Ep 829: ChatGPT Voice is Like Jarvis: How to use the New Feature and the 7 biggest unlocks

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life.

Many people talking about the new ChatGPT voice are calling it their AGI moment, and I don't disagree. Yeah, it's changed how I work in the short amount of time that it's been out, but there's usually at least a handful of updates every quarter that do that. But this update from ChatGPT Voice changes how I think about work, and that has rarely happened in the nearly 4 years I've been doing the Everyday AI Show. It's like if you've ever seen Iron Man, ChatGPT Voice is literally like having your own Jarvis. Whether you're in front of your computer or across the country like I was the past couple of days, you can ramble hands-free in natural language and the desktop version of ChatGPT will literally just control itself. It'll do anything on your computer that any skilled human could do, but you don't have to be there doing any of it. And I really put this through its paces yesterday as I took about an hour-long walk in controlling my home computer. Through my iPhone and my computer at home was creating documents, using other desktop programs like Cloud Code, organizing my cloud storage, triaging emails and drafting replies, and planning this very show by transcribing YouTube reviews and say bookmarks that I just didn't have time to get to, creating images, all that stuff. I just gave commands and the work was just done. So obviously OpenAI can't call this thing Jarvis mode, but you can see how that that might be a more fitting description than the moniker of ChatGPT Voice. But I will say this, I've used literally thousands of AI tools and features over the past 4 years of doing Everyday AI, and ChatGPT Voice has instantly become the number 1 most useful ever. So what is this ChatGPT Voice thing and why is it so special and why haven't you used it yet? Well, we'll break all those down as we put AI to work on Wednesdays on Everyday AI. All right, here is the big picture. ChatGPT Voice became an agent management interface. So this just came out. It hasn't even been out a full week since OpenAI announced it, and it lives inside of the desktop version of ChatGPT Work and Codex, which is the same thing, by the way. So ChatGPT Voice is powered by the new GPT Live model. A little confusing, but we'll break it down. And that somewhat confusing rollout, I think, made for this being the undisputed most criminally underhyped release in AI ever, right? Sometimes you see these like little releases come out and everyone's talking about it like, oh my gosh, this thing's gonna, you know, break how we all work. It's gonna take away all these jobs. It's gonna do all these things, right? ChatGPT Voice, I don't think people are talking about it, right? I'm very ears to the ground when it comes to these kind of things and people should be talking about this and using this more. And that's why we're gonna be going over it on today's show. So stick with me for the next 25-ish minutes. And here's what you'll learn. You're going to know how the ordinary voice mode differs from the new ChatGPT Voice launched inside Work or Codex. You're going to know which evidence and approval rules that you need to keep authoritative sounding updates accountable because you can't just trust the voice and never go back to it. And I'm going to give you these 7 advanced ChatGPT Voice uses that you're going to want to start using and apply to your workflows. Immediately. Let's get into it. Welcome to Everyday AI. My name's Jordan Wilson, and well, this thing's for you. This is your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me not just keep up with AI, but how you can use all of these updates. I tell you what matters. You go put it to work, grow your company and your career. So please, if you haven't already, subscribe to the podcast on Apple Podcasts or Spotify and make sure you go to our website at youreverydayai.com. We're gonna be not only highlighting the big takeaways from today's show, but also giving you all of the other AI news you need to know to stay ahead. All right, so let's get into it. So on Wednesdays, we put AI to work on Wednesday. So the latter half of this will be kind of demo-based. So FYI, if you do want to see the video version of this, it will obviously be somewhat visual because if everything goes correctly, which for an unedited, unscripted demo of a super smart AI, taking over my computer. Who knows what could go wrong, right? But you can always catch the video version of this on our website at youreverydayai.com or on LinkedIn, YouTube, Twitter, machines. All right. But let's first start with the two basic settings that you have to understand. It's kind of settings slash, you know, getting your mindset right on what's named