WEBVTT
00:00:02.560 --> 00:00:10.240
AI just got faster, more accessible, and cheaper all in the same week, but there's a good chance you might have missed all of it.
00:00:10.480 --> 00:00:13.359
I mean, let's see what happened just recently.
00:00:13.599 --> 00:00:22.000
Google dropped its new Gemini 3.7 flash at half the price of the version that they just released three weeks ago.
00:00:22.160 --> 00:00:27.120
Then Elon Musk's team shipped a smarter Grok at the exact same price.
00:00:27.280 --> 00:00:34.880
And OpenAI answered with a new mode that runs its most powerful model 14 times faster.
00:00:35.119 --> 00:00:38.079
I mean, I still don't even know how that's even possible.
00:00:38.240 --> 00:00:50.000
But here's why you should care, even if you've never touched an API, because price and speed are the two things standing between AI demos and AI that actually is running inside of your business.
00:00:50.159 --> 00:00:56.960
Because every time one of those two things drops, a use case that didn't make sense last month suddenly makes a ton of sense.
00:00:57.119 --> 00:00:58.560
And that was just the model news.
00:00:58.719 --> 00:01:10.079
I mean, Claude started working inside your browser tabs, your spreadsheets can now turn themselves into live dashboards, and ChatGPT quietly shipped a feature that watches everything you do on your Mac.
00:01:10.319 --> 00:01:11.519
And I'm here for it.
00:01:11.680 --> 00:01:24.319
So that's why on Fridays we bring you Friday features, our weekly update of the latest AI features that you can actually use today, not announcements, teasers.
00:01:24.480 --> 00:01:27.519
Yeah, you can go use all these things right now.
00:01:27.680 --> 00:01:29.280
All right, let's get into it.
00:01:29.439 --> 00:01:31.200
Welcome to Everyday AI.
00:01:31.359 --> 00:01:34.159
If you're new here, my name's Jordan Wilson, and we do this every day.
00:01:34.239 --> 00:01:45.120
It's a daily, unedited, unscripted live stream, podcast, and free dealing newsletter helping business leaders like you and me not just keep up with everything that's happening, but I tell you what matters, what doesn't, how to use it.
00:01:45.200 --> 00:01:47.680
And you take that information to grow your company and your career.
00:01:47.840 --> 00:01:50.640
All of a sudden it's like, whoa, you're the smartest person in AI.
00:01:50.719 --> 00:01:52.560
Everyone's looking at you like, what's your secret?
00:01:52.719 --> 00:01:57.519
Well, your secret is this, but also our website, your everydayai.com.
00:01:57.680 --> 00:02:00.799
So if you didn't know, uh you can go check that out.
00:02:00.959 --> 00:02:07.120
We have literally 830 plus episodes you can go watch, listen to, and read about all for free.
00:02:07.200 --> 00:02:10.479
It is a free generative AI university on our website.
00:02:10.639 --> 00:02:13.840
So uh let's get into it on today's show.
00:02:14.000 --> 00:02:14.879
Here's what you're gonna learn.
00:02:14.960 --> 00:02:23.039
You're gonna know the new Mac feature that watches everything you do at work, then tells you what you did all day and maybe finishes the work that you forgot to do.
00:02:23.280 --> 00:02:29.360
You're gonna know why the smartest cheap model in AI just got 50% cheaper.
00:02:29.599 --> 00:02:32.159
And the catch that's hiding in the fine prints.
00:02:32.400 --> 00:02:38.560
And you're gonna know how Claude can now do real work inside your browser tabs, login and all.
00:02:38.800 --> 00:02:40.879
All right, so let's take a look.
00:02:41.039 --> 00:02:43.840
Actually, live stream audience, haven't shouted you out in a bit.
00:02:44.000 --> 00:02:45.520
Big bogey face, good to see you.
00:02:45.680 --> 00:02:54.159
Brent joining us from London, Brian from Minnesota, Jose joining us from Santiago, uh, Miko joining us from Tokyo.
