for a lot of the design changes. And it's super useful. And so, yeah, I mean, this is a very complete product. You can also full screen it. One thing that you can do is you can full screen and then you can kind of message it here. It's kind of in a more immersive way to do some vibe coding tasks or if you want to just focus on the document that you're working on. I find this to be incredibly useful.
- [48:50]
A
Have you been using codecs a lot with Images 2.0?
- [48:56]
B
Yes. Launched 100%. So great question. Codex has GPT Image 2, which is the best image model in the world. And again, this is not like up for debate. I'm not, again, I'm really not trying to shill anything. You notice here that it's making changes and it's using browser use to test stuff. I didn't even ask for that, so just wanted to put that out there. Yeah, the GPT image model is actually insane. It's better across every single category and it's built into codecs and so you can ask it to create websites, you can ask it to create 10 different versions of thumbnails and codecs will just prompt GPT image 2 and then it can generate images and then it can put those images in a word doc, you can put them in a presentation or a website or anything. So, yeah, it's a super useful feature and I think there might be a skill for it. It's slash image or GPT. No. Okay. You can just ask it to use GPT Image and it'll just do it. And they also have Sora built in too. Like Sora is an official plugin.
- [50:04]
A
Yeah, but are you going to use Sora?
- [50:06]
B
No. Yeah, no. God, no. Sora. I don't even have it installed. I just, you know, some people out there, you know, they might want to create some brain rot videos or whatever they're doing.
- [50:16]
A
Yeah, I think the people that listen to this, like most people here are like, I want to start a company where I have a company and I want product market fit and I want to scale it and they're looking at how can I use this tool to just supercharge my productivity. And so I think that with respect to Sora, Sora is now delisted from the App Store, right?
- [50:43]
B
Yeah, it doesn't exist anymore.
- [50:45]
A
Doesn't exist anymore. So it's like it's interesting, it's fun, but I think the people listening to this are way more interested in using Remotion.
- [50:53]
B
Totally. But mind you, that's not like a reference to the app, the plugin, it's the video model. And SORA is not the best video model right now. So that's why it's not interesting. And I think that's what OpenAI realized. They're like, oh, it's going to take so much effort to create the best state of the art video model. That's what people want to use. So you can create a skill that connects to FAL to use the best video models and you can attach it to this app and you could have those videos show up inside the Remotion thing. You can put videos inside of the Remotion video that play in the video itself. So if you wanted to create a talking head video of you, I've seen you, you have used AI video before. I'm certain you could easily set that workflow to work directly inside codecs and you could add subtitles directly in it. I'm sure there's going to be a company that only builds like an AI video editor like Remotion that shows up inside codecs and they're going to make a killing. Right. That's all coming. And so you'll be able to just edit, create videos directly inside Codex because any interface that you can possibly think of that would be on the Internet will show up here because it's a full web browser.
- [52:03]
A
And with working with Falcon, you know what is the best video model? Is it CDance 2.0? Is there something else that you use?
- [52:13]
B
Yeah. For whatever reason, I've like completely lost interest in AI videos. Maybe it's gotten so good for me. I just focus so much more on like I find coding and agents so much more interesting than video models. I'm not up to date. I would love in the comments if someone could let us know what the current state of the art model is. I'm really curious. I've heard Seed, Dance, Cling and a few others are really good. I'm curious where Google is. Have you heard anything about Veo? Like, is that still a good model?
- [52:44]
A
Yes.
- [52:45]
B
3.
- [52:45]
A
Was that the latest one?
- [52:46]
B
I think it was the latest. Maybe like 3.1 or something.
- [52:49]
A
Yeah, I mean I had PJ ASA on, I don't know, three, four months ago and the stuff that he was able to show me in terms of creating commercials using VEO 3.0, I believe was incredible. Haven't heard of A big launch from Google since then. But I, I loved, I loved what, I loved what he showed me. I also played with Sea Dance 2 last week and I was also very, very impressed.
- [53:28]
B
Yeah, I always try it and I'm like, this is awesome. And then I'm like, what am I going to use this for? For me, I make educational content. I like talking. It's hard for me. And I've just been using Remotion a lot because very similar to you in the first one to two minutes of my long form videos, I use graphics to kind of outline the. You know, I saw, I watched one of your Open Claw videos recently and you have a really cool graphic of like Open Claw with a spotlight coming down on the Open Claw. It's just a very good opening graphic and that's how I use, I use it as, as overlays over my long form content. That's what I'm excited about. I just have been using it not at all for, for AI generated video. But also PJ is a legend by the way.
