When ChatGPT Isn’t Enough, Open Codex

[00:00:00]Codeex feels harder than it actually is.

[00:00:02]Not because the tool is complicated, because the name sends your brain the wrong way. You hear Codeex and you think code developer, not for me. If you're new here, I'm Dylan. I run an AI consultancy. And right now, almost every coaching client that I have is asking me about codecs. And here's the shortcut.

[00:00:22]If you understand chatbt, you already understand most of codeex. All you need is a translation layer. and I'll give you that plus five examples of when it's worth opening codecs over chatbt. So let's get into it. There are many differences between chatbt in the browser and codecs on your desktop. But this one here is the primary one that matters the most is when you're using chatbt in the browser, you have to bring the data to the AI. So your files, the context, the prompts, everything. And when doing this, the AI has to hold all that context in its head at any given moment. And the more information you put into the AI's head, the less focus it has and the less likely it is to achieve the task that matters to you. Basically, the AI gets dumber over time the more information you give it. Now, with codecs, this is completely different.

[00:01:07]It's flipped. The AI comes to your data instead of you bringing your data to the AI. And it seems like a minor thing, but this is a critical because the AI's ability to maintain its intelligence and focus over time sustains. And the reason it's able to sustain that intelligence is the AI picks and chooses what data it looks at at any given moment. So it's not uploading the entire file in its head, but it's taking different segments of that file for the given task you've given it instead of looking at everything at once, which means the AI's head is less full and its intelligence is increased. There are many differences between chat between the browser and Codex in your desktop, which we'll talk through some of those. But what sits at the core is this shift where what we're doing is we're bringing the AI to the data instead of bringing the data to the AI. And this unlocks all types of valuable use cases, more complex use cases that people want to use for their jobs and their companies. Now, let's get a lay of the land of what Codeex looks like and also the questions you should ask yourself when you first open up this application. The first three questions you always want to ask yourself is where should the work live? That's the first question. Is it going to be in a basic chat or do you want it to be in a very specific folder inside your documents, your desktop or whatever else?

[00:02:13]After that we want to ask ourselves is how how hard should the AI think? So this really comes down to the level of reasoning. So is it light or is it extra high? After this we have the final question which is all around freedom.

[00:02:23]How much freedom do we want to give the AI? And this takes effect in permissions. So when you start out with codecs you probably should have the default permissions which means you're going to improve many of the actions the AI takes. But as you build more confidence in the AI you may take those restraints off and give it full access for certain activities. So those are the three questions we want to ask. Quick pause in your regular programming. This video is brought to you by me as always.

[00:02:45]So, two quick things. First off, Blow is a 30-day AI insight series, completely free. You'll get 30 insights in your inbox of how you can apply AI to your business and your work. The second thing is if you'd like to work with me, blower a series of offerings to see if there's a good fit from the two of us. Now, let's get back into the video. Now, let's actually look at Codeex and share some of the insights around the model choice, reasoning, and permissions.

[00:03:08]Okay, so here is codeex.

[00:03:11]As you can see, it looks very similar to chatbt. Let's actually do a side by side. There we go. We have chatbt on the left side inside the browser here. And then we have codecs on the right hand side on my desktop. They look very similar. They have input boxes. They have models to choose from, both in a similar location. They have a little plus sign for additional things you can do. They also have a dictation application. it it shouldn't be that overwhelming because it's very similar and how they're structured. So, let me show you some of the things I just shared previously. So, we'll start with permissions. With permissions, if I select this little orange thing here, this drop down, it shows me three options. When you open up Codeex to start, you're probably going to have two. It'll be default permissions and auto review. You'll likely want to start with default. That means that you're going to improve many of the actions the AI does. But, as you build more trust with AI, you can remove the restraints on the permissions. You can go from default to auto. Auto means the AI is going to be a little bit more free and you're going to have to approve less of the time. And eventually you can get to full access where you fully trust the AI and let it work autonomously. Now, a quick note on the permissions. You'll probably start with these two. If you want to have access to full access, the way that you have to do this, you have to go to your settings. So, if you expand this menu here, you would go down to settings. Once you select settings, you'll select settings again. And then once you're in here on the right hand side, you'll see I have permissions. I have default auto review and full access. If you want full access, you need to turn this on here. Once you've turned it on, you should have this option here that says full access.

