WEBVTT
00:00:01.360 --> 00:00:10.560
If you listened to our Monday AI News That Matters segment, I told you we were kind of in for a wild week.
00:00:10.720 --> 00:00:22.879
That's because after a somewhat slow-ish week last week, I said on Monday that this week was going to be a banger in terms of new AI features you can actually use.
00:00:23.359 --> 00:00:34.159
Sure, I knew ahead of time a few things that were gonna come out, but even I was kind of blown away just yesterday at the sheer amount of new updates that all dropped within hours of each other.
00:00:34.399 --> 00:00:44.399
I mean, no joke, on Thursday alone, we got huge new AI features from Google, Microsoft, Anthropic, and OpenAI in a matter of hours.
00:00:44.640 --> 00:00:50.000
That's why on Fridays we break down what's new in AI that you can actually use.
00:00:50.079 --> 00:00:51.920
And we do it in a very specific way.
00:00:52.079 --> 00:00:59.280
We give you the AI features that are available for you today, not rumors, leaks, or wait lists.
00:00:59.759 --> 00:01:05.040
These are instant AI upgrades that you should be putting into play today.
00:01:05.200 --> 00:01:08.719
And we've got a lot to cover on today's show.
00:01:08.879 --> 00:01:19.840
So stick with me for the next 25-ish minutes, and I'm gonna tell you how ChatGPT went full Jarvis mode and the new app shot secret that unlocks the future of work.
00:01:20.000 --> 00:01:27.599
I'm gonna let you know the new Claude feature that took a page out of the Codex playbook, but actually made it a little bit better.
00:01:27.840 --> 00:01:37.040
And I'm gonna fill you in on the kind of new Google agent that you can start using today, but with one caveat.
00:01:37.439 --> 00:01:38.959
All right, let's get into it.
00:01:39.040 --> 00:01:40.400
Welcome to Everyday AI.
00:01:40.480 --> 00:01:42.719
My name's Jordan Wilson, and we do this every day.
00:01:42.879 --> 00:01:52.560
It's your unedited, unscripted daily live stream podcast and free daily newsletter helping business leaders like you and me keep up with the avalanche of AI news and updates.
00:01:52.719 --> 00:01:54.879
I tell you what matters, how to use it.
00:01:55.040 --> 00:02:00.959
You take that information, poof, you're the smartest person in AI and you can grow your company and your career.
00:02:01.200 --> 00:02:05.840
So it starts here, but make sure you go to the website at your everydayai.com.
00:02:06.000 --> 00:02:08.080
We're gonna be recapping the highlights from today's show.
00:02:08.159 --> 00:02:14.319
So if you miss any of the details, maybe you're out on a walk and you want to go back and say, wait, how did this work?
00:02:14.479 --> 00:02:19.280
It's all gonna be in the newsletter as well as all of the other AI updates you need to know.
00:02:19.599 --> 00:02:21.599
All right, let's get straight into it.
00:02:21.759 --> 00:02:24.560
I'm a big fan of the Friday features show.
00:02:24.719 --> 00:02:27.520
I wish I would have started it like three and a half years ago.
00:02:27.680 --> 00:02:36.479
We've been doing it now for about six months, uh, but just for our podcast audience, I'm always showing uh on my screen some basic, you know, information from the company.
00:02:36.560 --> 00:02:37.680
You're not missing anything.
00:02:37.759 --> 00:02:39.439
So uh just FYI.
00:02:39.520 --> 00:02:41.599
But let's dive in live.
00:02:41.759 --> 00:02:49.840
I'm gonna be sharing my screen, uh, just kind of showing the uh windows for the different announcements that we're gonna be going over.
00:02:50.080 --> 00:02:54.879
All right, so let's start first with your health.
00:02:55.120 --> 00:03:00.319
Yeah, this is actually um one I'm personally excited about.
00:03:00.479 --> 00:03:09.520
I know some people aren't gonna be, you know, diving in headfirst necessarily to give their health data to anyone, let alone an AI company.
00:03:09.759 --> 00:03:14.000
For me, I'm absolutely wildly gonna be using this.
00:03:14.159 --> 00:03:20.319
So here's what's new ChatGPT Health is finally rolled out to everyone.
00:03:20.479 --> 00:03:24.319
It is launched for all US users today.
00:03:24.960 --> 00:03:38.400
And it's expanding the feature that was first introduced in January, but it was introduced as a wait list uh feature, and it connects your Apple Health and supported medical records directly into Chat GPT.
00:03:38.560 --> 00:03:39.840
So, what the heck can it do?
00:03:39.919 --> 00:03:40.639
Why is it useful?
00:03:40.800 --> 00:03:55.840
Well, it can compare your lab results, summarize changes since a prior appointment, track your medications, factor in your sleep activity and workouts, essentially anything that you know your Apple Health might track and all the different applications that connect in there.
00:03:56.000 --> 00:04:03.039
Well, it can bring in that data and anything else that you can update or upload into Chat GPT Health.
00:04:03.280 --> 00:04:10.800
So OpenAI says that health conversations obviously live in a dedicated space, siloed from all your other chats, right?
00:04:10.879 --> 00:04:16.959
So yeah, you're not gonna have all of that information popping up via like a memory uh into your other chats.
00:04:17.120 --> 00:04:20.959
So, like I said, this is already rolling out to US users.
00:04:21.040 --> 00:04:29.920
You do have to be 18 or older, but the cool thing is this is available on free plans, on the paid plans, uh, and also on iOS.
00:04:30.720 --> 00:04:38.800
So one small caveat though, free users, you're just gonna be using kind of the model that you have the most access to, which is GPT 5.5 instant.
00:04:38.879 --> 00:04:45.040
Um, and then for other paid users, you'll be able to use the more powerful GPT-5.6 sold.
00:04:45.680 --> 00:04:48.639
All right, so why is this useful?
00:04:48.879 --> 00:04:49.199
Right?
00:04:49.360 --> 00:04:56.079
Because right now, obviously, there's no one place for all of your health information to live.
00:04:56.240 --> 00:05:01.920
Like, I can't tell you the amount of conversations I've had with people about this over the last few years.
00:05:02.079 --> 00:05:06.480
And, you know, there was a company out there that was kind of, you know, AI native.
00:05:06.560 --> 00:05:08.240
Uh, they were called Forward Health.
00:05:08.319 --> 00:05:09.279
They are no more.
00:05:09.680 --> 00:05:22.480
Um, and I was kind of excited about that concept because I'm like, there needs to be, you know, something that's not tied to a single provider that allows you to just essentially bring in all your health records and to have a smart system that just knows all that.
00:05:22.639 --> 00:05:32.879
So obviously, there was nothing previously that would have stopped you from doing this inside of a large language model like Claude or ChatGPT, Gemini, Copilot, et cetera.
00:05:32.959 --> 00:05:34.959
But it wasn't really set up for this, right?
00:05:35.040 --> 00:05:40.560
So think of this as a specialized version of Chat GPT that's siloed from everything else.
00:05:40.639 --> 00:05:46.319
So it keeps all of your health data uh separate, but it is literally built for this.
00:05:46.639 --> 00:05:55.680
So, you know, right now your health info is scattered across different patient portals, EMRs, apps, and it just pulls it all into the conversational layer.
00:05:55.839 --> 00:05:59.839
And there's also some kind of practical prompts baked up for you, right?
00:06:00.000 --> 00:06:07.199
So how your cholesterol might be trending, or how you can summarize your blood work before an appointment, what you should ask your doctor tomorrow, etc.
00:06:07.600 --> 00:06:16.480
And that context kind of bleeds uh usefully into everyday tasks, like factoring a dietary restriction into your restaurant picks.
00:06:17.120 --> 00:06:19.040
So, who's gonna find this valuable?
00:06:19.199 --> 00:06:30.399
Well, this is obviously a consumer play, but I think this just goes to show where OpenAI has started to focus more recently, uh, right?
00:06:30.560 --> 00:06:37.680
Not necessarily just on the models, but how to make their models more useful for more people in more ways.
00:06:37.839 --> 00:06:52.079
So, you know, whether you are just someone that wants to get a little bit more serious about your health, you just have questions, or if you're managing chronic conditions, uh, if you're a caregiver trying to track something in your family health or just preparing for your upcoming appointment.
00:06:52.319 --> 00:07:02.079
So obviously, open AI says this is not intended for diagnosis or treatments, and health conversations aren't used to train foundation models.
00:07:02.399 --> 00:07:05.279
Um, so pretty big one.
00:07:05.439 --> 00:07:06.480
I'm excited to get this.
00:07:06.560 --> 00:07:13.920
I signed up for the uh the wait list previously, wasn't uh wasn't kind of uh selected to be part of that first group.
00:07:14.000 --> 00:07:16.800
So this is one I'm gonna be jumping into head first.
00:07:17.279 --> 00:07:20.000
All right, our next AI update.
00:07:20.160 --> 00:07:28.319
A new update, well, new-ish feature from Anthropic, just kind of borrowing from the codex playbook, but I think they made it a little bit better.
00:07:28.560 --> 00:07:32.160
So Claude has some new voice mode upgrades.
00:07:32.240 --> 00:07:33.519
Oh no, this is not the one.
00:07:33.600 --> 00:07:37.600
Sorry, that's a little bit uh better than uh ChatGPT's.
00:07:37.680 --> 00:07:43.360
So this one is voice mode, not quite as good, but also this one is a little different.
00:07:44.720 --> 00:07:47.360
Yeah, just so many new updates.
00:07:47.519 --> 00:07:51.040
Even I'm getting confused, and this is all I do every single day.
00:07:51.199 --> 00:08:02.480
All right, so this new update from Anthropic is essentially expanding how and where their voice mode can be used.
00:08:02.800 --> 00:08:09.920
So Anthropic has upgraded the Claude voice mode to run on Opus and Sonnet for the first time.
00:08:10.079 --> 00:08:23.759
That's because previously it ran only on their least powerful model, which is haiku, which is why for me, I absolutely never used voice mode uh inside Claude on my phone for that very reason.
00:08:23.839 --> 00:08:25.680
Uh, but now I probably will.
00:08:25.839 --> 00:08:32.720
The cool thing as well is you can switch models mid-conversation in voice mode now works with connectors.
00:08:32.879 --> 00:08:46.159
So that is the big difference right now with the um the one thing that's maybe a little bit better is that their upgraded voice mode works with connectors, but it's not bi-directional.
00:08:46.240 --> 00:08:51.759
So it's still kind of the quote unquote older, dumber uh version of a voice mode.
00:08:51.919 --> 00:08:55.279
So you you've heard me kind of rave about the new GPT Live.
00:08:55.440 --> 00:09:00.559
That's because GPT Live can listen to you um and speak at the same time, right?
00:09:00.639 --> 00:09:02.960
You can interrupt it, it might interrupt you.
00:09:03.039 --> 00:09:06.480
Um, and it's kind of listening, thinking, and doing all these things at the same time.
00:09:06.639 --> 00:09:14.639
So this is still the old school walkie talkie mode where you talk, you wait, Claude responds.
00:09:14.960 --> 00:09:20.000
But the big benefit here is it can use connectors, so pulling in your live data.
00:09:20.080 --> 00:09:26.320
So that's not something that GPT Live can do, although you can individually upload files into GPT Live.
00:09:26.399 --> 00:09:29.440
And OpenAI did say that they are working on bringing connectors.
00:09:29.600 --> 00:09:37.440
So here's who has access: you have to have a paid Claude account to get that upgraded Sonnet in Opus voice mode.
