You Have to Try the New ChatGPT Voice!

So one of my favorite and also most useful things about AI is voice. It's just the bandwidth. You can put in 5 times more than typing, no matter how fast you type. And this week we had so many voice updates, not just OpenAI reworking how their voice works from the ground up. We're going to explore that today, but also Claude releasing in their web application a proper voice model. Finally, that was like the biggest weakness of Claude. So we're going to have a look at all of that, how it integrates into the different products. All across ChatGPT, ChatGPT Work. It's even available in Codex. And it's just at a different level. It looks similar, but it's not. And I think a few little tips could change the way you interact with AI permanently. And that's what we're going to do today in this week's episode of News You Can Use, the show where we used to look at every news story and filter ones that matter. But now it's more like, what's the one thing this week that really makes a change and how you'll use this stuff? And then we look at other stuff on my radar in the end. Let's begin by looking at all of these voice functions that launched this week. So first of all, what released? Well, an overhaul to the ChatGPT interactive voice experience. You remember 2 years ago it went super viral when they had all of these demos of them speaking to AI. It was almost instant, felt fluent. They had a visually impaired person calling a cab and then using it to kind of navigate as it saw their camera. Get ready to wave it down. They never really fulfilled on many of those promises. It was super hyped. And then when we got the features, it was kind of nice. But most people that I know weren't using this interactive voice voice feature, this one particularly, this icon. So let's start by clarifying that there's a big difference between this one right here, the microphone, which is just voice input, or this one, which is interactive voice back and forth where the AI replies too. This is the one that got the big overhaul now. So where can you find or access this? Well, pretty much in every place that ChatGPT lives, as opposed to the Claude releases. We'll talk about that in a second. You can find it here in the web interface, chatgpt.com. I could just hit this. And you'll instantly see it's a bit different. Hey, what's up?

I was just thinking about how people decide what to work on next. What's been on your mind today?

And there you go. A few things to pinpoint there. As you see the interface, there's a chat below. So now I could be adding images and other things like so.

Yeah, you can drop images into the chat along with your text and I can work with them.

Just upload or drag and drop and tell me what you Like so, what do you see?

Got it, yeah. Thanks. I'll summarize the image first. It's a YouTube thumbnail that contrasts.

There you go. Additionally, there's some settings. So here in the top right, if you check out the intelligence level, instant is just the quickest and high is the smartest, but the slowest. As before, you can also pick the languages. I'm a Cove type of guy. You would just go through and select which one you like. It's lovely to meet you.

What would you like to Hey, good to see you. What's new in your world?

So this is where it begins. And this is the point at which you might think, okay, not that big of a deal. But I think where it's really interesting is the fact that this works in all other interfaces too. So let's get the next one out of the way, which is mobile. If you update your mobile app, it can do the same thing. And I kind of like this. I was testing it on my morning walk today. And you know, you just press the icon, it works the same way. But now you can upload images and even use your camera as you move through. So look, for example, I can say, hey, what do you see?

Yeah, that changes everything. Let me take a quick look. It looks like a retro interior corner, striped green and brown.

Well, yeah, so that's really nice. But here, and this wasn't obvious to me immediately, if you go to the three dots here under more, you can switch to either share screens or live video. So let's do that. Okay, switch to live video mode, confirmed. First time I'm doing this, by the way. All right, can you see? Oh, up here it switched to advanced. I'm going to switch.

Got it. Live video is on now. I see your screen on a ChatGPT settings page with a share screen option in the menu.

Okay, okay. So you see the phone screen, not the camera screen. Is that right?

Yeah, right now I'm seeing your screen, not the camera view.

Ah, I see. So live video is not available with this new model. If I want that, I need to switch to the old ChatGPT voice where which you can do here up top.

Checking.

The difference is that one is colored. The new one is black and white. So let's explore the last option here, which is more and then share screen.

Right, in most cases you'd open more.

And again, you will notice this one neither, this one isn't available with the new voice model. So you can only do camera and photos for now. Which does limit the mobile application. But let's get to my favorite part, which is the desktop app. This is the one where they really shipped. So what I'm talking about is the app that you need to download here, download ChatGPT for desktop. And once you have it, you just have the standalone application in your app bar here at the bottom. And they reworked this recently. If you're a little lost here, hey, you're not alone. I feel the same way. What they basically did is they now have the same lineup that Claude has, if you're familiar with that, if this helps. Cloud, it's Claude, right? That would be ChatGPT. Then it's Claude Cowork, that would be ChatGPT Work. And then the third one is Claude Code, which is made for building apps. They have Codex, which you can access by going here to ChatGPT and then Codex in here. So the desktop app contains all three levels now. Now you know how to navigate to them. And all three of these are empowered with the new voice function. So for anybody who hasn't tried any of this, I recommend just going into Work and then launching the voice chat from here. And this This is where the capabilities unfold. Hey, what's up?

I'm here and ready. What are we thinking about today?

All right, so I want to show your capabilities. And one amazing thing about this is that, hey, if I just move this out of the way, I'm just going to move this ChatGPT window to the side so you can still kind of see it, but you could totally move it out of the way. But what I want to show off is that if I'm doing my work, for example, if I'm on the site, this just works. There's no old model, new model, it just works. What do you see?