- [54:08]
A
I love that. Absolute legend. I think a part of what, what's daunting for a lot of people in AI in general is they hear about something like Codex and they're overwhelmed with how do I actually make this a part of my life. I just had coffee right before this with a friend of mine and I was like, dude, are you on, you know, Codex? How are you using Claude? How are you using Gemini? And he was just like to be honest, I use notion and Notion AI and like I ping pong with Notion, you know, ChatGPT and Gemini sometimes, but that's like basically it. And I was like, well, what's stopping you from setting up some automations and creating some skills and this and that? And he's like, I'm just overwhelmed and I don't know.
- [54:56]
B
Yeah, yeah, I mean Notion, first of all, Notion's done a great job. Honestly. I, you know, I think they're. I don't use Notion much anymore because the platform got a little too overwhelming to onboard other people onto like if you don't know how to use Notion, it's becoming a little annoying to like teach people how to use Notion. It's, it's kind of overwhelming a little bit. But they have a really good agent sales system built in and if you want to go deep in Notion AI and learning how to build agents, do it like, you know, it's great. I just personally want to create other types of documents and I find codecs like very soon you'll Be able to connect. Like Notion will just show up in the browser because right now you can't sign into anything and have it stay signed in. It's not persistent, but I know for a fact that they're releasing a full browser and so there is a Notion skill already. You'll be able to control Notion and so you're going to be able to use these models for cheaper inside codecs to use AI inside Notion. To me, it doesn't make much sense to use the agents within Notion. It makes sense to use the agents within Codex and then use the interface of Notion if you like it. That's how I see it.
- [56:09]
A
But if you say you use Notion as your second brain, I'm sure there's a plugin, a Notion plugin that connects, right?
- [56:17]
B
Yeah, absolutely. Plugin Notion. I have it set up. I use Notion for one specific thing and that is for all of my videos. I create one Notion page and I drop in. You can actually upload videos to it and I can comment on parts of the script and include attachments on the videos. It's a niche use case. So I use Notion and I set it up with one click. And if you actually go to Notion, actually, we don't need to get into this, but it's super easy to choose what Team Spaces has access for codecs. So when you go to Notion, you can very easily set up permissions and you can only give it access to one database if you want. If you're worried about it messing with your Notion or privacy concerns, you can give it access to only specific databases, which I find really useful.
- [57:10]
A
I want to end on this note. I want to end on. For someone who's just getting started with Codex, if there's three or four things that they should do once they've got it installed, what would those for things be?
- [57:27]
B
It's very funny. And by the way, we didn't plan this. I put this right before the podcast. I was like, if there's projects to get started, to test, I actually wrote them down and we covered one of them. I think the first thing that you should do is you should have fun on all these platforms. I think the worst thing that people do when they. When they show up on the platform is to, like, expect to get productivity out of it right away without kind of like play around for 30 minutes, take an hour to just do fun projects, and that this gives you an idea of, like, the possibilities and then hone it down and get very specific and get disciplined. And so I think one of the Best things that you can do is make a game and have browser use play a game against itself. Just use the browser use, test it. It'll get your brain thinking about where this is going in the future and how this will be useful in the future, which may actually change how you set up your documentation now for all of your different tasks. Like maybe it'll, maybe you can start preparing for it now because this is not a year from now, this is like three months away. AI will be able to control any browser like as good of a human. And this is something I can't convey by explaining it. You have to try it out. So that's why I recommend trying it out. Another thing that you should do is tell AI to do in depth research and say, put it on max effort and say, I want you to research this topic and put it all, organize all the information you find into a spreadsheet, then when you're done, turn that into a document and presentation and just get the sense that you can create a in depth research report. You can put it in a spreadsheet, you can create a document and a presentation, right? So once you do these two things, you now understand browser use and all the types of documents it can create. And then I said make a 3D simulation. This one's more just for fun if you want to. You can also try and build a mobile app in Swift. If you have a company website and you're like, oh, you've always wanted to create a mobile app, you can literally download the code to your web app or website, upload it and say, hey, I want you to build a mobile app in Swift for this. For now you still need to have a Mac and you need to have Xcode downloaded, but it can one shot mobile apps like full mobile apps. There's a few things that you need to learn, but you'll be mind blown at what it can create in one shot. To the point where it's actually gotten scary for me. I'm like, oh my God, every app that can exist will exist. That's how good AI is getting at creating mobile apps. And then the main thing, after you've kind of played around, like I said, list out all the things that you do on a day to day basis. Pick the most annoying thing that you do every day and then try and do it with computer use. Or just try to create a workflow that accomplishes it. Maybe you use plugins and then tell your agent to turn it into an automation. Just tell your agent to turn it into an automation and if you can create one useful automation in a couple of hours, that might be worth it. And so those are just some projects that you could get started with right now that will give you an overarching view of what this tool can do. And I think by then you'll be like, okay, I understand.