00:09:37.679 --> 00:09:40.000
Otherwise, you will be defaulted to Haiku.
00:09:40.320 --> 00:09:46.159
So free accounts do have a very limited usage and only a single connection.
00:09:46.320 --> 00:09:49.120
Um, and all prompts then obviously go through haiku.
00:09:49.519 --> 00:09:53.440
And right now, voice mode does support 11 languages.
00:09:53.919 --> 00:09:55.759
So, why is this useful?
00:09:56.000 --> 00:10:06.559
Well, if you want to be able to chat with your connectors and you have a paid quad plan, this is really good, especially if you don't mind kind of working in walkie-talkie mode.
00:10:06.720 --> 00:10:14.639
Uh, for me, I've really enjoyed kind of the uh the bi-directional or duplex uh new way of talking.
00:10:14.799 --> 00:10:17.200
It seems like that's the future of the voice interface.
00:10:17.360 --> 00:10:27.039
So as long as you don't mind kind of this more kind of waiting and holding and you know, only being able to talk or listen at once, not that bad.
00:10:27.200 --> 00:10:28.480
So, who's gonna find this useful?
00:10:28.639 --> 00:10:43.039
I mean, anyone that's a power uh user of Claude, if you're doing any hands-free work, commuters, you know, executives between meetings or just operators who want to think out loud, uh, pretty big update and release here from Anthropic.
00:10:43.679 --> 00:10:44.320
All right.
00:10:44.480 --> 00:10:50.639
Uh, next we have a yes, another new image mode from Microsoft.
00:10:50.879 --> 00:10:57.840
All right, so here's what's new in MAI Images or sorry, MAI Image 2.5 Pro.
00:10:58.159 --> 00:11:01.600
Uh, all these image models are always a mouthful.
00:11:01.759 --> 00:11:02.559
Um, right.
00:11:02.879 --> 00:11:13.679
We should just, I don't know, Microsoft get a fun like nano banana uh name, and then we can call it, you know, why not like Clippy, uh Clippy Pro, you know, 2.5.
00:11:13.840 --> 00:11:14.240
All right.
00:11:14.320 --> 00:11:18.960
So this is launched and it's launched in a lot of different places.
00:11:19.200 --> 00:11:34.960
So it is in Foundry uh preview, it's in Cop uh it's in Microsoft um Copilot PowerPoint, um, and it's in also a variety of other places, but the new image mode, it's pretty good.
00:11:35.039 --> 00:11:42.720
Um, I don't think we have a lot of benchmarks on it just yet, but I would assume that we get those probably uh within a couple of days.
00:11:42.960 --> 00:11:56.559
So this is Microsoft's newest, highest Vidal A professional grade image model, built, they say, for superior high-quality imagery, detailed editing, and precise in image text rendering.
00:11:56.879 --> 00:12:09.039
So this does uh Microsoft says handles text to image generation plus controllable edits, object removal, replacement in painting, text updates, all while preserving composition.
00:12:10.000 --> 00:12:11.440
Uh, so who has access?
00:12:11.519 --> 00:12:14.080
So yeah, it is now available in Microsoft Foundry.
00:12:14.320 --> 00:12:22.559
If you do have a Microsoft uh 365 co-pilot plan, you can use this right now inside of PowerPoint, or you can use it via the API.
00:12:22.639 --> 00:12:31.279
The pricing is five dollars uh per 1 million text input tokens or$8 per 1 million image uh input tokens.
00:12:31.759 --> 00:12:36.720
All right, and then uh$106 per 1 million image output tokens.
00:12:36.879 --> 00:12:38.720
So why is this useful?
00:12:38.960 --> 00:12:48.720
Well, I'll say this right now, if you're one of those organizations that can only use Microsoft Copilot and you can't use anything else, then this is great.
00:12:48.799 --> 00:12:54.080
This is a nice upgrade over the previous MAI image 2.0.
00:12:54.399 --> 00:12:56.879
Uh, but it's still presumably, right?
00:12:56.960 --> 00:13:00.159
We'll see what the benchmarks say, at least from my trained eye.
00:13:00.320 --> 00:13:02.399
This is still fairly far behind.
00:13:02.559 --> 00:13:10.480
Uh GPT Images 2, which is the best AI image uh you know model in the world, and also the Nano Banana 2.
00:13:10.559 --> 00:13:21.840
Uh, so those are kind of the top two, and then you have some uh some more uh you know Chinese open uh versions of these, but the best two still are GPT images and Google's Nano Banana.
00:13:21.919 --> 00:13:24.720
Uh so not really in the same tier.
00:13:24.879 --> 00:13:29.279
We'll see what the benchmarks say, at least by my kind of quick trained eye.
00:13:29.440 --> 00:13:31.759
I still think that there's probably a quality drop-off.
00:13:31.840 --> 00:13:45.360
But regardless, if you are someone, especially I think in um if you are building decks in PowerPoint via Copilot, being able to use this image generator in there is gonna be big, right?
00:13:45.440 --> 00:13:55.360
So yeah, I know that so many people out there are kind of living inside of PowerPoint, and maybe your visuals and the old clip art just won't do.
00:13:55.519 --> 00:13:57.360
Um, I think this is gonna be pretty big.
00:13:57.440 --> 00:14:03.600
So, you know, if you're trying to create product imagery, marketing visuals, brand assets, um big.
00:14:03.759 --> 00:14:07.600
It is big for those purposes when you can't use anything else.
00:14:07.840 --> 00:14:08.480
All right.
00:14:09.200 --> 00:14:11.679
Our next AI update you can use today.
00:14:11.840 --> 00:14:20.960
Yeah, it was this big of a week that I didn't even really honestly get a chance to play with a brand new model from Google Gemini.
00:14:21.200 --> 00:14:27.919
Aside from just some daily driving tasks, haven't really given the new Gemini 3.6 Flash a run.
00:14:28.159 --> 00:14:32.879
Uh, so here is what's new with Gemini 3.6 flash.
00:14:33.039 --> 00:14:41.759
There's actually three different models, but I say most people are focused on Gemini 3.6 flash, but they also unleashed uh Gemini 3.5 flash light.
00:14:41.840 --> 00:14:55.200
I know confusing, so they released both uh the 3.5 upgrades and updates and new variations in the 3.6 flash, which is the first of the 3.6 flash series.
00:14:55.360 --> 00:14:59.360
Even more confusing, the pro series is still stuck on 3.1.
00:14:59.519 --> 00:15:06.480
So yeah, now your latest and greatest Google Gemini models technically have three different step changes.
00:15:06.639 --> 00:15:13.840
You have 3.1 Pro, because we are still waiting for 3.5, or maybe they'll skip straight to 3.6 Pro.
00:15:14.080 --> 00:15:24.480
So you have 3.1 Pro, then you have the 3.5 uh Flash Light and 3.5 Flash Cyber, and then the only one on the 3.6 tier is Flash.
00:15:24.720 --> 00:15:26.879
So extremely confusing.
00:15:27.039 --> 00:15:32.639
They're they're all for different purposes, but let's break down at least uh what's new in these models.
00:15:33.200 --> 00:15:36.879
So uh Gemini 3.6 flash.
00:15:36.960 --> 00:15:44.000
Uh essentially not that much smarter per se, but it is cheaper and more token efficient.
00:15:44.159 --> 00:15:56.159
So Google says that Gemini 3.6 Flash consumes 17 fewer tokens than 3.5 flash while taking fewer reasoning steps and tool calls in multi-step workflows.
00:15:56.240 --> 00:16:03.200
Uh, then you have 3.5 flashlight, which is the cheaper and faster version of Flash.
00:16:03.279 --> 00:16:12.720
All right, so Flash is like your, you know, cheaper version, and then Flashlight is the cheaper and faster and more lightweight version.
00:16:12.960 --> 00:16:25.919
So the 3.5 flashlight is the fastest model in the 3.5 series at 350 output tokens per second, built for low latency, high throughput work like agentix search and document processing.
00:16:26.000 --> 00:16:32.240
And then you also have that third model, Gemini 3.5 cyber, which is tuned specifically for cybersecurity tasks.
00:16:32.480 --> 00:16:34.159
So who has access?
00:16:34.320 --> 00:16:35.039
Well, everyone.
00:16:35.200 --> 00:16:46.480
If you're a paid user, you can go into Gemini, you know, your Gemini.google.com, the Gemini API inside AI Studio, Android Studio, Anti-Gravity, Gemini Enterprise, uh, right?
00:16:46.639 --> 00:16:53.039
Literally wherever you're using Google Gemini for the most part, uh, you can find the new 3.6 flash there.
00:16:53.200 --> 00:16:56.720
Uh, Flash Cyber, though, is limited access only.
00:16:56.879 --> 00:17:07.359
Uh, that's the you know, government and trusted partners, similar to you know, a Mythos or OpenAI has their uh daybreak, I believe is that what their version is called.
00:17:07.599 --> 00:17:10.960
So, why is this useful and who should be using it?
00:17:11.119 --> 00:17:24.480
Well, if you are a developer, especially if you live inside of the Google ecosystem, if you're working on, you know, just document processing in bulk, if you're working on uh any agentic tasks, that's where it's gonna pay off, right?
00:17:24.640 --> 00:17:26.720
It's cheaper and more token efficient.
00:17:26.880 --> 00:17:34.880
You pay less per token and burn fewer tokens in the process, and that's kind of the uh agenc economics 101 story there.
00:17:35.039 --> 00:17:47.200
Uh so 3.6 does also uh 3.6 flash does deliver higher precision coding with fewer unwanted edits and reduction, uh reduction in loops, essentially, right?
00:17:47.359 --> 00:18:05.680
So if you're a dev running production agents inside Gemini, or if you found the previous version of 3.5 flash to work in terms of price per performance, I think that 3.6 flash, although it's not a big bump in terms of capabilities, it is just a little faster and cheaper and more efficient.
00:18:05.839 --> 00:18:10.799
So uh any developers that have already been using the Flash series is going to be good uh for you.
00:18:10.880 --> 00:18:19.519
And then also Flash Light, important here, will power the future Google search AI overviews where speed really matters the most.
00:18:20.000 --> 00:18:24.880
So the big news here is well, we still don't have a new pro version.
00:18:25.039 --> 00:18:28.160
Yeah, we were uh kind of reports this week.
00:18:28.240 --> 00:18:30.319
If you follow our newsletter, we covered this.
00:18:30.559 --> 00:18:46.319
Said that the pro version may get delayed a few more months, and then we similarly uh got news from uh frequent, uh open a uh frequent, uh everyday AI guest, Logan Kilpatrick, been on the show a handful of times.
00:18:46.400 --> 00:18:50.160
Uh, did say that Google is pre-training Gemini 4.
00:18:50.240 --> 00:18:51.039
So who knows?
00:18:51.279 --> 00:18:57.279
Maybe we technically won't get a Gemini 3.5 Pro or a 3.6 Pro.
00:18:57.519 --> 00:19:01.759
Maybe we'll just be in for a longer wait and we'll get a Gemini 4 Pro.
00:19:02.000 --> 00:19:06.319
Not sure, but Google did say that they're pre-training Gemini 4.
00:19:06.400 --> 00:19:09.920
And we saw reports that the Pro series is delayed.
00:19:10.000 --> 00:19:17.039
So if you put two and two together, maybe that just means there will be no 3.5 or 3.6 Pro, and we'll just be going straight to 4 Pro.