[00:04:38]That's permissions. What about model choice? So on the right hand side, we have the drop down here where we have the models we can choose from. So here I have 5.5 and 5.4. Most of you probably should be using 5.5. In addition to choosing the model, you then can choose the amount of intelligence or reasoning.

[00:04:55]So this is basically how long the AI is going to think for a given task. We have low and extra high. It's important to note that each of these correlates to speed as well. So, how long it's going to take for the AI to respond to you as well as cost. Everybody pays for a certain plan with chatbt. And within that plan, you get a certain amount of usage. So, if you set this to extra high, it's going to use more of that usage as well as take a little bit longer. But is important to use high reasoning and fat 5.5 for really complex tasks. So, don't shy away from it just because it takes a little bit longer and might cost a little bit more. And finally, we have speed, which is standard and fast. So standard is self-explanatory. It's the standard speed the AI has when you ask it to do something. And then fast makes it go faster but also increases the usage. So it's important to note if you're impatient or you have to get something done quickly, you can do fast but it is going to cost more within the description you have. Now a quick note on how do I know how much usage I've used within a week or a given session.

[00:05:47]There's two ways to find it. One way is pretty easy. If you just go to settings here, it'll be a button that says usage limits, but I'm not going to open it here so I don't share my email. A more difficult way to find it just so I can show you what it looks like is if you're inside of the chat on the bottom here where it says work locally, you can see it says rate limits. When I select this, you can see I have 100% for the next 5 hours that I have 92% for the week.

[00:06:08]Those are the two ways it measures it.

[00:06:09]So you have a 5 hour limit and then you have a weekly limit. You need to keep an eye on this depending on your plan. If you're paying for the $200 plan, you're probably never going to exceed any of this. I' I pay for the $200 plan and I never get below 75%. And I use it all the time. But if you're on the $100 or $20 plan, keep an eye out on this, especially if you're using 5.5 extra high with speed enabled. So that was Codeex and we kind of walk through some of the small things involved and I'll talk you through a lot more features.

[00:06:33]But before we get into that, I want you to grab this prompt and drop it into Codeex just to see what it can do. If you've just gotten exposed to Codeex and you're trying this for the first time, take this prompt, drop it into a folder that has some files inside of it that are likely backups or duplicates that you're okay with the AI editing, and then just see how it performs. This is a very basic prompt. I'm just telling the AI to inspect the folder, tell me what you see, and then suggest one small task you can complete safely. It'll then look at the folder and you can see how the AI interacts with your folders and your files, and then you can improve its actions as it goes. And this prompt is boring intentionally just so you can get exposed to how the AI interacts with your files and your folders and your desktop. Now, after you've done that, we'll talk through the translation. So, I want to do a comparison, a sidebyside between Chatbt and Codeex and the primary features that matter most in both of them and how they're really not that different. It's just that in codeex they're probably two to threex more powerful and useful for most people in their businesses and their jobs. So here's our road map, our side by side and I'm going to walk through each one of these. I'm going to show you inside the UI as well as discuss it here on this so we have an idea where we're headed. So the first one is chats. We have chats both in chatbt and codeex.

[00:07:36]I've kind of already shown you this. So we have the chat box here. We have the chat box here. When you talk to the AI, it's basically going to go back and forth in a single thread as if you were in chatbt. The same thing and also the same name. So that's chats. Then we have projects. So I'm sure a lot of you have created custom GPTs and/or projects inside of chatbt. And these are simply tailored AIs to a given activity. So the AI is very good at doing that specific thing over and over and over. You and I both have many things we do on a daily, weekly, and monthly basis. This is a perfect opportunity to customize an AI to do that thing for us. Within codeex, it's the same exact thing, but instead of a project being by itself, a project is actually a folder on your desktop or your or your documents or really anywhere you want it to be because when you create a folder on your computer, you can open that folder through codecs.

[00:08:22]So, actually, let me show you on the lefth hand side we have projects, right?

[00:08:25]So, we have projects and we have custom GPTs. Let me do the side by side here.

[00:08:29]So, here we have projects inside of chatbt. You can create a new project.