what. And this is important. Because there's a lot of confusion on this because most people are like, okay, well, yeah, ChatGPT voice, I know it's, it's not new, right? I can just, I've been able to talk to ChatGPT for a while. So yes and no. I essentially want to take you through 4 different phases or buckets of using your voice to interact with ChatGPT. All right. So number 1, dictation, that's been around for a very long time. Literally, some people think that that's what this is. That's not. So dictation, right? You can click the little microphone button. On the ChatGPT interface and it will— you can talk and it's just going to turn your speech into text. That's dictation. That's not what we're talking about. Then there's the old voice mode, which was technically called Advanced Voice Mode. There was a different voice mode before that. All right. But that was more of the walkie-talkie, right? That's the turn-by-turn conversation mode. So that was what, you know, a month ago, that's what OpenAI had. That's what Claude has today. Nothing wrong with that, but that's not what we really want out of a voice mode. What we want is, well, the actual engine driving this, and that is GPT Live. So that is what is called this kind of full duplex voice mode. So that means, and you'll see an example of this, the voice mode GPT Live can talk and listen at the same time. It can, as it's talking to you, it's using another model in the background to call tools, look up information on the internet. You know, maybe it's running a little bit of code to give give you an answer to something that's not as easy to find as you might think. All right, so that's that natural interruptible conversation. It's just like talking with a real human, but not like walkie-talkie style, right? So not to be confused, but that's not what we're talking about here, right? That came out 2 weeks ago, and maybe you've already used that, and it's really good. This is different. This is ChatGPT Voice. So this is essentially that voice model, that full duplex GPT live voice model that has been available in ChatGPT for about you know, 2 or 3 weeks now. Now that lives inside of ChatGPT Work or Codex. So this is the version that you download on your computer. So there is no ChatGPT Voice when you go to chatgpt.com. I know that's confusing. That's going to be using the GPT Live. So ChatGPT Voice is available on the desktop version. All right. And that's kind of where you're, the thing that's Jarvis mode, right? Where it can literally access all of your data. It can use your computer. It can do literally anything that a human sitting down in front of your computer can't. Whereas, you know, the normal GPT Live, the new smarter thing, it's cool to chat with, right? But it's not gonna actually do your work. Now we all finally have a super smart, uh, you know, coworker sitting down that we can literally just talk to like a real person. All right, so that's kind of the, the iteration of the different types of voice inside of ChatGPT. All right, so basic setting number 2, which I kind of talked about it. So ChatGPT Voice itself is just the GPT Live model plus bringing in this agentic execution layer. So aside from it being able to actually complete work for you, which the GPT Live can't really do, at least not fully, right? The difference here is it has access to all of your data and connectors, which GPT Live doesn't, right? So that's the one thing that is missing from GPT Live that I was using it, right? If you want to say, hey, check my email, or if you have some MCP servers connected inside of ChatGPT, you know, it can't access those. It can— you can upload a file to GPT Live and it will be able to go through and talk about that file, but it can't access all of your dynamic data. As an example, I probably have not a ton, right? Because my tech stack is not super fat, but I probably have at least 30 different either connectors or MCPs. So now with ChatGPT Voice on the desktop version of ChatGPT Work or Codex, I can instantly talk and grab data from any of those. But also I can output or have the agent output work for me, but I can just command it all via Jarvis mode., right? So here's how it works. All right. You, uh, the, the ChatGPT voice can listen, it can speak and coordinate all at the same time and actually do your work. All right. So there is a little button, right? So if you hover over it, it just says start voice chat. So not to be confused, uh, with the button on the left, which is just the microphone button, which is the dictate button. All right. So you're going to start a voice chat and this little orb, right? This little ChatGPT orb. I don't know if there's an official name for it. I just call it the, you know, you can call it the voice visualization or just the orb. So I'm just going to call it the orb.