00:19:17.359 --> 00:19:18.319
We'll see.
00:19:18.880 --> 00:19:22.000
All right, let's go to our next one.
00:19:22.319 --> 00:19:27.440
And yeah, this one, I'm both excited but a little bummed.
00:19:27.599 --> 00:19:30.880
But this is kind of how Google rolls out some of their products.
00:19:31.039 --> 00:19:39.519
Maybe I'm just being uh a little snobbish in uh, you know, wanting the best tools for where I actually need them.
00:19:39.680 --> 00:19:48.079
But yes, we finally have a new 24-7 personal agent from Google that just works.
00:19:48.240 --> 00:19:50.160
I'm talking about Gemini Spark.
00:19:50.319 --> 00:19:55.599
So not technically new, but new to most everyone.
00:19:55.680 --> 00:20:05.759
Uh, because this did launch in May, but only those two uh on the ultra plan, which is that$200 a month plan uh from Google Gemini.
00:20:05.920 --> 00:20:08.400
So we didn't cover it on the show in May, right?
00:20:08.480 --> 00:20:14.240
For the most part, I only cover uh on the Friday features those that come out to the base paid plans, right?
00:20:14.400 --> 00:20:19.039
Your$20 a month plan, not your plans that are$100,$200 a month.
00:20:19.279 --> 00:20:31.039
So that was launched for ultra subscribers in May, but now Google has released uh Gemini Spark to all pro subscribers as well.
00:20:31.200 --> 00:20:35.519
So that's Google's personal AI agent for automating complex tasks.
00:20:35.680 --> 00:20:36.720
It's rolling out.
00:20:36.880 --> 00:20:38.400
I have access to it right now.
00:20:38.559 --> 00:20:46.960
So it performs multi-step tasks on your behalf, research, planning, complex workflows while you focus on well, whatever else.
00:20:47.200 --> 00:20:50.319
So it's kind of built on these three different concepts.
00:20:50.559 --> 00:20:55.039
So you have your tasks, which is your high-level goal, you have your skills, right?
00:20:55.119 --> 00:21:01.200
The open skills standards, those are your uh reusable instructions with context, and then schedules.
00:21:01.279 --> 00:21:05.519
So those are automated triggers like times or conditions.
00:21:06.160 --> 00:21:06.480
All right.
00:21:06.559 --> 00:21:12.880
So, like I said, if you are on that base uh paid plan, you should get access to it now.
00:21:12.960 --> 00:21:19.279
Uh, if especially if you're in the US, Google did say that other countries are going to be getting access to it soon.
00:21:19.519 --> 00:21:22.960
So here's the downside or the caveat.
00:21:23.440 --> 00:21:26.480
I have no clue when it's going to be rolling out to workspace plans.
00:21:26.640 --> 00:21:29.200
Uh, this is usually one of those things that they say soon.
00:21:29.359 --> 00:21:36.160
So if your business runs on Google, on a paid Google Gemini account, maybe you have access to it, maybe you don't.
00:21:36.240 --> 00:21:42.160
But if you are on a standard paid uh, you know, Google Gemini plan like me, right?
00:21:42.240 --> 00:21:50.720
So my personal Gmail, I have a uh paid Gemini account for my personal Gmail, I have a paid Gemini account for my work emails.
00:21:50.880 --> 00:21:52.319
So on the work side, we don't have this.
00:21:52.400 --> 00:21:58.880
So for me, it's like, well, I don't know how much I'm gonna be using Gemini Spark in my personal Gmail, right?
00:21:58.960 --> 00:21:59.599
I just use it more.
00:22:00.160 --> 00:22:06.160
For testing and for certain purposes in Gemini that I don't have access to in my workspace account.
00:22:06.319 --> 00:22:08.880
Um, so why is it useful?
00:22:09.039 --> 00:22:12.480
Well, it works inside of the Google apps, right?
00:22:12.640 --> 00:22:30.720
So uh any acts, uh any app that you give Spark access to inside of the Gemini workspace, it can essentially go and you know monitor for updates in those apps, uh, you know, push uh updates from you know docs to slides or from your Gmail to your calendar, uh, right.
00:22:30.799 --> 00:22:35.599
So this also works uh with your personal intelligence, that new feature from Google.
00:22:35.759 --> 00:22:42.799
There's a remote browser uh and computer with code execution, chats, canvas, uh, all those different things.
00:22:43.039 --> 00:22:52.240
So an example that Google gave is you know, saying uh Google Doc, the task is to plan and manage your business trip coming up.
00:22:52.400 --> 00:23:05.200
It'll look through your uh, you know, your Gmail, your calendar and your schedule, and it will tell you, oh, your flight is delayed, or you know, a travel booking skill, you know, plus a Gmail writing skill.
00:23:05.359 --> 00:23:09.359
You can rebook the room and send a confirmation or something like that.
00:23:09.519 --> 00:23:22.079
So Spark can now open, read, and edit Google Docs directly, with that expanding to spreadsheets and presentations, uh, including reading comments and editing shared team docs.
00:23:22.240 --> 00:23:27.119
So I think Gemini Spark was previewed a super long time ago.
00:23:27.200 --> 00:23:30.480
Um, and at the time, I was personally very excited.
00:23:30.640 --> 00:23:42.720
But it took obviously many months for it to actually roll out, and it's still, uh at least according to my research, not available for every single paid workspace plan.
00:23:42.960 --> 00:23:52.400
So obviously, this is great if you run uh Gemini in your business, but not all business paid accounts have access to this.
00:23:52.720 --> 00:23:56.000
All pro accounts on the personal Gmail side do.
00:23:56.240 --> 00:24:06.799
So if I'm being honest, six months ago, if this rolled out to my workspace account, I would be talking about Gemini Spark every single hour because it's great.
00:24:07.039 --> 00:24:13.440
But right now, you can do most of these things, anyways, uh, with codecs or clawed code, clawed desktop.
00:24:13.519 --> 00:24:18.000
So that's why I'm not like gonna be, you know, raving about Gemini Spark.
00:24:18.079 --> 00:24:22.880
You know, it's kind of interesting that in a lot of cases, uh, OpenAI anthropic.
00:24:22.960 --> 00:24:32.079
And the crazy thing is, even sometimes Microsoft um offers some of these features and functionalities before Google does with their own products, right?
00:24:32.240 --> 00:24:41.359
So even Microsoft rolled out a uh essentially an open claw-esque integration, and you can connect with Google.
00:24:41.519 --> 00:24:47.200
So uh, although it's exciting, technically Google a little bit behind even in their own backyard.
00:24:47.359 --> 00:24:54.720
Although I do know that millions of people will like this update, and that's why I decided to cover it on the show.
00:24:55.279 --> 00:24:59.519
All right, uh, two more, and we are getting to our big one.
00:24:59.680 --> 00:25:12.480
But before we talk about how uh Chat GPT can now be your Jarvis, we have to talk about the this is the one where uh Anthropic took a uh page out of the codex playbook and made it a little bit better.
00:25:12.640 --> 00:25:19.279
Uh, that is the new record a skill feature uh that just shipped this week in Claude Co-Work.
00:25:19.599 --> 00:25:31.519
So this is essentially you can screen record yourself doing a task once, and then Claude analyzes the recording and turns it into a reusable, re-uh runnable skill.
00:25:32.000 --> 00:25:43.839
So Claude captures screen activity, cursor movements, keystrokes, and the new addition that we don't have just yet in codex, at least not by default, is your voice narration.
00:25:44.079 --> 00:25:52.400
So then uh it processes all of that together uh in the recording and it turns it into a structured skill that saves it to your library.
00:25:52.559 --> 00:25:53.759
So who has access?
00:25:53.920 --> 00:25:57.279
It is rolled out to all paid plans now.
00:25:57.519 --> 00:26:01.680
So um free users can still use existing skills, right?
00:26:01.759 --> 00:26:04.880
But you can't obviously record new ones.
00:26:05.519 --> 00:26:11.839
So it was a little confusing to find this, so it's only available on co-work.
00:26:12.000 --> 00:26:15.440
Um, so you have to go into the chat interface.
00:26:15.599 --> 00:26:19.440
So now obviously Claude Desktop changed their interface a little bit.
00:26:19.599 --> 00:26:21.440
It used to be the three different tabs.
00:26:21.519 --> 00:26:23.519
It was chat, cowork, and code.
00:26:23.680 --> 00:26:27.839
Now it says chat and code, but in chat there's a co-work tab, right?
00:26:28.000 --> 00:26:31.039
So you have to be in chat, you have to go find your co-work tab.
00:26:31.119 --> 00:26:35.599
You hit the plus button, and then you can find uh the new recorder skill.
00:26:35.680 --> 00:26:37.839
So it actually took me a little bit of clicks to find it.
00:26:37.920 --> 00:26:39.039
I'm like, wait, where is this?
00:26:39.119 --> 00:26:40.240
So you do need to update this.
00:26:40.319 --> 00:26:41.680
This is desktop only.
00:26:41.839 --> 00:26:50.240
So update your Claude desktop app, go into chat, click the cowork tab, click the plus button, and then and actually super useful.
00:26:50.319 --> 00:26:51.920
So I did a little bit of testing.
00:26:52.000 --> 00:26:53.359
Uh, it's really good.
00:26:53.519 --> 00:26:54.799
So, why is it useful?
00:26:54.960 --> 00:27:01.839
Well, until now, creating a skill meant either chatting with Claude and having it create one for you.
00:27:02.000 --> 00:27:14.559
You know, you can obviously create those skill MD files uh by hand, but so many times it's just like at the end, uh, and and and you should be doing this FYI if you're not already doing this.
00:27:14.799 --> 00:27:28.240
Um, the way I look at new chats or tasks in any uh, you know, large language model, especially any agentic one, when I'm done and I get an output where I'm like, this is good, you should be turning that into a skill, right?
00:27:28.400 --> 00:27:36.799
So this is instead of having to work through it or to hand hold right Claude or Codecs or anti-gravity or anything, right?
00:27:36.880 --> 00:27:39.680
At the end, you would normally say turn this into a skill.
00:27:39.839 --> 00:27:44.480
So the difference here with the teach Claude a skill is well, you don't have to work through it.
00:27:44.640 --> 00:27:52.240
You just click start recording, you do your work, you dictate through it, you say, Here's what I'm doing, A, B, and C.
00:27:52.400 --> 00:27:55.839
I'm opening this site, here's what I'm looking for on this page.
00:27:56.000 --> 00:27:59.039
I download it, I open it in this program, right?
00:27:59.119 --> 00:28:03.200
And it'll just record all of that and turn it into a skill.
00:28:03.359 --> 00:28:13.759
So, like I said, technically taking a uh page out of the codex uh replay uh or no record and replay skill uh that did this exact same thing.
00:28:13.839 --> 00:28:18.400
But the downside with codexes is by default there is no voice narration, right?
00:28:18.480 --> 00:28:26.400
There is kind of some workarounds um around that, but it is nice that in this new one for Claude Cowork, it is just enabled by default.
00:28:26.480 --> 00:28:33.440
You don't have to have any workarounds to actually narrate your way through that skill, uh, which is very helpful, obviously.
00:28:33.680 --> 00:28:35.440
So, who's gonna find this valuable?
00:28:35.599 --> 00:28:37.200
Well, it literally anyone.
00:28:37.279 --> 00:28:42.720
So if you're a Claude desktop user and you use Claude Cowork, it's gonna be great.
00:28:42.799 --> 00:28:47.440
Um, I would love to see uh Anthropic roll this out for Claude Code as well.