[00:08:33]You kind of have a general idea of how these function. They have system prompts in the background that the I checks every single time and then some knowledge files that it can reference.

[00:08:40]Inside of codecs, you have projects as well. And a project correlates to a folder. So each one of these are folders on my computer and the names I've given them. What I did is I opened the AI through this folder. So anytime that I interact with this AI in this folder, it's going to potentially have a tailored activity to do within that folder. And here you can see if I expand these, these are conversations that I've had with the AI inside of this folder.

[00:09:05]New conversation. So each one of these is a completely different conversation inside the same folder. Now the question here is how do we create a system prompt for a project inside of codecs? Because in projects they have the instructions that you can add and you just basically add your prompt to the instructions. So the AI always looks at the instructions every time it interacts with you before it does anything else. The way that you do this inside of codeex is by simply having the AI create a file for you called agents.mmd.

[00:09:29]You can create this yourself, but it's easier if you just outsource it to the AI. And all you have to do is tell the AI, "This is what I want to do in this folder. Here's a specific outcome that I'm seeking to achieve. Now, I need you to create an agent.mmd file for me in this specific folder to achieve that."

[00:09:42]That's it. And a MD file shouldn't be overwhelming or scary either because it's just text. It's just text with a few hashtags put in different places. So that's how we prime the AI inside of codecs.

[00:09:54]After that we have apps. So in both chatbt and codeex we can connect our AI to other systems. So inside of chatbt I'm sure many of you if you go to more you've used apps before. So inside of the apps we have all these different applications that we can connect the AI to. Now the thing with these apps is they're sometimes useful and the AI does a pretty good job using them. But inside of codecs it's night and day. It is probably two to threex times more effective at using applications and connecting your AI to them than they are inside of the browser. And this really comes back to the point that I made previously. The AI manages its context more effectively, which means it can use a tool for longer periods of time. And so where do we find apps inside of codecs? Well, inside of codecs, they call them plugins. And a plugin is simply a bundle of two things. A plugin is going to have an app, which is what you know about. So this is basically the AI connecting to that thing. So this is going to be Gmail, Notion, Slack, whatever else. And then also often times plugins have skills as well. And this is also why AI is so much more effective at using these inside of codecs because it has a series of skills on how to use and connect to this app effectively. Let me actually show you what this looks like inside of codeex. So here we have plugins. When I select plugins, let me expand this out a bit more. You can see I have a series of options that I can choose from. These are all things that I can connect my AI to. Now, when I select one of these, let's say I I click uh Gmail, and in here you can see I have the app itself, the thing the AI is going to connect to, but I also have a skill, actually two skills. I have a Gmail skill, which basically tells AI how to utilize this specific app more effectively. Then also, I have a triage skill. Often times, many of these skills come with these plugins when you when you download them. But also, if you want, you can add these skills to these plugins if you choose. And a quick note on skills is I'm sure if you've used Claude or Cloud Co-work in the past, a skill is basically a series of steps the AI is going to take on a consistent basis. It's very similar to a project inside of Claude or Chatbt in the browser, but it's just a thing that you can use across conversations. Instead of you having to use it within a given folder, you can use it anywhere. So those are plugins, apps, and skills. The next thing we have here is schedule tasks and automations. So again, inside of chatbt, honestly, most people probably don't use this, but there is a scheduled task where you tell the AI to do something for you and it does it on a recurring basis. The reason that it's not useful is because oftentimes AI struggles to use other tools and or the access it has to the other tools is limited by just being able to read stuff instead of write stuff and that doesn't provide that much value to people. But with automations inside of codeex, again to the point that I made around here, since the AI is so good at using tools and it can read and write to the systems that you use, such as writing emails for you or writing to your CRM or things like that, you can have an AI set up a series of automations that works on a consistent basis. So you can say, "Every Monday at 9:00 a.m., I want you to send me a weekly briefing on what I have coming up," which is a kind of a hint of what we're going to talk about in the future. So where do I find these inside of Codex? So if I go to the menu, you can see right here it says automations.