All right.

So that's when this kind of ChatGPT voice mode is working. So also a little confusing. That's the same orb that comes on if you're just using GPT Live inside of ChatGPT.com. All right. So you just got to be using this if you're trying to follow along at home, which I encourage you to do. You know, make sure you're obviously using the desktop version of ChatGPT Work or Codex. Which again, a little confusing, is the exact same app. There's just a dropdown at the top because people are always like, wait, I don't have Codex. Yes, you do. It's ChatGPT Work. And people are like, I don't have ChatGPT Work. And it's like, yes, you do. It's Codex. It's the same thing. All right. So you have to be doing this inside of that download, the downloadable program on your computer. And well, it's just neural. It's going to be able to listen and run these models in the background and literally do anything that you can do on your computer. So there's multiple layers and kind of the biggest differentiator that I think versus ChatGPT Voice and GPT Live, the voice layer just talks. Now we have this new layer that we're combining it with all of the capabilities of Codex is it just is going to work, right? It's going to execute your work for you. And just for ease of not saying the wrong thing too many times from here on out, I'm just going to say Codex. All right. But it's the same thing as ChatGPT Work, right? But there's too many ChatGPTs flying around. So I'm just going to say Codex, right? That is the version that you download and run on your computer.

All right.

So let's quickly go into the 7 advanced ChatGPT Voice uses and then we're going to go straight into the demo and hopefully it all works. But I'm going to be doing some of these advanced use cases as we go along. These are also kind of best practices as well. So Number 1, and this is a difference in how these threads work. So you cannot go and pick up an old thread and start a new ChatGPT voice conversation in there. All right. You can't do it. Unfortunately, that would be great if you could. Um, instead, uh, you are going to, uh, start a new chat. Okay. So whether that's inside of a current project Or if you just go down and start a new chat that's not attached to a project. So, um, you can start any new chat, um, by just clicking the, uh, start new voice chat button. Um, and that's how you start, but here's the thing, right? It knows everything and it controls everything. All right. And that is just. I can't emphasize how big that is, right? I've done this many times. I've done live demos of it. I just checked again, right? Code, uh, let's, let's talk about Claude Code, right? Claude Code doesn't know anything within what's happening under its own roof, right? Unless it just happens to commit something to its memory, um, which it just does that haphazardly. Otherwise it knows nothing, right? So all of those conversations, you know, let's just say that you're talking about Claude Code work. Right? All of those conversations, unless it just commits something to memory, which, you know, maybe it does 1 to 2% of the content in there. If you start a new chat, it knows absolutely nothing, which makes something like Claude Desktop, in my opinion, extremely useless. Right? That's why I don't really use it. I mean, I use it to audit things that I build inside Codex, right? Uh, but the big thing with this new voice mode is it can understand any thread anywhere. All right. So you can direct it to, hey, go look in this project, right? Find this project folder so it can research any thread, see what was done, understand it. It can send a message to any of those threads. So it is just absolutely crazy how a voice mode, because the old advanced voice mode couldn't do that right now. You can. And this is again, Combining the power of Codex, uh, with this technology, right? So I'll just walk you through how I was using this on my walk this morning. I know the name of most of my projects inside Codex, and I just said, hey, walk me through every single major thread, um, in these, you know, project folders one by one. At the end, um, you know, give me a brief update and what next steps may be. If you're not sure of a next step, suggest one for me based on, you know, what you know about me, what's in my Agent's MD file for that folder or for, you know, Codex in general. So it knows my overarching goals and it knows the project goals. And then it's going to go in there. It's going to look at all that context and then just make me make decisions, right? This is why I think on my walk, sorry, yesterday, I was more productive in that 1 hour on my phone just talking in my AirPods than I would have been probably in 3 or 4 hours doing the exact same thing in front of my computer. I talk about this all the time. Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on GenAI. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft, and NVIDIA have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use gen AI. So whether you're looking for ChatGPT training for thousands or just need help building your front-end AI strategy, you can partner with us too. Just like some of the biggest companies in the world do. Go to youreverydayai.com/partner to get in contact with our team, or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on GenAI. The internet is made to distract you, right? And even for me, it can be distracting working in Codex, right? Because I have all my browser tabs open inside Codex. So something about just the combination of rambling and being able to control any of these threads and just having that real focused time. I have so much done right in that 1-hour walk. But anyways, kind of advanced tip number 1 is know that you can with your voice any new chat can understand, read, write, and do anything to any thread. So it doesn't really matter for me from an organizational perspective. I would really like to just be able to go into a certain thread and pick it up from there, right? I have all these, uh, Chief of Staff threads inside of each project folder. That would be preferable for me because I like things to be a little organized, right? And then you have this chat that's down there in your recents, but overall it's okay. Because when it does make changes, it will actually go in there and send a message in those chats. So it's not like it gets lost. And I do have another advanced tip later that can help with this. All right, so, uh, number 2, GPT Live can't use connectors and Codex with ChatGPT Live can. So I already referenced this, but this is huge because GPT Live by itself, it's great, um, right? It's more of I think, an advanced party trick. And I do