00:28:47.599 --> 00:28:49.599
Um, I think that would be really helpful.
00:28:49.759 --> 00:29:06.640
Uh, just because seemingly, um, I'm guessing Claude Cowork um is not kind of riding that same wave of popularity that it was in earlier uh 2026, just because the the interface kind of got changed over a little bit.
00:29:06.799 --> 00:29:12.880
And I think more and more people are starting to use Claude Code and just more of the agentic features in the normal chat section.
00:29:13.039 --> 00:29:16.079
So you can only set this up in the cowork, which is a downside.
00:29:16.160 --> 00:29:21.599
But literally anyone that uses Claude and you're doing the same processes over and over.
00:29:21.920 --> 00:29:23.519
I mean, this is big, right?
00:29:23.599 --> 00:29:30.160
I love, I absolutely loved and use the this codec skill um all the time for my workflows.
00:29:30.240 --> 00:29:34.079
So I'm personally happy uh to see this in uh claude code.
00:29:34.240 --> 00:29:37.039
The other thing, here's the here's the big unlock, y'all.
00:29:37.200 --> 00:29:40.319
You can do this in I talked about this in codecs as well.
00:29:40.400 --> 00:29:50.960
Uh, you can do this in Claude Code or Codex, go through, record the skill, and skills are shareable and they're fairly uh openly supported across all different platforms.
00:29:51.039 --> 00:30:03.039
So you can just create a skill, uh, very detailed skill as an example using this new feature from Anthropic, and then use that skill inside of Google or use that skill inside of Codex or Chat GPT.
00:30:03.519 --> 00:30:07.119
All right, and our last big feature.
00:30:07.279 --> 00:30:14.799
And y'all, this this one, this is one of those instances where I'm like, okay, this is crazy after I tried uh this.
00:30:14.880 --> 00:30:18.079
So yeah, hasn't even been out a full 24 hours.
00:30:18.240 --> 00:30:25.119
Uh, but open AI went to full Jarvis mode and they launched Chat GPT voice on desktop.
00:30:25.200 --> 00:30:30.000
And it's a lot more than it sounds like, but let me go through the bullet point details first.
00:30:30.160 --> 00:30:43.680
So OpenAI launched ChatGPT Voice in the desktop app, uh, powered by GPT Live, which I already referenced, that lets you talk through work and coordinate tasks across chat work and codecs.
00:30:43.920 --> 00:31:00.000
So per OpenAI announcement, you can control your computer and direct multiple agents running in Chat GPT work or codex using just your voice with GPT Live, speaking, listening, and coordinating work simultaneously.
00:31:00.319 --> 00:31:04.559
So here is how open AI kind of describes the main benefits.
00:31:04.720 --> 00:31:13.200
So it says you can work across projects, you can start a new task, coordinate work in progress, and pick up where you left off without micromanaging.
00:31:13.359 --> 00:31:21.039
You can work across your desktop, you can use your files, apps, and connected tools from Slack and GitHub to Notion and beyond.
00:31:21.200 --> 00:31:23.519
And you can explore, plan, and learn.
00:31:23.599 --> 00:31:30.480
You can talk through a pull request, ask questions about a code base, or learn a new topic through natural back and forth.
00:31:30.640 --> 00:31:31.839
So who has access?
00:31:32.000 --> 00:31:40.480
Well, if you are on any paid plan and you're using the desktop version of Chat GPT Workslash Codex, it's the same thing.
00:31:40.720 --> 00:31:53.039
It is out now, and it works also with uh ChatGPT's remote feature, which is pretty cool because you don't even need to be in front of your computer, you can just have your phone, right?
00:31:53.119 --> 00:32:11.279
I have one uh right, I'm traveling right now, I'm not in Chicago, but I have essentially a home Mac Studio that never gets shut off, and I can be controlling it right now and just with my voice, and it can be doing literally anything, and I can be watching and seeing it go in real time.
00:32:11.680 --> 00:32:20.559
So this just kind of brings that uh GPT live mode that was announced about two weeks ago uh to the desktop.
00:32:20.640 --> 00:32:34.960
So before that was only available via mobile, so it doesn't just bring it to the desktop, but this is the first time where I'm like, wait, you technically don't even need to type or use your mouse anymore.
00:32:35.119 --> 00:32:45.920
Then yes, I haven't been able to run it through many hours of demos, but um the possibilities for this one are absolutely bonkers, y'all.
00:32:46.079 --> 00:32:47.359
And it is so good.
00:32:47.519 --> 00:32:49.039
So let me talk about a couple things.
00:32:49.200 --> 00:32:52.240
Number one, it works with ChatGPT's remote feature, which is great.
00:32:52.319 --> 00:33:02.079
That's just on your mobile app, you can click remote and you can technically uh control any computer that you have connected to that chat GPT account.
00:33:02.240 --> 00:33:03.119
So that's cool.
00:33:03.279 --> 00:33:05.200
But here's the other thing app shots.
00:33:05.519 --> 00:33:05.839
All right.
00:33:06.000 --> 00:33:23.519
So what app shots are, if you're you know a power, you know, codex or chat GPT work user, the default uh, you know, is the two command keys and it takes essentially a screenshot, but not just a screenshot, but every single thing it brings into context that you have um in that program.
00:33:23.680 --> 00:33:30.240
So as an example, right now on my screen, um, I have up some notes, some bullet points that I wanted to go over, right?
00:33:30.319 --> 00:33:31.920
But it's a very long list.
00:33:32.160 --> 00:33:37.440
So obviously what's on my screen is probably only like five percent of what's in that document.
00:33:37.599 --> 00:33:50.799
So if I use the app shots, it's gonna obviously take a screenshot of what is actually on my screen, but it brings in the context of everything else in uh that program that I have up that is not even on the screen.
00:33:50.960 --> 00:33:56.160
So the cool thing is with this new uh, you know, I kind of wish we could have got a fun name for it.
00:33:56.240 --> 00:33:57.359
I know they can't use Jarvis.
00:33:57.519 --> 00:33:58.960
I'm just using Jarvis, right?
00:33:59.200 --> 00:34:07.680
Like you can have this app shot feature and functionality while you're using Chat GPT uh live voice on desktop.
00:34:08.000 --> 00:34:21.280
So in my kind of playing around, I had, you know, six or seven different programs open on my computer, hands-free, not typing and just telling Chat GPT work or codecs what to do.
00:34:21.440 --> 00:34:22.480
Open this file.
00:34:22.639 --> 00:34:23.519
Okay, great.
00:34:23.679 --> 00:34:26.079
Um, can you find this change that I made?
00:34:26.239 --> 00:34:27.519
It's probably somewhere in the middle.
00:34:27.599 --> 00:34:41.440
And it can go, it can find that using the app shots, uh, right, bringing in all that other context without having me to always tell it, oh, somewhere in this document, you know, I I think that I outlined uh something about that that certain KPI, right?
00:34:41.519 --> 00:34:43.760
It's all there and it's absolutely bonker.
00:34:43.920 --> 00:34:48.880
So uh, you know, why we keep saying Jarvis is this is literally the scene out of Iron Man, right?
00:34:49.039 --> 00:34:56.239
When Tony, uh Tony Stark is at his computer just saying, Hey, Jarvis, you know, pull up this, go do this, right?
00:34:56.320 --> 00:34:59.519
Now we actually have this for the first time and it works.
00:34:59.679 --> 00:35:09.360
So, not just having the remote is cool, the app shots is kind of that uh that hack, but what I'm really excited about is for the future of this.
00:35:09.519 --> 00:35:16.079
So, obviously, open AI's uh computer use with the new GPT 5.6 model took a huge leap forward.
00:35:16.159 --> 00:35:24.639
But the thing I'm actually excited about is when we get that super blazing fast uh uh Cerebras uh integration.
00:35:24.719 --> 00:35:28.559
So, you know, OpenAI did say that it was coming in July.
00:35:28.880 --> 00:35:43.440
So when we can bring this with uh this ridiculously fast uh version of GPD 5.6, what you are able to accomplish in front of your computer without even sitting down is amazing.
00:35:43.599 --> 00:35:52.000
So I don't know, maybe maybe this is gonna bring back the popularity of having like a treadmill desk or something like that, or just being able to go touch grass.
00:35:52.079 --> 00:35:57.039
I'm like, okay, is this gonna also, you know, the XR glasses and all these other things?
00:35:57.199 --> 00:36:04.159
I'm actually geeking out and excited about this one because I don't know, like I don't like sitting in front of a screen all day.
00:36:04.320 --> 00:36:07.039
I like getting up and walking around a lot, right?
00:36:07.119 --> 00:36:11.119
And obviously, you can have voice dictation apps, but that's different.
00:36:11.199 --> 00:36:15.920
So this is just a completely different way to work, a different way to interface with smart AI.
00:36:16.079 --> 00:36:17.920
It's one I'm super excited about.
00:36:18.000 --> 00:36:19.599
So this is useful to anyone.
00:36:19.679 --> 00:36:39.199
So if you're using uh the new Chat GPT work desktop app, which is essentially the uh developer or the non-uh technical version of codex, uh, this is in theory a way to completely change how you interface with the computer, not just how you use Chat GPT and how you work, but literally how you interface with a device.
00:36:39.360 --> 00:36:41.760
So um I'm geeked out about that one.
00:36:41.920 --> 00:36:43.920
I am too, or I hope you are too.
00:36:44.079 --> 00:36:46.960
Maybe I'll do a uh Wednesday demo of that one.
00:36:47.119 --> 00:36:51.039
So that's a wrap on all the things that are new this week.
00:36:51.119 --> 00:36:58.880
Let me give you a very quick update of our seven AI features you can use today uh if you have paid plans.
00:36:59.039 --> 00:37:02.159
So ChatGPT Health is out and launched.
00:37:02.320 --> 00:37:04.960
Claude has some nice voice mode upgrades.
00:37:05.039 --> 00:37:11.199
You no longer have to talk to the dumber haiku model, you can talk to the smart uh opus and sonnet.
00:37:11.280 --> 00:37:15.039
Uh, we have the new Microsoft MAI image 2.5 Pro.
00:37:15.440 --> 00:37:21.119
Uh Google released Gemini 3.6 Flash and Gemini 3.5 Flash Lite.
00:37:21.280 --> 00:37:23.519
Uh, we got the new Gemini Spark.
00:37:23.599 --> 00:37:29.519
Uh Google's always on 24-7 AI agent that is available now to pro users.
00:37:29.760 --> 00:37:31.519
Anthropics, very useful.
00:37:31.599 --> 00:37:38.079
Claude, record a skill, uh, including that voice, uh, being able to narrate your skill, which is great.
00:37:38.239 --> 00:37:43.760
And then last but not least, openai went full Jarvis mode with Chat GPT voice on desktop.
00:37:43.840 --> 00:37:44.639
I hope this was helpful.
00:37:44.719 --> 00:37:45.760
If so, do me a favor.
00:37:45.840 --> 00:37:53.440
If you're not already, please subscribe to the podcast on Spotify or Apple, and then go to your everydayai.com.
00:37:53.519 --> 00:37:58.480
We're going to be recapping the highlights from today's show, anything you missed, and everything else in today's newsletter.
00:37:58.559 --> 00:37:59.679
So thank you for tuning in.
00:37:59.760 --> 00:38:02.960
We'll see you back Monday and every day for more everyday AI.
00:38:03.039 --> 00:38:04.079
Thanks, y'all.