[00:12:48]So when I select automations, they give you some examples to start with, but these are all technical, so you can ignore them. I'm sure they'll swap them out soon with other stuff. But to create an automation, it's very simple. You just go new automation. You give it a title. You give it a prompt. And in the prompt, you want to be specific on what you want the AI to do. Also, you can choose where that AI works. So we want to probably want it to work locally. And then we also can select the project or the folder it works inside of on that consistent basis. And then we tell it how frequently we want it to work. So every Monday at 9:00 a.m. we choose the model, we choose the level of reasoning and then we create it. That's it.

[00:13:22]Another big thing here is browser use.

[00:13:24]So some of you may have used Atlas browser from OpenAI or the Cloud Chrome extension where Claude can actually navigate your browser for you and things like that. I've used all the tools and by far the best browser use tool that I've seen right now is going to be the one that's inside of Codex. And I'll share a very specific use case with you later that no other model can do that.

[00:13:43]This one saved me probably six hours of time. Now, how do we get access to this?

[00:13:47]So, inside of codeex, if I go back here and go to a new chat, all you have to do is do at@ browser and you see that little plugin right there, it says browser. All I have to do is select that and then a blue thing pops up and it's going to use that plugin which includes a skill on the AI of how to use the browser. Now, I said the browser is inside of codeex. So, where is do I where do I find that? Well, on the upper right hand corner, you have this little icon here. When I select this and I select this little plus sign here, I have the browser option. So when I select browser, this is a live browser inside of codeex. So I can I can go to Google if I could spell it. I search Google on Google and it gives me a bunch of Google stuff. Now the cool thing here is that maybe I need to use Workday or maybe I need to use QuickBooks or maybe I need to use Google Cloud's console and I don't necessarily know how to use it.

[00:14:36]Well, I can have the AI navigate it for me, get the information that's necessary and give it back to me. And it's kind of trippy when you watch the AI navigate the browser as quickly as it does and as effectively as it does. And then finally after browser we have memory. So in chat GBT we have memory where the AI kind of remembers our preferences and what we care about. There's something similar in codecs but again it's much more powerful because we're not reliant on the AI's limits because with the memory inside of chatbt it can only hold so much in its head before it gets filled up and it gets stupid. But with codecs, the AI can write files to your desktop or different folders and it can reference parts of those files in the future. So it can have tons and tons and tons of memories for all types of activities. And when you're doing any given activity, it can search that specific memory, grab it, pull it in, and include your preferences or insights that have arrived in the past. So that's the side by side and the translation between chatbt and codecs.

[00:15:26]Now we want to ask ourselves when is actually worth opening up codecs. There are tons of examples as to why you would choose one over the other. So what I'm going to share with you are five examples that are relevant for the broadest audience and also are going to have an immediate impact on your productivity either individually or within your organization. We'll start with the first example and this is a common thing that I see a lot of my clients want to do inside of the browser but they can't which is you have the AI create maybe a beautiful dashboard or an Excel sheet for you and you're like okay well I want to update this dashboard or this Excel sheet every day every week every month. Well, if you do that in the browser with attribute or claude, it has to rewrite the entire Excel sheet or the entire dashboard over and over and over again. And that is both timely, it takes too much time, but it increases the chances of what the AI is going to mess up what it did previously. So, if it's rewriting the entire thing, it might make a mistake. But with Codeex, once it's created that dashboard, Excel sheet, PowerPoint, whatever else, and it's made that for you on week 14, it's inside of a folder. So when you have new data that you want the AI to add to it, all you have to do is drop in that new data in that same folder and say, "Okay, now I want you to update this dashboard or Excel sheet with this new data.

[00:16:27]Change nothing else. Just add this new data. Then you can do it again and then you can do it again and do it over and over and over. You can even automate it with automations. So this is a huge use case that a lot of my clients get value from. Now the next example is having the AI look through tons and tons of files, organize them, and derive the insights from them. So maybe you have a massive folder with a bunch of files inside of it related to a client or a project or something like that. Well, you can have the AI go through this. It can rename the files based off of what they actually are and organize them more effectively. Not just renaming the files, but also going into the files and removing duplicates, merging data where necessary and identifying edge cases that have been set aside that you probably should look at based off of its organization and its review. It can not just organize stuff for you, but also through the process of doing so, write out insights that you may want to know about. So a summary of what's changed and why, but also it could add lessons learned. So in the future, if it were to do something in relation to this data set, it knows that within this data set, you actually prefer the term account name for company name. So it'll use the same terminology that you use. And this is one of the big use cases for codeex because it can look at all your files, organize them, drive insights, and do a bunch of other things to them without you having to drop in one file at a time like you have to do inside the browser.