00:02:54.319 --> 00:02:56.719
We're we're all over the place this morning.
00:02:56.879 --> 00:02:57.759
Uh, love to see it.
00:02:57.840 --> 00:02:59.840
Michelle, good to good to see you join us.
00:03:00.000 --> 00:03:02.560
All right, a lot of new AI features, y'all.
00:03:02.639 --> 00:03:09.599
So on Fridays, it's one of my favorite shows, and it's actually been one of our uh more listened to shows because yeah, it's hard to keep up.
00:03:09.680 --> 00:03:14.800
And some of these things literally, I I had the show planned last night, woke up this morning.
00:03:14.879 --> 00:03:16.479
I'm like, all right, there's already new things.
00:03:16.560 --> 00:03:18.719
You know, some things had to get the cut, right?
00:03:18.800 --> 00:03:23.280
It's kind of crazy because there's some big updates this week this week that didn't even make our top seven.
00:03:23.360 --> 00:03:27.280
I mean, uh what let me just read these before we get into the ones we're actually going to cover.
00:03:27.439 --> 00:03:32.639
I mean, the Microsoft Copilot started unifying its co-pilot offerings into the super app.
00:03:32.719 --> 00:03:35.360
So that's actually happening, wasn't even a top seven thing.
00:03:35.439 --> 00:03:41.039
Uh, ChatGPT upgraded its voice mode to include projects and file uploads, which is pretty big.
00:03:41.120 --> 00:03:42.879
I've been using that one a lot since it came out.
00:03:43.039 --> 00:03:51.599
Uh, Google Gemini released 13 new connectors, and Google also released a slew of new AI features for Google Ads and Google Analytics.
00:03:51.759 --> 00:03:55.599
So let's get into the ones that we're actually going to be covering today.
00:03:55.759 --> 00:04:05.840
And let's start with well, if you care specifically about the fastest, cheapest model that is still the most intelligent.
00:04:06.080 --> 00:04:11.599
Technically, if you look at things like artificial analysis, they actually have a nice little uh model filter.
00:04:11.759 --> 00:04:18.079
And if you say those are the three things that are most important for you, the new Gemini 3.7 flash is the recommended model.
00:04:18.240 --> 00:04:19.839
So it is brand new.
00:04:19.920 --> 00:04:25.040
And if you're wondering, like, wait, Jordan, didn't you just cover this new Gemini Flash like very recently?
00:04:25.120 --> 00:04:26.560
Yeah, three weeks ago.
00:04:26.639 --> 00:04:32.959
That's how long it took Google to go from Gemini 3.6 Flash to Gemini 3.7 Flash.
00:04:33.040 --> 00:04:34.480
So uh let's see what's new.
00:04:34.639 --> 00:04:44.000
Google released Gemini 3.7 Flash, its new AI model that it calls the most intelligent workhorse model uh for coding and agents.
00:04:44.079 --> 00:04:51.279
And it's a big step up from the version released just three weeks ago with major gains in coding, debugging, and building web apps.
00:04:51.439 --> 00:04:52.959
So who has access to this?
00:04:53.120 --> 00:04:56.560
If you're a paid subscriber, it's available already.
00:04:56.720 --> 00:05:01.519
So uh interestingly enough, it's not yet available inside of Google Gemini, right?
00:05:01.600 --> 00:05:11.759
So Gemini.google.com, but it is available in its Gemini Spark model, which is available, uh, the uh kind of agentic offering for Google, which is available in Gemini.
00:05:12.000 --> 00:05:13.199
I know that's a little confusing.
00:05:13.360 --> 00:05:21.839
Uh, but if you go into Google's AI studio, it's in there, as well as Android Studio and the Gemini API and Gemini Enterprise platform.
00:05:21.920 --> 00:05:31.680
So yeah, if you're a normal paid Gemini user, which is I think the majority of our listeners, and you go in there, you're not gonna find it there, but it is powering the Gemini Spark.