hope, and I know the team has alluded to this online, that they are going to eventually bring connectors to GPT Live. But at least for me and how I want to use, you know, ChatGPT, I need it to be connected to those, you know, 30 or so, you know, connected integrations and MCP servers, right? If not, it's just, you know, a better way to browse the internet. Um, which is fine, right? But I really want to use this to get work done. And that is why this combination inside Codex is amazing. And I don't know why I literally didn't see anyone talk about this or reference this, right? I've, you know, I follow the chatter and, you know, I've watched a couple of videos of people giving their opinions on this and no one is mentioning this, right? This is the biggest unlock right there to be able to go and go through your entire inbox, to go through your, uh, you know, everything in Slack and to be able to talk about it and just say, yes, send this. Yes. Draft this. No, save that to my Google Drive, right? Just to be able to go through all those things is— this is what people have been wanting for so long. This is the Jarvis, right? Hey, what's important? Go look in all of, you know, go look in my CRM, go look in my, I don't know, ERP, alphabet soup, whatever you want to say, my Drive, my Linear, whatever programs you're using that are connected inside Codex. Go through all those, triage those. Ask me what you want, you know, what are the most important decisions I need to make? And if you have read-write access, right, you can just start making those decisions and Codex can actually just go do the work in those programs for you. All right. So that's number 2. Connectors is a big unlock. All right. Number 3, you can schedule, right, the same way that you would inside of ChatGPT. So yeah, this is another one of those things that you can't overlook. Just that you're using voice mode. You can't overlook just the normal power of ChatGPT, right? So one of the things, if you use ChatGPT online or the ChatGPT Work online, is you can have repeatable tasks. So even when you're talking, right, I did this a couple of times. I'm like, ah, I'm gonna keep missing this thing, right? So, you know, I said, hey, you know, every, you know, Monday at 7 AM, you know, run this task, this thing that we just did together that, you know, me and ChatGPT just collaborated on via this ChatGPT voice that needs to be run, you know, every Monday morning or every Friday afternoon or, you know, every day at 4 PM. This, what we just did, let's turn that into a skill and then let's have this be a scheduled run. So another big thing. So, you know, as an example, maybe every single day you want to go in through and see any new lead that's coming via HubSpot. You want to enrich that data with Clay and then, you know, use the Kondo MCP. That's one I use to help manage my LinkedIn messages to see if you have a LinkedIn connection with those people. Right. So you can literally do that all via voice first. It's going to learn how to do it all. And then you can say turn that into a skill. Run that every single day at 9 a.m. and then you can literally, once it's done, go and talk and be like, hey, what happened here? Right? That process right there as an example, going into your HubSpot, seeing who's new, going in, enriching all that in Clay, and then, you know, seeing which of these people, right? If especially if you're in B2B enterprise sales or SaaS sales, right? You know, and you have a big LinkedIn network, you know, like I do, that's something that takes a lot of time to do all those things manually, right? That might be every single day. I don't know if you're an SDR, BDR, That might be a third of your day. It can be done for you, right? Set it up once in ChatGPT Voice, set it up as a schedule, and then every day you can also be like, hey, check in, walk me through, gimme a game plan on this. And you know, in a 10-minute ChatGPT Voice conversation, you've already done what would've normally taken you 3 hours and 10 minutes. All right, let's go next. Number 4, stack multiple tasks at once. Yeah, you don't just have to give one, you can just kind of ramble on and just keep giving more and more assignments. You know, as you are going back and forth, remember, sometimes ChatGPT will interrupt you. You can interrupt it, which is actually fine because that's how a normal conversation goes, right? Normal business conversation. Sometimes you just interrupt each other, right? You might not do that in a big business meeting or if you're meeting with the president of the company. But if you're working with colleagues or people on your team, you probably interrupt each other all the time because if you see someone's going down the wrong path, you don't necessarily want to give them 30 seconds or 3 minutes to keep going down that path. You're like, oh no, no, wait, hold up. That's not what I was talking about. Let's, let's rewind a little bit, right? You can do the exact same thing, but also just stack those multiple tasks and the voice will just only return when it needs more information from you or when it has updates. All right. The big unlock here, use your iPhone. My gosh, but it's a little tricky because remember, all right. Codex doesn't have an app. It is in the ChatGPT app. So confusing part here. If you go into the ChatGPT app and you're looking for that little button I told you to push, that's just going to launch GPT Live, which is not what you want. So you're going to go onto your iPhone. You're going to click Remote. Then at the bottom of the screen, you're going to click that, that button to start a ChatGPT Voice. All right. So don't just open your ChatGPT app on your iPhone and click the Voice chat button, you need to first click remote because that's essentially then at that point, and this is what I was doing today, that's how you are literally on the phone, right? For me, quite literally hands-free. I had an AirPod, you know, one AirPod in my ear and I was literally controlling my entire computer working on multiple projects, right? Going through all those folders and it was downloading, uploading, opening programs, changing things, doing the things that I would normally be doing. Doing it all hands-free, just giving, uh, you know, my, my Jarvis instructions. All right. All right. Number 6, a spoken update is not proof. All right. And I talked about this earlier. You should require a durable artifact. What does that mean? All right. When you go back, especially, right, and this is something I really found yesterday morning, um, I was looking back through the, the, um, transcript of what I said and I'm like, wait, how did we get this done?