[00:17:37]Now, as promised, our next example is in relation to browser use. So, like I said, there's all types of use cases for this. The primary use case most people are going to get value from is if you need to get data from a piece of software that you don't really use that often or you don't necessarily know how to use at all. You can log into the software within the browser that I showed you inside of Codeex and you can tell the AI exactly what you need from the software and it'll grab it for you or do that task for you. Now, the example I wanted to share with you here is I was creating a sequence, an email sequence for one of my clients and we had to add basically 150 emails to that sequence. No other browser AI could do this. It would have taken me probably six to seven hours to do this. I dropped it inside a codeex and it did it in about 30 minutes without any intervention. I was truly impressed. Our next example is another common one. So say that you have to answer a hard question, something that probably includes many different notes, systems, and things like that. Well, you can ask the AI that question assuming that it has access to all the systems that it needs to pull the information from. So maybe this question includes some notes that you have on your folder. Maybe it has to research the web for something.

[00:18:36]Maybe it has to go to your inbox or maybe even your calendar. aggregating all this data and putting it into a one-page summary for you on helping you make that decision. And depending on the prompt you provided to the AI, it could also include sources to any of the facts that it pulled from that data so you can quickly audit to ensure that the AI is correct on what it pulled out from those documents. It also can include any assumptions that it made so you can double check its work as well as any open-ended questions that it's uncertain on that you can answer for it. So this use case is basically answering hard questions across many documents and systems. And then finally, one of the fan favorites is connecting your AI to your email and not just having it read your email, but also write emails for you as well as give you weekly briefings and updates. Now, the reason this is interesting is that sure, you can do a daily or weekly briefing inside of Codex or Chatbt, but that means you have to be inside of the tool to get it. With this setup, as long as you connect the AI and Codeex to the plugin for Gmail or Outlook, you can have it draft a weekly or daily update on what you did last week or yesterday and what you have coming up for the next week or the next day, and it'll then insert that as a draft in your inbox. So, you could be on the move, open up your phone, look at your Gmail or Outlook, and see exactly what you have coming up and exactly what you've done in the past. It's a quick win that's extremely valuable for a lot of people I work with. Now, these are just five of the examples. Like I said, there are many examples as to why you would choose one or the other, but these are the five that I think will be the most impactful for the audience watching this right now. So, let's do a quick recap on the three things that we talked about. The first thing is when you're opening up Codeex for the first time, there are three questions you want to answer. First off, where do you want the work to live? Do you want it to be inside of a folder or a general chat?

[00:20:06]Most of the time, it's going to be in a folder. After that is how long do you want the AI to think? So, you want it to think for a short period of time or a long period of time. That'll be determined on two things. Are you willing to wait and also are you willing to pay the cost because of the importance of the task. Next we have permissions. So this really comes down to how much freedom do we want to give the AI? How much trust do we want to put into it? I'd recommend starting out you start with default permissions. Once you build trust in the AI then you can give it full permissions. After this we talked about the translation piece. Most of what you use inside of chatbt is identical in codecs just way more powerful. And we mapped each one of those features side by side. And finally, it's important to know that chatbt is still really useful in the browser and you should probably use it a good amount of the time. But that's for quick work that's ad hoc and smaller.

[00:20:46]But for the really complex tasks that are sometimes more valuable for your job and your company, you want to consider utilizing codecs because you can outsource more that's more complex and there's a higher likelihood that it's going to achieve it effectively. And that's it. So, as a reminder, two quick things. First off, Blow is a 30-day AI insight series, completely free. You'll get 30 insights in your inbox of how you can apply AI to your business and your work. The second thing is if you'd like to work with me, blower a series of offerings to see if there's a good fit between the two of us. Now, here's the catch with Codeex. It runs one of the smartest models in the world right now.

[00:21:20]And if your prompts are still written like before, you're actually getting in the model's way. There are four small tweaks that change that. And I made a video on them right here, so you can go ahead and check it out. I'll see you next time, internet.