00:05:31.759 --> 00:05:38.560
And if you're using Google Anti-Gravity, so many different ways that you can access Google's uh uh AI.
00:05:38.639 --> 00:05:40.720
It's like you need a roadmap sometimes.
00:05:40.879 --> 00:05:50.879
All right, but here's the key thing API pricing is 75 cents per million input tokens and$3.75 per million output tokens.
00:05:51.199 --> 00:05:56.560
All right, so let's read just quick uh kind of what Google said about their newest release.
00:05:56.720 --> 00:05:58.240
Let me zoom in on my browser here.
00:05:58.399 --> 00:06:00.399
Apparently, I'm old and can barely see.
00:06:00.560 --> 00:06:11.839
All right, so Google says today we're building on the progress of our widely used Flash series by introducing Gemini 3.7 Flash, our most intelligent workhorse model yet for coding and agents.
00:06:12.000 --> 00:06:22.399
The release comes just three weeks after Gemini 3.6 Flash and is a direct result of developer feedback and algal uh algorithmic uh innovations that we look forward to bringing to future models.
00:06:22.560 --> 00:06:34.800
3.7 Flash delivers substantial improvements across software engineering, knowledge work, and web development workflows with an introductory price of half of the original uh 3.6 flash cost per million tokens.
00:06:35.040 --> 00:06:42.560
So if you recall uh when Gemini 3.6 Flash came out, I said I don't understand this model.
00:06:42.720 --> 00:06:50.560
Uh, because the whole point of the Flash series, all along from Google, has been to be extremely fast and extremely cheap.
00:06:50.720 --> 00:06:56.720
And Gemini 3.6 Flash, although it was fast-ish, it was not cheap, right?
00:06:56.800 --> 00:07:07.040
It was not its normal, you know, you would look at Gemini's Flash series and be like, all right, this is, you know, not like free, but this is a very low-cost model that you can put out at scale.
00:07:07.199 --> 00:07:12.399
So uh unfortunately, I think a lot of people were upset at Gemini 3.6 Flash because it wasn't really that.
00:07:12.639 --> 00:07:16.800
Gemini 3.7 Flash gets back to exactly that.
00:07:17.040 --> 00:07:20.800
So uh who's gonna find this useful?
00:07:20.959 --> 00:07:26.800
Well, it's useful because it got smarter, faster, and the price cut in half.
00:07:26.879 --> 00:07:33.120
Again, that's if you're paying on the uh API side, but the real thing is inside Google Gemini Spark.
00:07:33.199 --> 00:07:40.160
So if you haven't used Spark, that's kind of their uh new ish autonomous coding agent that is inside of Google Gemini.
00:07:40.319 --> 00:07:44.000
So it's kind of a mode, it's an agentic mode inside of Google Gemini.
00:07:44.240 --> 00:07:46.879
So for that, it's gonna make, I think, a big difference.
00:07:47.040 --> 00:07:49.199
So I tested out Gemini Spark.
00:07:49.360 --> 00:07:59.519
I wasn't super impressed, uh, you know, when it first came out to paid plans, mainly just because I've been using, you know, codex and you know, claude code, claude co-work for so long.
00:07:59.600 --> 00:08:07.519
And I'm like, yeah, I'm not really seeing anything that I can't do inside of those, but now with a better model, I think some people are gonna find it much more valuable.
00:08:07.759 --> 00:08:09.600
So who's gonna find this valuable?
00:08:09.759 --> 00:08:19.519
Well, if you're a business using AI on high via the API, uh, this is might be your new workhorse model, right?
00:08:19.600 --> 00:08:23.680
Uh, people don't really understand from a serving tokens perspective.
00:08:23.839 --> 00:08:28.240
Uh, Google is and has always been a leader in the space, right?
00:08:28.319 --> 00:08:37.919
Even though, even if you don't see a lot of people talking about the Gemini chatbot, when it comes to serving up tokens, uh, right, Google has been a leader and will continue to be.
00:08:38.159 --> 00:08:38.720
All right.