Right?

Because sometimes all you get is what you said and what ChatGPT said. So, you know, one thing I would encourage you to do in this voice conversation, or, you know, you can leave this in your agent's MD file inside of any project folder, is just say, hey, always require written evidence when you're done, right? Or, you know, build a, you know, an outline every single time that you complete a task, you know, make sure you outline what was done. Or maybe you have a folder a file, you know, whether it's locally, etc. Every time you, you know, complete a run, complete a task, complete a project, upload this file and say exactly what was changed, what was done, where the file location is, etc. That's huge. You always need written evidence because, yeah, it's nice to be able to just flow with voice mode, but sometimes you need that data. You need that data, right? Maybe you want it in a markdown file. Maybe you want to upload it to your Google Drive. Maybe you just want it inline in the, in the thread inside Codex, right, as a bullet point., but you need that written proof. And then last but not least, and this one's really cool, another one that I haven't seen many people mention, but there's literally on the desktop version, if you click that little orb, you can drag it. All right. And you don't even have to have Codex kind of active or open on your screen. This is one thing I'll hope I'll be able to, uh, to kind of demo. Um, but all you have to do is pull the little orb. Away, right? Which it's kind of hidden, right? You don't even know that you can do it a lot of times, right? One thing I do when there's new features, I just click around because, you know, OpenAI is known for, you know, hiding little fun Easter eggs. You know, if you're talking about like Codex pets or anything like that. But yeah, I just clicked and I was like, huh, you can drag this thing, right? And then I saw maybe one or two other people mentioned this. But yeah, for the most part, you can just have that little orb just chill on your desktop and you have this omnipresent agent that's just going to chill. It's going to do absolutely nothing until you tell it to. Right. So one little, little, I guess not cheat code, but best practice for this, especially, right. I like to listen to music a lot and you can't really do that because it's listening for any word. Right. So I will just have an AirPod in listening to music. And then on my Mac, I just, you know, set the output to my AirPods so I could still listen to music or be on a call with a colleague, right? Something like that. And then I can still, you know, I can still mute the mic. But if it's just me in front of my computer, that's all I really need to do. I can listen to my music in an AirPod and use my, my, my computer microphone to then talk to the Orb, right? That sounds weird talking to an Orb. All right. But it just allows you to have that omnipresent agent. All right. My gosh, what could go wrong? Here we go. Let's go live. I'm not, I'm not going to lie. Sometimes, especially because I have a lot of more kind of private, you know, personal private information in my Codex, you know, sometimes when I'm like, oh, I got to share my entire Codex, I'm, you know, sweating bullets a little bit, but I always want you guys to see, you know, see how this goes live and kind of, uh, see how it works. All right, let's try this live, live stream audience. If you could let me know if you can see my screen. All right. And hopefully we'll be able to hear the audio as well. So, uh, as a reminder, this is the, and yeah, if you do want to see the video version of this, if you're listening on the podcast, in the show notes, there's always a link to the episode page on our website and you can go watch the video version. So just skip to, you know, I don't know, about 26 minutes or so.