00:00:01.360 --> 00:00:10.560
If you listened to our Monday AI News That Matters segment, I told you we were kind of in for a wild week.
00:00:10.720 --> 00:00:22.879
That's because after a somewhat slow-ish week last week, I said on Monday that this week was going to be a banger in terms of new AI features you can actually use.
00:00:23.359 --> 00:00:34.159
Sure, I knew ahead of time a few things that were gonna come out, but even I was kind of blown away just yesterday at the sheer amount of new updates that all dropped within hours of each other.
00:00:34.399 --> 00:00:44.399
I mean, no joke, on Thursday alone, we got huge new AI features from Google, Microsoft, Anthropic, and OpenAI in a matter of hours.
00:00:44.640 --> 00:00:50.000
That's why on Fridays we break down what's new in AI that you can actually use.
00:00:50.079 --> 00:00:51.920
And we do it in a very specific way.
00:00:52.079 --> 00:00:59.280
We give you the AI features that are available for you today, not rumors, leaks, or wait lists.
00:00:59.759 --> 00:01:05.040
These are instant AI upgrades that you should be putting into play today.
00:01:05.200 --> 00:01:08.719
And we've got a lot to cover on today's show.
00:01:08.879 --> 00:01:19.840
So stick with me for the next 25-ish minutes, and I'm gonna tell you how ChatGPT went full Jarvis mode and the new app shot secret that unlocks the future of work.
00:01:20.000 --> 00:01:27.599
I'm gonna let you know the new Claude feature that took a page out of the Codex playbook, but actually made it a little bit better.
00:01:27.840 --> 00:01:37.040
And I'm gonna fill you in on the kind of new Google agent that you can start using today, but with one caveat.
00:01:37.439 --> 00:01:38.959
All right, let's get into it.
00:01:39.040 --> 00:01:40.400
Welcome to Everyday AI.
00:01:40.480 --> 00:01:42.719
My name's Jordan Wilson, and we do this every day.
00:01:42.879 --> 00:01:52.560
It's your unedited, unscripted daily live stream podcast and free daily newsletter helping business leaders like you and me keep up with the avalanche of AI news and updates.
00:01:52.719 --> 00:01:54.879
I tell you what matters, how to use it.
00:01:55.040 --> 00:02:00.959
You take that information, poof, you're the smartest person in AI and you can grow your company and your career.
00:02:01.200 --> 00:02:05.840
So it starts here, but make sure you go to the website at your everydayai.com.
00:02:06.000 --> 00:02:08.080
We're gonna be recapping the highlights from today's show.
00:02:08.159 --> 00:02:14.319
So if you miss any of the details, maybe you're out on a walk and you want to go back and say, wait, how did this work?
00:02:14.479 --> 00:02:19.280
It's all gonna be in the newsletter as well as all of the other AI updates you need to know.
00:02:19.599 --> 00:02:21.599
All right, let's get straight into it.
00:02:21.759 --> 00:02:24.560
I'm a big fan of the Friday features show.
00:02:24.719 --> 00:02:27.520
I wish I would have started it like three and a half years ago.
00:02:27.680 --> 00:02:36.479
We've been doing it now for about six months, uh, but just for our podcast audience, I'm always showing uh on my screen some basic, you know, information from the company.
00:02:36.560 --> 00:02:37.680
You're not missing anything.
00:02:37.759 --> 00:02:39.439
So uh just FYI.
00:02:39.520 --> 00:02:41.599
But let's dive in live.
00:02:41.759 --> 00:02:49.840
I'm gonna be sharing my screen, uh, just kind of showing the uh windows for the different announcements that we're gonna be going over.
00:02:50.080 --> 00:02:54.879
All right, so let's start first with your health.
00:02:55.120 --> 00:03:00.319
Yeah, this is actually um one I'm personally excited about.
00:03:00.479 --> 00:03:09.520
I know some people aren't gonna be, you know, diving in headfirst necessarily to give their health data to anyone, let alone an AI company.
00:03:09.759 --> 00:03:14.000
For me, I'm absolutely wildly gonna be using this.
00:03:14.159 --> 00:03:20.319
So here's what's new ChatGPT Health is finally rolled out to everyone.
00:03:20.479 --> 00:03:24.319
It is launched for all US users today.
00:03:24.960 --> 00:03:38.400
And it's expanding the feature that was first introduced in January, but it was introduced as a wait list uh feature, and it connects your Apple Health and supported medical records directly into Chat GPT.
00:03:38.560 --> 00:03:39.840
So, what the heck can it do?
00:03:39.919 --> 00:03:40.639
Why is it useful?
00:03:40.800 --> 00:03:55.840
Well, it can compare your lab results, summarize changes since a prior appointment, track your medications, factor in your sleep activity and workouts, essentially anything that you know your Apple Health might track and all the different applications that connect in there.
00:03:56.000 --> 00:04:03.039
Well, it can bring in that data and anything else that you can update or upload into Chat GPT Health.
00:04:03.280 --> 00:04:10.800
So OpenAI says that health conversations obviously live in a dedicated space, siloed from all your other chats, right?
00:04:10.879 --> 00:04:16.959
So yeah, you're not gonna have all of that information popping up via like a memory uh into your other chats.
00:04:17.120 --> 00:04:20.959
So, like I said, this is already rolling out to US users.
00:04:21.040 --> 00:04:29.920
You do have to be 18 or older, but the cool thing is this is available on free plans, on the paid plans, uh, and also on iOS.
00:04:30.720 --> 00:04:38.800
So one small caveat though, free users, you're just gonna be using kind of the model that you have the most access to, which is GPT 5.5 instant.
00:04:38.879 --> 00:04:45.040
Um, and then for other paid users, you'll be able to use the more powerful GPT-5.6 sold.
00:04:45.680 --> 00:04:48.639
All right, so why is this useful?
00:04:48.879 --> 00:04:49.199
Right?
00:04:49.360 --> 00:04:56.079
Because right now, obviously, there's no one place for all of your health information to live.
00:04:56.240 --> 00:05:01.920
Like, I can't tell you the amount of conversations I've had with people about this over the last few years.
00:05:02.079 --> 00:05:06.480
And, you know, there was a company out there that was kind of, you know, AI native.
00:05:06.560 --> 00:05:08.240
Uh, they were called Forward Health.
00:05:08.319 --> 00:05:09.279
They are no more.
00:05:09.680 --> 00:05:22.480
Um, and I was kind of excited about that concept because I'm like, there needs to be, you know, something that's not tied to a single provider that allows you to just essentially bring in all your health records and to have a smart system that just knows all that.
00:05:22.639 --> 00:05:32.879
So obviously, there was nothing previously that would have stopped you from doing this inside of a large language model like Claude or ChatGPT, Gemini, Copilot, et cetera.
00:05:32.959 --> 00:05:34.959
But it wasn't really set up for this, right?
00:05:35.040 --> 00:05:40.560
So think of this as a specialized version of Chat GPT that's siloed from everything else.
00:05:40.639 --> 00:05:46.319
So it keeps all of your health data uh separate, but it is literally built for this.
00:05:46.639 --> 00:05:55.680
So, you know, right now your health info is scattered across different patient portals, EMRs, apps, and it just pulls it all into the conversational layer.
00:05:55.839 --> 00:05:59.839
And there's also some kind of practical prompts baked up for you, right?
00:06:00.000 --> 00:06:07.199
So how your cholesterol might be trending, or how you can summarize your blood work before an appointment, what you should ask your doctor tomorrow, etc.
00:06:07.600 --> 00:06:16.480
And that context kind of bleeds uh usefully into everyday tasks, like factoring a dietary restriction into your restaurant picks.
00:06:17.120 --> 00:06:19.040
So, who's gonna find this valuable?
00:06:19.199 --> 00:06:30.399
Well, this is obviously a consumer play, but I think this just goes to show where OpenAI has started to focus more recently, uh, right?
00:06:30.560 --> 00:06:37.680
Not necessarily just on the models, but how to make their models more useful for more people in more ways.
00:06:37.839 --> 00:06:52.079
So, you know, whether you are just someone that wants to get a little bit more serious about your health, you just have questions, or if you're managing chronic conditions, uh, if you're a caregiver trying to track something in your family health or just preparing for your upcoming appointment.
00:06:52.319 --> 00:07:02.079
So obviously, open AI says this is not intended for diagnosis or treatments, and health conversations aren't used to train foundation models.
00:07:02.399 --> 00:07:05.279
Um, so pretty big one.
00:07:05.439 --> 00:07:06.480
I'm excited to get this.
00:07:06.560 --> 00:07:13.920
I signed up for the uh the wait list previously, wasn't uh wasn't kind of uh selected to be part of that first group.
00:07:14.000 --> 00:07:16.800
So this is one I'm gonna be jumping into head first.
00:07:17.279 --> 00:07:20.000
All right, our next AI update.
00:07:20.160 --> 00:07:28.319
A new update, well, new-ish feature from Anthropic, just kind of borrowing from the codex playbook, but I think they made it a little bit better.
00:07:28.560 --> 00:07:32.160
So Claude has some new voice mode upgrades.
00:07:32.240 --> 00:07:33.519
Oh no, this is not the one.
00:07:33.600 --> 00:07:37.600
Sorry, that's a little bit uh better than uh ChatGPT's.
00:07:37.680 --> 00:07:43.360
So this one is voice mode, not quite as good, but also this one is a little different.
00:07:44.720 --> 00:07:47.360
Yeah, just so many new updates.
00:07:47.519 --> 00:07:51.040
Even I'm getting confused, and this is all I do every single day.
00:07:51.199 --> 00:08:02.480
All right, so this new update from Anthropic is essentially expanding how and where their voice mode can be used.
00:08:02.800 --> 00:08:09.920
So Anthropic has upgraded the Claude voice mode to run on Opus and Sonnet for the first time.
00:08:10.079 --> 00:08:23.759
That's because previously it ran only on their least powerful model, which is haiku, which is why for me, I absolutely never used voice mode uh inside Claude on my phone for that very reason.
00:08:23.839 --> 00:08:25.680
Uh, but now I probably will.
00:08:25.839 --> 00:08:32.720
The cool thing as well is you can switch models mid-conversation in voice mode now works with connectors.
00:08:32.879 --> 00:08:46.159
So that is the big difference right now with the um the one thing that's maybe a little bit better is that their upgraded voice mode works with connectors, but it's not bi-directional.
00:08:46.240 --> 00:08:51.759
So it's still kind of the quote unquote older, dumber uh version of a voice mode.
00:08:51.919 --> 00:08:55.279
So you you've heard me kind of rave about the new GPT Live.
00:08:55.440 --> 00:09:00.559
That's because GPT Live can listen to you um and speak at the same time, right?
00:09:00.639 --> 00:09:02.960
You can interrupt it, it might interrupt you.
00:09:03.039 --> 00:09:06.480
Um, and it's kind of listening, thinking, and doing all these things at the same time.
00:09:06.639 --> 00:09:14.639
So this is still the old school walkie talkie mode where you talk, you wait, Claude responds.
00:09:14.960 --> 00:09:20.000
But the big benefit here is it can use connectors, so pulling in your live data.
00:09:20.080 --> 00:09:26.320
So that's not something that GPT Live can do, although you can individually upload files into GPT Live.
00:09:26.399 --> 00:09:29.440
And OpenAI did say that they are working on bringing connectors.
00:09:29.600 --> 00:09:37.440
So here's who has access: you have to have a paid Claude account to get that upgraded Sonnet in Opus voice mode.