00:08:39.360 --> 00:08:43.039
Uh, so oh, Steven Leeds says uh he's been waiting for this one.
00:08:43.120 --> 00:08:43.440
All right.
00:08:43.519 --> 00:08:47.120
Hopefully you enjoy uh Gemini 3.7 flash.
00:08:47.279 --> 00:08:49.279
All right, let's get into our next one.
00:08:49.360 --> 00:08:55.039
And this is the one that just came out like kind of like an hour, a couple hours ago.
00:08:55.200 --> 00:09:00.320
So, yeah, we may have a new King of the Open Source Hill.
00:09:00.480 --> 00:09:09.919
Yeah, we don't even have full benchmarks for it, so we don't know if it's gonna truly be the best open weights model in the world, but it might be.
00:09:10.080 --> 00:09:15.840
That's because we have a new version of ZAI's GLM5.
00:09:16.639 --> 00:09:23.600
So a new model it's calling, it's best yet for coding, and the strongest open weights coding models that it has measured.
00:09:23.759 --> 00:09:32.480
So uh an interesting wrinkle, it's the same underlying model as GLM 5.2, but every single gain just comes from better post-training.
00:09:32.559 --> 00:09:35.519
So, yeah, not like a completely new model per se.
00:09:35.600 --> 00:09:40.080
This is more of a step change that's very um, very normal.
00:09:40.240 --> 00:09:48.879
But the big surprise here in the big jump, at least according to ZAI's benchmarks, like I said, it's too fresh to have any third-party uh benchmarks yet.
00:09:49.039 --> 00:09:50.399
Big jump is cybersecurity.
00:09:50.480 --> 00:10:02.720
So ZAI says the model got better at finding and chaining software vulnerabilities than the company intended, and it has already found over a thousand critical security flaws in real software like Linux and Apple's web kits.
00:10:02.799 --> 00:10:12.799
Yeah, ones that have already uh, you know, according to ZAI, withstood, you know, the fable and the GPT-5.6 uh, you know, vulnerabilities that everyone has found.
00:10:12.960 --> 00:10:19.679
The thing that's interesting here, you know, so for our live stream audience, I have uh kind of their blog post announcement up on my screen.
00:10:19.759 --> 00:10:27.600
You can always watch the video version of our podcast on uh your everydayai.com, as well as you can go catch that link in the show notes.
00:10:27.759 --> 00:10:32.559
But the interesting thing was, yeah, uh some of these benches, right?
00:10:32.720 --> 00:10:44.639
I mean, they're they're definitely um they're definitely competing with the closed source proprietary models, specifically, you know, Fable 5 from Anthropic and GBD56 Soul.
00:10:44.879 --> 00:10:47.919
So the one that really stood out is Cyber Gym.
00:10:48.080 --> 00:10:48.399
All right.
00:10:48.480 --> 00:10:53.440
So this is kind of uh a metric on the cybersecurity side, which is very important.
00:10:53.600 --> 00:10:56.000
I talked about that uh, you know, this week.
00:10:56.159 --> 00:11:01.120
Was that this week that I had my uh yeah, my my rogue AI agent show?
00:11:01.200 --> 00:11:02.320
It's all blending together now.
00:11:02.399 --> 00:11:09.759
But you know, let's go back to when Anthropic teased uh its mythos and fable models, and then they sat on it, didn't release it for like three months.
00:11:10.000 --> 00:11:15.120
The main uh the main benchmark that it led with was Cyber Gym.
00:11:15.360 --> 00:11:15.919
All right.
00:11:16.000 --> 00:11:22.000
Uh and right now, you know, it is a pretty important uh benchmark, you know, Cyber Gym and Exploit Gym.
00:11:22.080 --> 00:11:28.559
Uh, but for Cyber Gym, this new open model is the best in the world, right?
00:11:29.120 --> 00:11:30.799
Which is crazy, right?
00:11:31.200 --> 00:11:34.879
Yeah, an open, uh, an open model uh that is the best.