I get you. You can keep me off to the side while you work and just talk through whatever's on your mind. If you want, I can help brainstorm, explain stuff, or Okay, what do you see? One sec. I'm looking at your current screen now. You've got Chrome open to ChatGPT's download page.

Yeah, exactly. So it took a screenshot and it kind of just hovers there. Another thing I like is it kind of just sits there. So when I'm working and sitting here in silence, it's not just gonna talk, talk, talk, talk, talk my ear off. I can kind of just do my thing and have ChatGPT open like this. And whenever I need it, I can invoke it. Okay, how about a little test? I'm just gonna open up Slack here with a message that has been sent to me by Daniel, who helps with some of these YouTube videos. And he was asking me to set up a ChatGPT Work schedule task because we're making a short tutorial video on that feature. This was from Monday. And he basically gave me a prompt here and told me to set it up in ChatGPT Work. I have not tried this previously. I want to see if Chat can do that, right? Hey, this is ChatGPT Work. This is a prompt. Can you just set the schedule task for me?

Sure, I'll check that out. Okay, I'm checking Daniel's message to pin down the exact schedule before creating anything. I scheduled the hourly check. It stays quiet unless OpenAI or Anthropic announces a major launch, then emails a consistent alert.

Holy, it did that. Okay, okay, hold up. Let's pause this and check in the interface. I should see this here under scheduled. Yeah, Major AI Release Monitor running here. It uses the Terra model. I could upgrade that. I could change all of this. It runs in my default project, which is good. This is amazing. It just created a scheduled task automatically, which I'll delete now. But you can see the capability. You really get all of ChatGPT, including the new Work features, which are powerful. And by the way, if you're wondering what else Work can do, I put together a PDF with 10 of my favorite use cases. You can get it for free in the description below, first link. And all of that you can now command with your voice. This does feel like this Jarvis fantasy that we all saw and probably liked in Iron Man.

Wake up, Daddy Sean.

Welcome home, sir. I know a lot of people come into AI with those expectations. It's becoming a reality now. I have two more use cases that I want to show you. I think one of them is kind of obvious, but it's a good reminder that ChatGPT is just the GOAT at this. And now with this new voice feature, it's even better. Let's say I'm doing something new, right? Whether that's, you know, building something or maybe I'm stumbling my way through Codex. I just switched to Codex. And I'm like, gosh, this is a new interface. Where do I even start? I want to build, I don't know, Pomodoro timer app.

Totally fair question. If you're new, a good first step is to keep it small, define a tiny goal, like one timer.

But concretely, like, what is the next tiny step that I could take here? Can you assist me?

Sure thing. Let me check the quickest next move for that timer. The next tiny step is type one sentence, ask into Codex, make me a simple Pomodoro timer with start, pause, reset, and a 25-minute timer. Default. Once it generates, test it, then ask for just one small tweak, like a 5-minute break mode.

Okay, can you explain the entire interface here? What's going on here on the left with all of these folders and projects?

Sure, I'll focus on that left panel and map out what you're seeing. The left panel view is the Add Simple Pomodoro Timer task. I'm checking the interface context so I can orient you properly. Quick orientation: the left side is your workspace library. At the top, you can start a new task. Projects and folders sit below to organize related work, and the items under them are past tasks you can reopen to pick up where you left off. The main area is whatever task you're currently working on, like your timer. You don't need to organize anything yet. Just stay in this task and describe the timer you want.

So, Diego, I think that's a simple demo, but this use case of, hey, you're doing something new and now you're not alone with it because Chat is right there, sees what you see. And if you're confused, you're just like, hey, I'm confused. What's going on here?

I'll figure that out right now. You're in the Add Simple Pomodoro Timer task. Nothing's wrong. You're just in the chat where you ask for the app and get made.

Great. Okay, now I could mute my microphone down here or even mute the voice chat if I just wanted to completely turn this off for a while. And then if I want to continue, well, there you go. I just continue or cancel the thing overall. So you can see how this fluently could integrate into your daily workload. It doesn't need extra screen real estate. You don't need to learn a new app. It's just ChatGPT in a more fluent interface. Like, it's not an extra window you need to manage or minimize or something. It's kind of just there and doing its thing. And this is where the desktop app shines, because that's where you unlock these capabilities of it sitting over on the side, it accessing local files, it being able to remote control Chat, Work, and Codex all at the same time. And again, everything you've been doing with Chat will work with this. More fluently. I'm sure there's some caveats and there's going to be some fantastic comments below pointing out, hey, Igor, actually this one thing doesn't work. But kind of operating under the assumption that it does is the right thing because most of the stuff I've seen works. I think it has some issues with projects from what I've seen. Like you can't just go into your project, which I do all the time, start a new chat and then work from there. Hey, can you see the context files in here?

I can see the current voice chat, but no Paul is visible here. Do you mean a person, a message, something else?

No, I mean attached files in this project.

Got it, I'll check for attachments. I'm not seeing any attached or context.