00:09:37.679 --> 00:09:40.000
Otherwise, you will be defaulted to Haiku.
00:09:40.320 --> 00:09:46.159
So free accounts do have a very limited usage and only a single connection.
00:09:46.320 --> 00:09:49.120
Um, and all prompts then obviously go through haiku.
00:09:49.519 --> 00:09:53.440
And right now, voice mode does support 11 languages.
00:09:53.919 --> 00:09:55.759
So, why is this useful?
00:09:56.000 --> 00:10:06.559
Well, if you want to be able to chat with your connectors and you have a paid quad plan, this is really good, especially if you don't mind kind of working in walkie-talkie mode.
00:10:06.720 --> 00:10:14.639
Uh, for me, I've really enjoyed kind of the uh the bi-directional or duplex uh new way of talking.
00:10:14.799 --> 00:10:17.200
It seems like that's the future of the voice interface.
00:10:17.360 --> 00:10:27.039
So as long as you don't mind kind of this more kind of waiting and holding and you know, only being able to talk or listen at once, not that bad.
00:10:27.200 --> 00:10:28.480
So, who's gonna find this useful?
00:10:28.639 --> 00:10:43.039
I mean, anyone that's a power uh user of Claude, if you're doing any hands-free work, commuters, you know, executives between meetings or just operators who want to think out loud, uh, pretty big update and release here from Anthropic.
00:10:43.679 --> 00:10:44.320
All right.
00:10:44.480 --> 00:10:50.639
Uh, next we have a yes, another new image mode from Microsoft.
00:10:50.879 --> 00:10:57.840
All right, so here's what's new in MAI Images or sorry, MAI Image 2.5 Pro.
00:10:58.159 --> 00:11:01.600
Uh, all these image models are always a mouthful.
00:11:01.759 --> 00:11:02.559
Um, right.
00:11:02.879 --> 00:11:13.679
We should just, I don't know, Microsoft get a fun like nano banana uh name, and then we can call it, you know, why not like Clippy, uh Clippy Pro, you know, 2.5.
00:11:13.840 --> 00:11:14.240
All right.
00:11:14.320 --> 00:11:18.960
So this is launched and it's launched in a lot of different places.
00:11:19.200 --> 00:11:34.960
So it is in Foundry uh preview, it's in Cop uh it's in Microsoft um Copilot PowerPoint, um, and it's in also a variety of other places, but the new image mode, it's pretty good.
00:11:35.039 --> 00:11:42.720
Um, I don't think we have a lot of benchmarks on it just yet, but I would assume that we get those probably uh within a couple of days.
00:11:42.960 --> 00:11:56.559
So this is Microsoft's newest, highest Vidal A professional grade image model, built, they say, for superior high-quality imagery, detailed editing, and precise in image text rendering.
00:11:56.879 --> 00:12:09.039
So this does uh Microsoft says handles text to image generation plus controllable edits, object removal, replacement in painting, text updates, all while preserving composition.
00:12:10.000 --> 00:12:11.440
Uh, so who has access?
00:12:11.519 --> 00:12:14.080
So yeah, it is now available in Microsoft Foundry.
00:12:14.320 --> 00:12:22.559
If you do have a Microsoft uh 365 co-pilot plan, you can use this right now inside of PowerPoint, or you can use it via the API.
00:12:22.639 --> 00:12:31.279
The pricing is five dollars uh per 1 million text input tokens or$8 per 1 million image uh input tokens.
00:12:31.759 --> 00:12:36.720
All right, and then uh$106 per 1 million image output tokens.
00:12:36.879 --> 00:12:38.720
So why is this useful?
00:12:38.960 --> 00:12:48.720
Well, I'll say this right now, if you're one of those organizations that can only use Microsoft Copilot and you can't use anything else, then this is great.
00:12:48.799 --> 00:12:54.080
This is a nice upgrade over the previous MAI image 2.0.
00:12:54.399 --> 00:12:56.879
Uh, but it's still presumably, right?
00:12:56.960 --> 00:13:00.159
We'll see what the benchmarks say, at least from my trained eye.
00:13:00.320 --> 00:13:02.399
This is still fairly far behind.
00:13:02.559 --> 00:13:10.480
Uh GPT Images 2, which is the best AI image uh you know model in the world, and also the Nano Banana 2.
00:13:10.559 --> 00:13:21.840
Uh, so those are kind of the top two, and then you have some uh some more uh you know Chinese open uh versions of these, but the best two still are GPT images and Google's Nano Banana.
00:13:21.919 --> 00:13:24.720
Uh so not really in the same tier.
00:13:24.879 --> 00:13:29.279
We'll see what the benchmarks say, at least by my kind of quick trained eye.
00:13:29.440 --> 00:13:31.759
I still think that there's probably a quality drop-off.
00:13:31.840 --> 00:13:45.360
But regardless, if you are someone, especially I think in um if you are building decks in PowerPoint via Copilot, being able to use this image generator in there is gonna be big, right?
00:13:45.440 --> 00:13:55.360
So yeah, I know that so many people out there are kind of living inside of PowerPoint, and maybe your visuals and the old clip art just won't do.
00:13:55.519 --> 00:13:57.360
Um, I think this is gonna be pretty big.
00:13:57.440 --> 00:14:03.600
So, you know, if you're trying to create product imagery, marketing visuals, brand assets, um big.
00:14:03.759 --> 00:14:07.600
It is big for those purposes when you can't use anything else.
00:14:07.840 --> 00:14:08.480
All right.
00:14:09.200 --> 00:14:11.679
Our next AI update you can use today.
00:14:11.840 --> 00:14:20.960
Yeah, it was this big of a week that I didn't even really honestly get a chance to play with a brand new model from Google Gemini.
00:14:21.200 --> 00:14:27.919
Aside from just some daily driving tasks, haven't really given the new Gemini 3.6 Flash a run.
00:14:28.159 --> 00:14:32.879
Uh, so here is what's new with Gemini 3.6 flash.
00:14:33.039 --> 00:14:41.759
There's actually three different models, but I say most people are focused on Gemini 3.6 flash, but they also unleashed uh Gemini 3.5 flash light.
00:14:41.840 --> 00:14:55.200
I know confusing, so they released both uh the 3.5 upgrades and updates and new variations in the 3.6 flash, which is the first of the 3.6 flash series.
00:14:55.360 --> 00:14:59.360
Even more confusing, the pro series is still stuck on 3.1.
00:14:59.519 --> 00:15:06.480
So yeah, now your latest and greatest Google Gemini models technically have three different step changes.
00:15:06.639 --> 00:15:13.840
You have 3.1 Pro, because we are still waiting for 3.5, or maybe they'll skip straight to 3.6 Pro.
00:15:14.080 --> 00:15:24.480
So you have 3.1 Pro, then you have the 3.5 uh Flash Light and 3.5 Flash Cyber, and then the only one on the 3.6 tier is Flash.
00:15:24.720 --> 00:15:26.879
So extremely confusing.
00:15:27.039 --> 00:15:32.639
They're they're all for different purposes, but let's break down at least uh what's new in these models.
00:15:33.200 --> 00:15:36.879
So uh Gemini 3.6 flash.
00:15:36.960 --> 00:15:44.000
Uh essentially not that much smarter per se, but it is cheaper and more token efficient.
00:15:44.159 --> 00:15:56.159
So Google says that Gemini 3.6 Flash consumes 17 fewer tokens than 3.5 flash while taking fewer reasoning steps and tool calls in multi-step workflows.
00:15:56.240 --> 00:16:03.200
Uh, then you have 3.5 flashlight, which is the cheaper and faster version of Flash.
00:16:03.279 --> 00:16:12.720
All right, so Flash is like your, you know, cheaper version, and then Flashlight is the cheaper and faster and more lightweight version.
00:16:12.960 --> 00:16:25.919
So the 3.5 flashlight is the fastest model in the 3.5 series at 350 output tokens per second, built for low latency, high throughput work like agentix search and document processing.
00:16:26.000 --> 00:16:32.240
And then you also have that third model, Gemini 3.5 cyber, which is tuned specifically for cybersecurity tasks.
00:16:32.480 --> 00:16:34.159
So who has access?
00:16:34.320 --> 00:16:35.039
Well, everyone.
00:16:35.200 --> 00:16:46.480
If you're a paid user, you can go into Gemini, you know, your Gemini.google.com, the Gemini API inside AI Studio, Android Studio, Anti-Gravity, Gemini Enterprise, uh, right?
00:16:46.639 --> 00:16:53.039
Literally wherever you're using Google Gemini for the most part, uh, you can find the new 3.6 flash there.
00:16:53.200 --> 00:16:56.720
Uh, Flash Cyber, though, is limited access only.
00:16:56.879 --> 00:17:07.359
Uh, that's the you know, government and trusted partners, similar to you know, a Mythos or OpenAI has their uh daybreak, I believe is that what their version is called.
00:17:07.599 --> 00:17:10.960
So, why is this useful and who should be using it?
00:17:11.119 --> 00:17:24.480
Well, if you are a developer, especially if you live inside of the Google ecosystem, if you're working on, you know, just document processing in bulk, if you're working on uh any agentic tasks, that's where it's gonna pay off, right?
00:17:24.640 --> 00:17:26.720
It's cheaper and more token efficient.
00:17:26.880 --> 00:17:34.880
You pay less per token and burn fewer tokens in the process, and that's kind of the uh agenc economics 101 story there.
00:17:35.039 --> 00:17:47.200
Uh so 3.6 does also uh 3.6 flash does deliver higher precision coding with fewer unwanted edits and reduction, uh reduction in loops, essentially, right?
00:17:47.359 --> 00:18:05.680
So if you're a dev running production agents inside Gemini, or if you found the previous version of 3.5 flash to work in terms of price per performance, I think that 3.6 flash, although it's not a big bump in terms of capabilities, it is just a little faster and cheaper and more efficient.
00:18:05.839 --> 00:18:10.799
So uh any developers that have already been using the Flash series is going to be good uh for you.
00:18:10.880 --> 00:18:19.519
And then also Flash Light, important here, will power the future Google search AI overviews where speed really matters the most.
00:18:20.000 --> 00:18:24.880
So the big news here is well, we still don't have a new pro version.
00:18:25.039 --> 00:18:28.160
Yeah, we were uh kind of reports this week.
00:18:28.240 --> 00:18:30.319
If you follow our newsletter, we covered this.
00:18:30.559 --> 00:18:46.319
Said that the pro version may get delayed a few more months, and then we similarly uh got news from uh frequent, uh open a uh frequent, uh everyday AI guest, Logan Kilpatrick, been on the show a handful of times.
00:18:46.400 --> 00:18:50.160
Uh, did say that Google is pre-training Gemini 4.
00:18:50.240 --> 00:18:51.039
So who knows?
00:18:51.279 --> 00:18:57.279
Maybe we technically won't get a Gemini 3.5 Pro or a 3.6 Pro.
00:18:57.519 --> 00:19:01.759
Maybe we'll just be in for a longer wait and we'll get a Gemini 4 Pro.
00:19:02.000 --> 00:19:06.319
Not sure, but Google did say that they're pre-training Gemini 4.
00:19:06.400 --> 00:19:09.920
And we saw reports that the Pro series is delayed.
00:19:10.000 --> 00:19:17.039
So if you put two and two together, maybe that just means there will be no 3.5 or 3.6 Pro, and we'll just be going straight to 4 Pro.