Okay.

So at the bottom here, you'll see there's a normal microphone. Don't click that. That is the dictation. The other one is to start a new voice chat. All right. So let's just go ahead. I'm going to go to make this an unattached chat thread, and then I'm just going to riff a little bit here. All right. So if you hear me talking, I might hit the pause, the mute button so I can keep talking to you all and explain what's going on. But otherwise, assume that I'm just talking from here on out to ChatGPT. So let's get it going. I'm going to actually work on a project that I wanted to do because I put a lot of work into a recent speech that I did. And there are some things that happened in there that I'm like, huh, I should probably pull that and do, do something more with that speech that I did. So let's go ahead and get started here. Gotta love the live demos. I've literally never seen this, this screen come up before where it said voice chat couldn't start. Oh, that's funny. All right, let's try it again. Uh, hey.

Okay. What would you like to try again?

All right, cool. No, so, uh, can you look, it's maybe in my downloads folder, but, uh, maybe from like 2 or 3 weeks ago, um, I did this presentation for a firm and, uh, there's probably like a PowerPoint. It was about, um, ChatGPT work. Let me know if you can find it.

I'll check your recent presentation files for that. All right. So ChatGPT work PowerPoint in your downloads so you can open it from the file card on your screen.

Oh, cool. Great. So can you just kind of go through and bullet point what I talked about and what I didn't? And then while you're doing that, can you also open Chrome and then go to YouTube and find the livestream that OpenAI did a couple of weeks ago for the ChatGPT work announcement. It's probably like 2 weeks ago or so, and then copy that link from that OpenAI livestream and then use Computer Use and open up an app on my computer. It's called Parrot. It's an app that I built. And then I want you to take that YouTube URL from the OpenAI livestream, and I want you to paste it into Parrot and create a transcription. And I want you to look at that transcription, which will tell you, you know, what they talked about in that 30 or so minute livestream. And essentially, I want you to look and find what I did not cover in my Affirm presentation that was covered in the OpenAI livestream. So just take, take your time and start working on that.

I'll get started on that. I'm comparing your Affirm deck with the OpenAI Work Launch stream now using Chrome and Parrot for the transcript.

All right. So I just put this on mute, so you'll probably still hear ChatGPT. Oh, it's not on mute. All right. There we go. I think, I think I'm on mute now. I'm not, I have no clue what's going on. Okay. That's, that's weird. Another time. Um, this, it's, it's, it's showing the transcript here, but, uh, obviously I'm talking to the livestream audience now, not ChatGPT, even though it's still, uh, transcribing. So, uh, hey, Chat, are you still listening to this? Okay, interesting. Love, love doing these, love doing these, these live streams here. 'Cause I always run through and then do the demo and then I always get these weird things that happen that didn't happen before. So it was muted, but it was still dictating what I was saying, which didn't happen in all of my testing anyways. So let me explain what I did and what's happening now because it's already opened up this program that I created called Parrot. Right? Just this awesome program that I, you know, built up over the past however long. So I told it, go find this presentation I did 2 weeks ago. I don't remember what was in it because I literally had to build it, like rebuild it the night before because it was right when OpenAI announced ChatGPT Work and I didn't remember if I got everything from the livestream into that presentation. And I thought that maybe at some point I'll do an entirely new course or something like that on Codex or ChatGPT Work. So I'm like, well, I put together a lot of work in this presentation and I thought it was pretty good, but there's probably some things I didn't get to. So what's already happened is Codex went in or ChatGPT Voice inside Codex went in. It opened up this program that I built It grabbed the, the URL and it went in and pasted it. It looks like it has the transcript up and we'll see what's, what's happening here.