Yeah, this is the biggest limitation that I found for myself. Every time it starts a chat, it's just a brand new chat and not inside of a project, unfortunately. Okay, there's one more thing that I want to show you with this before we talk about other stories, and that is the fact that it also works with connectors. So they're called plugins in ChatGPT, right? And I connected a few plugins, including my main calendar, and now I I can talk to it. And what I want to end on is one of my main issues is scheduling something. So what is coming up today in terms of events for me?

Okay, I'll check your calendar for today.

It takes a second. Fair enough.

You're in the video shoot until 2. Then you have Fidel from 2 to 4:30, a brief Mastery Hub switch at 4:45, and Asana cleanup with Stephanie at 5. After that, networking at 6, Editing sync at 7, 2 overlapping commitments at 8, and dinner at 9.

That's amazing. That's exactly right for today. And, you know, I want to schedule something for tomorrow, a 30-minute sync up with our YouTube team. Propose the best time for that.

Got it. I'll look at tomorrow and find us a good slot. I'm checking tomorrow for a clean half-hour window with your team. Tomorrow's best slot is 11 to 11:30 AM your time. It's open, it's before lunch, and it leaves room before your 3 o'clock slides review. I can't see the team calendars, but given yours, that's the strongest option.

Honestly, that's a great take. Now my main gripe with this is that I would want to alter its behavior and install certain rules on how to book things, but I'm not sure how to do that if it's just going to open up a new chat every single time. I guess I could use my account-wide custom instructions under personalization and then put them in here. They did up these recently to 5,000 characters now, which is more than triple what it used to be. So I guess I could put my booking rules in here. But ideally, I would want to create a project with some rules on how I book my events and context on recurring commitments. And then tell it, hey, every time I want to book something, look into that project file rather than accessing my entire account. This is really still the biggest weakness of ChatGPT because in terms of the capabilities, it's getting close to a lot of the more advanced and connected applications like OpenClaw and Hermes. Its memory and its context management is still very lackluster. But this voice function, the smoothness of it, the interactivity goes beyond anything I currently have in my own OpenClaw. I guess I could install the new models and go through the API and kind of like build it myself. But this just works. It's a perfect implementation. Like I literally want the same thing here with this floating box. And you too can try this today. Now, I did say that I wanted to talk about Claude too. And you know, Claude is amazing. You know, I'm a big fan if you've been following the channel, but their voice function was the worst in the market. Now it's actually pretty good. They made major upgrades to this last weekend. But the big caveat is this interactive voice that they now too have only works in a regular chat. Hey Claude, what's up?

Hey Igor, not much, just here and ready. That first bit came through a little scrambled on my end, but no worries. What's on your mind today?

Nice. Not bad. But as soon as I switch into Cowork, this option disappears. Code is the same thing. Desktop app is the same thing. So really interactive voice for Claude. It's now in chat. So in other words, they're playing catch up to ChatGPT. I will say that finally this normal voice input is amazing and available across all of them. So they upgraded this too. And yeah, you can speak into it and it just works. As seamlessly as the ChatGPT one did for the past 2 years. If you're not familiar, the Claude voice input used to be so bad that the only thing you could do is just joke about it. It got every, I don't know, 4th word wrong in my opinion. Maybe it's just my accent. No, actually kidding. It was bad for everybody. And there you go. Those are the voice updates this week. Pretty great stuff. Now to round this video out, I want to talk about 2 quick stories. One of them is Opus 5 that came out. I usually create a dedicated video. Last week I was traveling and also when I tried it, First, I wasn't that blown away. But now I'm like, wow, it's actually really good if you use Fable to orchestrate and Opus 5 to deliver. It's a bit confusing. That model on the benchmarks, it looks better than Fable in many of them. But when building things, I don't know, I was a bit confused by it. I've seen articles saying that, hey, you should just completely wipe the context when you're working with it. And I just want to be clear, what I'm talking about here is for people building with it. Everyday usage, it's fantastic. It feels very similar to Opus 4.8. But to build things, I was initially confused by it. I guess I still am a little bit. But what I've seen people do is, yeah, use Fable to orchestrate and then switch to Opus. I've been doing that too for the past 3 days. Works well. But honestly, personally, it's just not one of those releases where I went from Opus 4.8 to Opus 5 and I was like, wow, it's just so, so, so much better. It is better. But yeah, that's kind of my take on it as of now. Would love to hear yours. And then secondly, this is a really fun one. If you're in Cloud, there's a new connector that you can set up. Connector. And it's called Anthropic Economic Index. If you're not familiar, this is the research on how AI is actually being applied in the real world. Under managed connectors, you can just install this. And if you have context on yourself, DNA files, even if it's relatively shallow, you can just ask, what can an AI do for me based on the Economic Index top 10 by productivity? And it's going to give you a list of statistically proven best use cases for your industry, your job. And then of course, because it's Claude, it can write the prompts to run those things. This is probably the biggest hack to finding use cases for yourself that I have today. Yeah. And now you know too. All right. That's pretty much it for this week. Let me know how you're using voice or what limitations you hit. And with that being said, my name is Igor and I hope you have a wonderful week.

Stories you actually care about. AI news you can use. Igor's here to help you figure it out.