00:19:17.359 --> 00:19:18.319
We'll see.
00:19:18.880 --> 00:19:22.000
All right, let's go to our next one.
00:19:22.319 --> 00:19:27.440
And yeah, this one, I'm both excited but a little bummed.
00:19:27.599 --> 00:19:30.880
But this is kind of how Google rolls out some of their products.
00:19:31.039 --> 00:19:39.519
Maybe I'm just being uh a little snobbish in uh, you know, wanting the best tools for where I actually need them.
00:19:39.680 --> 00:19:48.079
But yes, we finally have a new 24-7 personal agent from Google that just works.
00:19:48.240 --> 00:19:50.160
I'm talking about Gemini Spark.
00:19:50.319 --> 00:19:55.599
So not technically new, but new to most everyone.
00:19:55.680 --> 00:20:05.759
Uh, because this did launch in May, but only those two uh on the ultra plan, which is that$200 a month plan uh from Google Gemini.
00:20:05.920 --> 00:20:08.400
So we didn't cover it on the show in May, right?
00:20:08.480 --> 00:20:14.240
For the most part, I only cover uh on the Friday features those that come out to the base paid plans, right?
00:20:14.400 --> 00:20:19.039
Your$20 a month plan, not your plans that are$100,$200 a month.
00:20:19.279 --> 00:20:31.039
So that was launched for ultra subscribers in May, but now Google has released uh Gemini Spark to all pro subscribers as well.
00:20:31.200 --> 00:20:35.519
So that's Google's personal AI agent for automating complex tasks.
00:20:35.680 --> 00:20:36.720
It's rolling out.
00:20:36.880 --> 00:20:38.400
I have access to it right now.
00:20:38.559 --> 00:20:46.960
So it performs multi-step tasks on your behalf, research, planning, complex workflows while you focus on well, whatever else.
00:20:47.200 --> 00:20:50.319
So it's kind of built on these three different concepts.
00:20:50.559 --> 00:20:55.039
So you have your tasks, which is your high-level goal, you have your skills, right?
00:20:55.119 --> 00:21:01.200
The open skills standards, those are your uh reusable instructions with context, and then schedules.
00:21:01.279 --> 00:21:05.519
So those are automated triggers like times or conditions.
00:21:06.160 --> 00:21:06.480
All right.
00:21:06.559 --> 00:21:12.880
So, like I said, if you are on that base uh paid plan, you should get access to it now.
00:21:12.960 --> 00:21:19.279
Uh, if especially if you're in the US, Google did say that other countries are going to be getting access to it soon.
00:21:19.519 --> 00:21:22.960
So here's the downside or the caveat.
00:21:23.440 --> 00:21:26.480
I have no clue when it's going to be rolling out to workspace plans.
00:21:26.640 --> 00:21:29.200
Uh, this is usually one of those things that they say soon.
00:21:29.359 --> 00:21:36.160
So if your business runs on Google, on a paid Google Gemini account, maybe you have access to it, maybe you don't.
00:21:36.240 --> 00:21:42.160
But if you are on a standard paid uh, you know, Google Gemini plan like me, right?
00:21:42.240 --> 00:21:50.720
So my personal Gmail, I have a uh paid Gemini account for my personal Gmail, I have a paid Gemini account for my work emails.
00:21:50.880 --> 00:21:52.319
So on the work side, we don't have this.
00:21:52.400 --> 00:21:58.880
So for me, it's like, well, I don't know how much I'm gonna be using Gemini Spark in my personal Gmail, right?
00:21:58.960 --> 00:21:59.599
I just use it more.
00:22:00.160 --> 00:22:06.160
For testing and for certain purposes in Gemini that I don't have access to in my workspace account.
00:22:06.319 --> 00:22:08.880
Um, so why is it useful?
00:22:09.039 --> 00:22:12.480
Well, it works inside of the Google apps, right?
00:22:12.640 --> 00:22:30.720
So uh any acts, uh any app that you give Spark access to inside of the Gemini workspace, it can essentially go and you know monitor for updates in those apps, uh, you know, push uh updates from you know docs to slides or from your Gmail to your calendar, uh, right.
00:22:30.799 --> 00:22:35.599
So this also works uh with your personal intelligence, that new feature from Google.
00:22:35.759 --> 00:22:42.799
There's a remote browser uh and computer with code execution, chats, canvas, uh, all those different things.
00:22:43.039 --> 00:22:52.240
So an example that Google gave is you know, saying uh Google Doc, the task is to plan and manage your business trip coming up.
00:22:52.400 --> 00:23:05.200
It'll look through your uh, you know, your Gmail, your calendar and your schedule, and it will tell you, oh, your flight is delayed, or you know, a travel booking skill, you know, plus a Gmail writing skill.
00:23:05.359 --> 00:23:09.359
You can rebook the room and send a confirmation or something like that.
00:23:09.519 --> 00:23:22.079
So Spark can now open, read, and edit Google Docs directly, with that expanding to spreadsheets and presentations, uh, including reading comments and editing shared team docs.
00:23:22.240 --> 00:23:27.119
So I think Gemini Spark was previewed a super long time ago.
00:23:27.200 --> 00:23:30.480
Um, and at the time, I was personally very excited.
00:23:30.640 --> 00:23:42.720
But it took obviously many months for it to actually roll out, and it's still, uh at least according to my research, not available for every single paid workspace plan.
00:23:42.960 --> 00:23:52.400
So obviously, this is great if you run uh Gemini in your business, but not all business paid accounts have access to this.
00:23:52.720 --> 00:23:56.000
All pro accounts on the personal Gmail side do.
00:23:56.240 --> 00:24:06.799
So if I'm being honest, six months ago, if this rolled out to my workspace account, I would be talking about Gemini Spark every single hour because it's great.
00:24:07.039 --> 00:24:13.440
But right now, you can do most of these things, anyways, uh, with codecs or clawed code, clawed desktop.
00:24:13.519 --> 00:24:18.000
So that's why I'm not like gonna be, you know, raving about Gemini Spark.
00:24:18.079 --> 00:24:22.880
You know, it's kind of interesting that in a lot of cases, uh, OpenAI anthropic.
00:24:22.960 --> 00:24:32.079
And the crazy thing is, even sometimes Microsoft um offers some of these features and functionalities before Google does with their own products, right?
00:24:32.240 --> 00:24:41.359
So even Microsoft rolled out a uh essentially an open claw-esque integration, and you can connect with Google.
00:24:41.519 --> 00:24:47.200
So uh, although it's exciting, technically Google a little bit behind even in their own backyard.
00:24:47.359 --> 00:24:54.720
Although I do know that millions of people will like this update, and that's why I decided to cover it on the show.
00:24:55.279 --> 00:24:59.519
All right, uh, two more, and we are getting to our big one.
00:24:59.680 --> 00:25:12.480
But before we talk about how uh Chat GPT can now be your Jarvis, we have to talk about the this is the one where uh Anthropic took a uh page out of the codex playbook and made it a little bit better.
00:25:12.640 --> 00:25:19.279
Uh, that is the new record a skill feature uh that just shipped this week in Claude Co-Work.
00:25:19.599 --> 00:25:31.519
So this is essentially you can screen record yourself doing a task once, and then Claude analyzes the recording and turns it into a reusable, re-uh runnable skill.
00:25:32.000 --> 00:25:43.839
So Claude captures screen activity, cursor movements, keystrokes, and the new addition that we don't have just yet in codex, at least not by default, is your voice narration.
00:25:44.079 --> 00:25:52.400
So then uh it processes all of that together uh in the recording and it turns it into a structured skill that saves it to your library.
00:25:52.559 --> 00:25:53.759
So who has access?
00:25:53.920 --> 00:25:57.279
It is rolled out to all paid plans now.
00:25:57.519 --> 00:26:01.680
So um free users can still use existing skills, right?
00:26:01.759 --> 00:26:04.880
But you can't obviously record new ones.
00:26:05.519 --> 00:26:11.839
So it was a little confusing to find this, so it's only available on co-work.
00:26:12.000 --> 00:26:15.440
Um, so you have to go into the chat interface.
00:26:15.599 --> 00:26:19.440
So now obviously Claude Desktop changed their interface a little bit.
00:26:19.599 --> 00:26:21.440
It used to be the three different tabs.
00:26:21.519 --> 00:26:23.519
It was chat, cowork, and code.
00:26:23.680 --> 00:26:27.839
Now it says chat and code, but in chat there's a co-work tab, right?
00:26:28.000 --> 00:26:31.039
So you have to be in chat, you have to go find your co-work tab.
00:26:31.119 --> 00:26:35.599
You hit the plus button, and then you can find uh the new recorder skill.
00:26:35.680 --> 00:26:37.839
So it actually took me a little bit of clicks to find it.
00:26:37.920 --> 00:26:39.039
I'm like, wait, where is this?
00:26:39.119 --> 00:26:40.240
So you do need to update this.
00:26:40.319 --> 00:26:41.680
This is desktop only.
00:26:41.839 --> 00:26:50.240
So update your Claude desktop app, go into chat, click the cowork tab, click the plus button, and then and actually super useful.
00:26:50.319 --> 00:26:51.920
So I did a little bit of testing.
00:26:52.000 --> 00:26:53.359
Uh, it's really good.
00:26:53.519 --> 00:26:54.799
So, why is it useful?
00:26:54.960 --> 00:27:01.839
Well, until now, creating a skill meant either chatting with Claude and having it create one for you.
00:27:02.000 --> 00:27:14.559
You know, you can obviously create those skill MD files uh by hand, but so many times it's just like at the end, uh, and and and you should be doing this FYI if you're not already doing this.
00:27:14.799 --> 00:27:28.240
Um, the way I look at new chats or tasks in any uh, you know, large language model, especially any agentic one, when I'm done and I get an output where I'm like, this is good, you should be turning that into a skill, right?
00:27:28.400 --> 00:27:36.799
So this is instead of having to work through it or to hand hold right Claude or Codecs or anti-gravity or anything, right?
00:27:36.880 --> 00:27:39.680
At the end, you would normally say turn this into a skill.
00:27:39.839 --> 00:27:44.480
So the difference here with the teach Claude a skill is well, you don't have to work through it.
00:27:44.640 --> 00:27:52.240
You just click start recording, you do your work, you dictate through it, you say, Here's what I'm doing, A, B, and C.
00:27:52.400 --> 00:27:55.839
I'm opening this site, here's what I'm looking for on this page.
00:27:56.000 --> 00:27:59.039
I download it, I open it in this program, right?
00:27:59.119 --> 00:28:03.200
And it'll just record all of that and turn it into a skill.
00:28:03.359 --> 00:28:13.759
So, like I said, technically taking a uh page out of the codex uh replay uh or no record and replay skill uh that did this exact same thing.
00:28:13.839 --> 00:28:18.400
But the downside with codexes is by default there is no voice narration, right?
00:28:18.480 --> 00:28:26.400
There is kind of some workarounds um around that, but it is nice that in this new one for Claude Cowork, it is just enabled by default.
00:28:26.480 --> 00:28:33.440
You don't have to have any workarounds to actually narrate your way through that skill, uh, which is very helpful, obviously.
00:28:33.680 --> 00:28:35.440
So, who's gonna find this valuable?
00:28:35.599 --> 00:28:37.200
Well, it literally anyone.
00:28:37.279 --> 00:28:42.720
So if you're a Claude desktop user and you use Claude Cowork, it's gonna be great.
00:28:42.799 --> 00:28:47.440
Um, I would love to see uh Anthropic roll this out for Claude Code as well.