The parent transcript is captured, so I could ground this in what they actually covered. Here's the quick take. Your deck focused on the Desktop Agent and AgentsMD workflow. The stream emphasized connected workflows and shareable outcomes. Those connectors across Slack, Calendar, and Excel to PowerPoint.

Can you maybe just give me, can you maybe just give me bullet points? Can you maybe just give me bullet points of the biggest things that I did not cover in my Affirm presentation that was in the livestream?

Sure. The biggest things you didn't cover were connected work across tools like Slack, Gmail, and Calendar, end-to-end finance workflows like Excel to PowerPoint reporting, interactive sites for dashboards or internal tools, starting tasks on mobile.

That's a good point. I think that there was a new update a couple, like, uh, maybe it was Thursday, but I don't even think OpenAI put anything on their website about it. It was something in Sites. So see if you can look up maybe their changelog for Codex or sorry, for Sites. I think it was something about there's some new analytics. Can you look that piece up?

Yeah, I'll check that.

Also look in my transcripts for Your Everyday AI because I, you know, every Friday and Monday we go over kind of like new features and new updates. Uh, so can you just check this, uh, the last week's, um, shows? Cause I think there were a couple other, uh, Codex updates in there. So you can just, you know, open that up. Or I think there's a folder, um, on my desktop if you want to, or not my desktop, maybe in documents if you want to use computer use. Or you can just go to our website and look at the transcript there. Um, cause sometimes they only update their, uh, kind of changelog like once a week. Um, so we obviously put out shows, you know, every day. So just go, um, specifically on sites. I want you to get more information on that.

Got it. I'll check the official notes, the sites docs, and your recent transcripts for anything site-specific.

Okay. And while you're doing that, um, or maybe first, can you just, uh, create a very simple, um, document and then just download it and save it to my computer, uh, that just shows kind of what I talked about in the Affirm presentation. And then what I did not cover and just kind of download that to my computer, but also save it to my Google Drive just so I have it. And then, you know, name it so it's easy for me to see and understand what's in there too. Got it.

I'll put that together and save it both places.

Okay. So I realize what I did wrong now, guys, earlier when I'm like, wait, it's still catching me speak even— no, that was a human error. I accidentally turned off the speaker instead of the microphone. So that was my fault. So right now I have both muted. So, you know, ChatGPT is still talking to me, but it's not super important. You know, essentially, I just wanted to show you those capabilities because what just happened, even though I kind of fumbled through that demo quite a little bit, right? This is something I was going to do because I knew I put together, like I said, a lot of good information for this Affirm presentation, just going over some of the features. You know, I had some good visuals, some good content, all that stuff, right? So why would I start from scratch? But I knew that there was new things that needed to be updated. I didn't have 35 minutes to go rewatch the, you know, ChatGPT livestream. So I just had it literally use its built-in browser, go find the YouTube channel for OpenAI, find that livestream, take that link, it opened up a program that it wasn't even pinned to my desktop, so it just found it. It opened up the program that I built, it transcribed it, and then it compared kind of everything that was in that PowerPoint that it opened in its file browser to the transcript from the OpenAI livestream. Now, right now it's putting it together and it's already— it just told me that it's put that document together and it's done. So that's good. And it also said it confirmed the missing sites update. So, all right. I do want to show maybe one or two other things quick. So I'm going to start a new voice chat and here we go. Hey, can you use the Beehive MCP and look at my poll from yesterday? And tell me what the results from the poll were.

Yeah, I'll check yesterday's Beehive poll results quickly. All right.