- [60:30]
A
And by then you'll be like that Riley Brown guy. He was right about that Codex thing.
- [60:34]
B
Yeah. And you might have doubt now. You can only convey so much information by talking or even showing you got to feel with AI, I think to a large degree you have to feel it. You do these things. If you're still not sold, message me. I don't think you'll exist. I don't think that person will exist. Just, you gotta try.
- [60:53]
A
I'll include links for where to find Riley in the show, notes in the description. Riley, just put my phone number.
- [60:58]
B
I want a direct line.
- [61:00]
A
I'll include Riley's phone number. Actually, be careful what you text Riley because he's got Chronicle set up. So, you know, whatever you text him might just show up on the Internet.
- [61:11]
B
It goes straight to Sam Altman.
- [61:13]
A
Yeah, Riley, thanks so much for coming on the show. Do you have one last word of wisdom to leave people after this Codex masterclass?
- [61:26]
B
Honestly, just get out there and experiment. Right now is the best time to experiment. And what I've noticed, being in San Francisco, I've noticed that the people who have been, like, really successful creating startups are people who just like, want to experiment and tinker. And they don't have so many people who are like, type A want to get an immediate benefit after 30 minutes of using a tool. And I don't think that's where the benefits come. I think the benefits come from tinkering, messing around with tools. And when you get. When you have this weird inclination of, you're like, okay, wait, what is a model? Why is this model different than The Deep Seq V4 model? It's like, go down that rabbit hole, use AI, learn about it. And then through that process, you'll learn the things you need to learn to start or grow your business. I think all of the magic in AI happens in these rabbit holes. And then once you find something, just go all in on it. That's kind of my final advice. And I think Codex is a great kind of surface to do it on.
- [62:24]
A
You're totally right, by the way. I lived in San Francisco for almost nine years, and I think it's not just so much that the tinkerers win, is that people, the tinkerers who aren't afraid of looking dumb win. Right. So I think when you're trying some of these tools, you might feel dumb initially, and it's because you aren't a master of it yet. But the only way to become a master is to actually put in the time and work. So I think the reward is infinite. Right. The ability to use these tools really well, like Codex, and be able to create. For me, this isn't just a tool that's helpful for content creators. This is helpful for literally anything that you can dream of. Lead magnets. Use one example. Lead magnets. Content, of course, but ideas for products that you want to create, businesses you want to create. Automating your personal life. Automating a dream team for, you know, a business that you're starting, just, you know, and everything in between. So, yeah, keep. Keep playing. Just stop listening to this. But, you know, also, like. And comment for. For us that, you know, I want. I want this episode to get a lot of love. So Riley comes back on, but. Yeah, get your hands.
- [63:54]
B
Come on, dude. This is my fifth time on the show, man. I mean, I'll come back on anytime, even if it gets 10 views.
- [64:00]
A
You came on when we were just a little show, dude.
- [64:03]
B
Honestly, there was an argument to be made that I put you on the map. No, I'm just playing the video. The first vibe coding video. It's funny. A lot of people bring that up. They're like, oh, I first saw you on Greg Eisenberg's podcast when I was struggling through that first episode.
- [64:19]
A
And that's why we do this, right? You know, I hope. I hope so. Yeah. Thanks again for coming on. And I hope someone learned, 1, 2, 3, 4, 5 things that just gave them not just the creative juices to actually use these tools, but the motivation to keep going. So thanks a lot, Riley.
- [64:37]
B
Appreciate you having me on. This is awesome.