00:28:47.599 --> 00:28:49.599
Um, I think that would be really helpful.
00:28:49.759 --> 00:29:06.640
Uh, just because seemingly, um, I'm guessing Claude Cowork um is not kind of riding that same wave of popularity that it was in earlier uh 2026, just because the the interface kind of got changed over a little bit.
00:29:06.799 --> 00:29:12.880
And I think more and more people are starting to use Claude Code and just more of the agentic features in the normal chat section.
00:29:13.039 --> 00:29:16.079
So you can only set this up in the cowork, which is a downside.
00:29:16.160 --> 00:29:21.599
But literally anyone that uses Claude and you're doing the same processes over and over.
00:29:21.920 --> 00:29:23.519
I mean, this is big, right?
00:29:23.599 --> 00:29:30.160
I love, I absolutely loved and use the this codec skill um all the time for my workflows.
00:29:30.240 --> 00:29:34.079
So I'm personally happy uh to see this in uh claude code.
00:29:34.240 --> 00:29:37.039
The other thing, here's the here's the big unlock, y'all.
00:29:37.200 --> 00:29:40.319
You can do this in I talked about this in codecs as well.
00:29:40.400 --> 00:29:50.960
Uh, you can do this in Claude Code or Codex, go through, record the skill, and skills are shareable and they're fairly uh openly supported across all different platforms.
00:29:51.039 --> 00:30:03.039
So you can just create a skill, uh, very detailed skill as an example using this new feature from Anthropic, and then use that skill inside of Google or use that skill inside of Codex or Chat GPT.
00:30:03.519 --> 00:30:07.119
All right, and our last big feature.
00:30:07.279 --> 00:30:14.799
And y'all, this this one, this is one of those instances where I'm like, okay, this is crazy after I tried uh this.
00:30:14.880 --> 00:30:18.079
So yeah, hasn't even been out a full 24 hours.
00:30:18.240 --> 00:30:25.119
Uh, but open AI went to full Jarvis mode and they launched Chat GPT voice on desktop.
00:30:25.200 --> 00:30:30.000
And it's a lot more than it sounds like, but let me go through the bullet point details first.
00:30:30.160 --> 00:30:43.680
So OpenAI launched ChatGPT Voice in the desktop app, uh, powered by GPT Live, which I already referenced, that lets you talk through work and coordinate tasks across chat work and codecs.
00:30:43.920 --> 00:31:00.000
So per OpenAI announcement, you can control your computer and direct multiple agents running in Chat GPT work or codex using just your voice with GPT Live, speaking, listening, and coordinating work simultaneously.
00:31:00.319 --> 00:31:04.559
So here is how open AI kind of describes the main benefits.
00:31:04.720 --> 00:31:13.200
So it says you can work across projects, you can start a new task, coordinate work in progress, and pick up where you left off without micromanaging.
00:31:13.359 --> 00:31:21.039
You can work across your desktop, you can use your files, apps, and connected tools from Slack and GitHub to Notion and beyond.
00:31:21.200 --> 00:31:23.519
And you can explore, plan, and learn.
00:31:23.599 --> 00:31:30.480
You can talk through a pull request, ask questions about a code base, or learn a new topic through natural back and forth.
00:31:30.640 --> 00:31:31.839
So who has access?
00:31:32.000 --> 00:31:40.480
Well, if you are on any paid plan and you're using the desktop version of Chat GPT Workslash Codex, it's the same thing.
00:31:40.720 --> 00:31:53.039
It is out now, and it works also with uh ChatGPT's remote feature, which is pretty cool because you don't even need to be in front of your computer, you can just have your phone, right?
00:31:53.119 --> 00:32:11.279
I have one uh right, I'm traveling right now, I'm not in Chicago, but I have essentially a home Mac Studio that never gets shut off, and I can be controlling it right now and just with my voice, and it can be doing literally anything, and I can be watching and seeing it go in real time.
00:32:11.680 --> 00:32:20.559
So this just kind of brings that uh GPT live mode that was announced about two weeks ago uh to the desktop.
00:32:20.640 --> 00:32:34.960
So before that was only available via mobile, so it doesn't just bring it to the desktop, but this is the first time where I'm like, wait, you technically don't even need to type or use your mouse anymore.
00:32:35.119 --> 00:32:45.920
Then yes, I haven't been able to run it through many hours of demos, but um the possibilities for this one are absolutely bonkers, y'all.
00:32:46.079 --> 00:32:47.359
And it is so good.
00:32:47.519 --> 00:32:49.039
So let me talk about a couple things.
00:32:49.200 --> 00:32:52.240
Number one, it works with ChatGPT's remote feature, which is great.
00:32:52.319 --> 00:33:02.079
That's just on your mobile app, you can click remote and you can technically uh control any computer that you have connected to that chat GPT account.
00:33:02.240 --> 00:33:03.119
So that's cool.
00:33:03.279 --> 00:33:05.200
But here's the other thing app shots.
00:33:05.519 --> 00:33:05.839
All right.
00:33:06.000 --> 00:33:23.519
So what app shots are, if you're you know a power, you know, codex or chat GPT work user, the default uh, you know, is the two command keys and it takes essentially a screenshot, but not just a screenshot, but every single thing it brings into context that you have um in that program.
00:33:23.680 --> 00:33:30.240
So as an example, right now on my screen, um, I have up some notes, some bullet points that I wanted to go over, right?
00:33:30.319 --> 00:33:31.920
But it's a very long list.
00:33:32.160 --> 00:33:37.440
So obviously what's on my screen is probably only like five percent of what's in that document.
00:33:37.599 --> 00:33:50.799
So if I use the app shots, it's gonna obviously take a screenshot of what is actually on my screen, but it brings in the context of everything else in uh that program that I have up that is not even on the screen.
00:33:50.960 --> 00:33:56.160
So the cool thing is with this new uh, you know, I kind of wish we could have got a fun name for it.
00:33:56.240 --> 00:33:57.359
I know they can't use Jarvis.
00:33:57.519 --> 00:33:58.960
I'm just using Jarvis, right?
00:33:59.200 --> 00:34:07.680
Like you can have this app shot feature and functionality while you're using Chat GPT uh live voice on desktop.
00:34:08.000 --> 00:34:21.280
So in my kind of playing around, I had, you know, six or seven different programs open on my computer, hands-free, not typing and just telling Chat GPT work or codecs what to do.
00:34:21.440 --> 00:34:22.480
Open this file.
00:34:22.639 --> 00:34:23.519
Okay, great.
00:34:23.679 --> 00:34:26.079
Um, can you find this change that I made?
00:34:26.239 --> 00:34:27.519
It's probably somewhere in the middle.
00:34:27.599 --> 00:34:41.440
And it can go, it can find that using the app shots, uh, right, bringing in all that other context without having me to always tell it, oh, somewhere in this document, you know, I I think that I outlined uh something about that that certain KPI, right?
00:34:41.519 --> 00:34:43.760
It's all there and it's absolutely bonker.
00:34:43.920 --> 00:34:48.880
So uh, you know, why we keep saying Jarvis is this is literally the scene out of Iron Man, right?
00:34:49.039 --> 00:34:56.239
When Tony, uh Tony Stark is at his computer just saying, Hey, Jarvis, you know, pull up this, go do this, right?
00:34:56.320 --> 00:34:59.519
Now we actually have this for the first time and it works.
00:34:59.679 --> 00:35:09.360
So, not just having the remote is cool, the app shots is kind of that uh that hack, but what I'm really excited about is for the future of this.
00:35:09.519 --> 00:35:16.079
So, obviously, open AI's uh computer use with the new GPT 5.6 model took a huge leap forward.
00:35:16.159 --> 00:35:24.639
But the thing I'm actually excited about is when we get that super blazing fast uh uh Cerebras uh integration.
00:35:24.719 --> 00:35:28.559
So, you know, OpenAI did say that it was coming in July.
00:35:28.880 --> 00:35:43.440
So when we can bring this with uh this ridiculously fast uh version of GPD 5.6, what you are able to accomplish in front of your computer without even sitting down is amazing.
00:35:43.599 --> 00:35:52.000
So I don't know, maybe maybe this is gonna bring back the popularity of having like a treadmill desk or something like that, or just being able to go touch grass.
00:35:52.079 --> 00:35:57.039
I'm like, okay, is this gonna also, you know, the XR glasses and all these other things?
00:35:57.199 --> 00:36:04.159
I'm actually geeking out and excited about this one because I don't know, like I don't like sitting in front of a screen all day.
00:36:04.320 --> 00:36:07.039
I like getting up and walking around a lot, right?
00:36:07.119 --> 00:36:11.119
And obviously, you can have voice dictation apps, but that's different.
00:36:11.199 --> 00:36:15.920
So this is just a completely different way to work, a different way to interface with smart AI.
00:36:16.079 --> 00:36:17.920
It's one I'm super excited about.
00:36:18.000 --> 00:36:19.599
So this is useful to anyone.
00:36:19.679 --> 00:36:39.199
So if you're using uh the new Chat GPT work desktop app, which is essentially the uh developer or the non-uh technical version of codex, uh, this is in theory a way to completely change how you interface with the computer, not just how you use Chat GPT and how you work, but literally how you interface with a device.
00:36:39.360 --> 00:36:41.760
So um I'm geeked out about that one.
00:36:41.920 --> 00:36:43.920
I am too, or I hope you are too.
00:36:44.079 --> 00:36:46.960
Maybe I'll do a uh Wednesday demo of that one.
00:36:47.119 --> 00:36:51.039
So that's a wrap on all the things that are new this week.
00:36:51.119 --> 00:36:58.880
Let me give you a very quick update of our seven AI features you can use today uh if you have paid plans.
00:36:59.039 --> 00:37:02.159
So ChatGPT Health is out and launched.
00:37:02.320 --> 00:37:04.960
Claude has some nice voice mode upgrades.
00:37:05.039 --> 00:37:11.199
You no longer have to talk to the dumber haiku model, you can talk to the smart uh opus and sonnet.
00:37:11.280 --> 00:37:15.039
Uh, we have the new Microsoft MAI image 2.5 Pro.
00:37:15.440 --> 00:37:21.119
Uh Google released Gemini 3.6 Flash and Gemini 3.5 Flash Lite.
00:37:21.280 --> 00:37:23.519
Uh, we got the new Gemini Spark.
00:37:23.599 --> 00:37:29.519
Uh Google's always on 24-7 AI agent that is available now to pro users.
00:37:29.760 --> 00:37:31.519
Anthropics, very useful.
00:37:31.599 --> 00:37:38.079
Claude, record a skill, uh, including that voice, uh, being able to narrate your skill, which is great.
00:37:38.239 --> 00:37:43.760
And then last but not least, openai went full Jarvis mode with Chat GPT voice on desktop.
00:37:43.840 --> 00:37:44.639
I hope this was helpful.
00:37:44.719 --> 00:37:45.760
If so, do me a favor.
00:37:45.840 --> 00:37:53.440
If you're not already, please subscribe to the podcast on Spotify or Apple, and then go to your everydayai.com.
00:37:53.519 --> 00:37:58.480
We're going to be recapping the highlights from today's show, anything you missed, and everything else in today's newsletter.
00:37:58.559 --> 00:37:59.679
So thank you for tuning in.
00:37:59.760 --> 00:38:02.960
We'll see you back Monday and every day for more everyday AI.
00:38:03.039 --> 00:38:04.079
Thanks, y'all.