So I just muted chat, but what I'm showing for our livestream audience here is I'm actually minimizing— I'm going to minimize Codex here. And you'll see here kind of this little orb floating around on my desktop.

OpenAI's new Jarvis Live Mode is leading with 32. Claude Opus 5 Hands-On has 21, and 6 people picked other.

Okay, uh, could you just start that over? I missed the beginning.

Sure. Yesterday's AI Working Wednesdays poll has 59 votes. OpenAI's new Jarvis Live Mode is leading with 32. Claude Opus 5 Hands-On has 21, and 6 people picked other.

Uh, cool. Can you tell me, um, what my open rate was, um, yesterday, uh, or sorry, on Monday, July 27th, versus the Monday before that? I just want to see kind of what the open rates were over those last ones.

Okay, I'll pull up both Mondays and compare them.

So essentially what I'm doing right now, I just muted both, is, well, I got rid of Codex and now I just have this little floating orb here, right? So I'm dragging it around my screen. I can put it literally anywhere. It's showing me, even though I have it muted, what the open rate was. It says this yesterday or 2 days ago, it was about 55%, and the Monday before it was 58%. That's still because, you know, opens keep trailing in a couple of days after. So pretty cool. You can see right there some of the basic things. So I went through showing you multiple tasks at once, kind of stacking those using different connectors and MCPs, right? I used my, you know, Google Drive. That's a connector. I used an MCP, which is a model context protocol for my newsletter because it doesn't have an official ChatGPT connector. I, you know, did things like browsing the web, you know, pulling, you know, opening other programs on my computer. And, you know, it was clicking around, you know, you probably saw it. It happened very quickly because computer use is very fast inside GPT-5.6-SOL. So that's a wrap for the demo portion. It didn't go too bad, right? Maybe we'll see how this actually turns out. Hopefully it recorded all that audio, but Here's the takeaway. This is a completely new way to work, right? And almost a new way to think about work. Um, right? Because even before, right? Honestly, like even just 10 days ago before this update came out, for the most part, when I wanted to be the most productive, I was still sitting in my chair. And, you know, clicking around, right? Which is not bad necessarily, right? But distraction, overwhelm. And sometimes what I found is when you're not actually looking at a screen and when you're able to just kind of ramble and yap, right? Sometimes you get much better results because you can communicate things sometimes in a much clearer way. Because sometimes, especially if you're typing, right, not everyone can You know, especially if you work in an open office, you can't always ramble, right? Or if you're working on something private, maybe you work at home and you can't, you know, talk about certain things, you know, with other people around. This is something that for me, I, I'm not saying I've struggled to adjust how to work, but I've had to rethink what it means to work, right? And I had a great day working, right? I had a great day working yesterday. Went out and like I said, took an hour walk and it was not only being able to go get outside and touch grass here in Chicago, but I actually think that I got a lot more done in that 1 hour than I would've gotten done 2 or 3 hours doing something very similar using Codex or even ChatGPT Voice in front of my computer with all my screens because it's distraction. And then I would've said, oh, I, I have to go do all these other things, right? So it is definitely the new ChatGPT voice is like having a Jarvis. And at least for me, I'm, even though I've been Codex pilled since the day it came out, sometimes I still find myself still doing some of the work, right? And this has been a refreshing change in that because I'm actually just using my brain and thinking a little bit more, trying to think more strategically and creatively and, you know, being able to communicate things in a different way and, you know, being able to describe all the nuance and, you know, having ChatGPT Voice sometimes poke holes in my plan. So it's been for me a lot more talking and iterating and working through my voice, which for me is really easy and maybe it will be for you, but I hope this episode was helpful. Maybe this won't be for you, but there's a good chance that using the new ChatGPT Voice is, could be the biggest paradigm shift in how you've worked since maybe ChatGPT came out, or even more so than that. So, uh, thank you for tuning in. If you haven't already, again, please go to youreverydayai.com, sign up for the free daily newsletter. Thanks for tuning in. We'll see you back tomorrow and every day for more Everyday AI. Thanks, y'all.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.