00:11:35.039 --> 00:11:46.639
So it scored an 84.5 on Cyber Gym, uh ahead of the other two models uh that were kind of tied there for first place previously, which is Fable 5 and GPT 5.6 soul.
00:11:46.879 --> 00:11:51.679
So a little bit more about GLM5.3 and who has access.
00:11:51.840 --> 00:11:53.759
Well, it's available to everyone now.
00:11:53.840 --> 00:12:06.960
So if you are on ZAI's GLM coding plan, or if you have their Z code installed, which Z code is pretty much just a pixel per pixel copy of Codex, anyways.
00:12:07.039 --> 00:12:13.279
Well, if you use Codex and you're like, Oh, I want to try this new GLM5.2, you can try Z code and it's pretty much the same thing as Codex.
00:12:13.519 --> 00:12:17.039
Uh, but it's been rolled out to all uh subscribers.
00:12:17.200 --> 00:12:20.000
So the open weights are not out yet.
00:12:20.159 --> 00:12:25.039
All right, so ZAI says it will release them in about two weeks after further safety testing.
00:12:25.120 --> 00:12:28.559
Uh, so calling it open source today is a little premature.
00:12:28.639 --> 00:12:36.080
It will be open weights, and uh, we at least know that GLM uh sorry, ZAI has a positive track record of doing this.
00:12:36.159 --> 00:12:47.440
This is what happened with GLM 5.2, this is what happened with Kimi K3, with Quen uh, you know, 3.8, all the Chinese open models they've released, and then they usually will release the weights, you know, in about two weeks.
00:12:47.519 --> 00:12:53.840
So technically, the weights are not available today if you're super dork and care about those things, uh, but they will be available soon.
00:12:53.919 --> 00:12:55.120
So, why is this useful?
00:12:55.279 --> 00:12:58.159
Uh, well, it's been a budget favorite.
00:12:58.240 --> 00:13:07.679
So the GLM coding plan has been a budget favorite for AI coding, and this makes the budget option meaningfully better at complex long-running tasks.
00:13:07.840 --> 00:13:13.759
So, uh, the company said on its own tests, it's roughly 50% better at coding than the last version, which is crazy.
00:13:13.919 --> 00:13:30.080
And if you recall back a couple of months when GLM 5.2 was released, this was kind of the first open model that people started to look at and say, wait, has China kind of closed the gap on US uh frontier proprietary closed models?
00:13:30.159 --> 00:13:31.759
And it was like kind of right.
00:13:31.840 --> 00:13:39.200
Um, I think before it was always, you know, US models were tier one and Chinese open models were tier two.
00:13:39.519 --> 00:13:48.000
So the US models are still ahead, but it's now it's one A and one B with the Chinese open models.
00:13:48.080 --> 00:14:05.759
So GLM, Kimi K3, uh and Quen 3.8 are probably in that 1B tier uh with you know Anthropics, uh Opus V and Fable 5, along with um OpenAI's G uh GPT 5.6 soul are kind of still in that 1A tier by themselves.
00:14:06.000 --> 00:14:09.120
So uh pretty, pretty interesting now.
00:14:09.200 --> 00:14:15.440
Yeah, Josh here on YouTube, good comment said actually OpenAI now has new cybersecurity daybreak blue and red.
00:14:15.600 --> 00:14:16.320
That is correct.
00:14:16.399 --> 00:14:17.679
They rolled that out earlier this week.
00:14:17.759 --> 00:14:20.320
If you read our newsletter, you caught that as well.
00:14:20.480 --> 00:14:22.960
All right, more new models.
00:14:23.200 --> 00:14:24.799
Do you see a pattern here?
00:14:24.960 --> 00:14:26.320
A lot of new models this week.
00:14:26.399 --> 00:14:28.559
They're coming out faster and faster than ever.
00:14:28.720 --> 00:14:36.559
And this one, I was kind of surprised with Grock 4.6 is actually a pretty good model, right?
00:14:36.639 --> 00:14:40.799
It's probably joined that one that you know, 1B uh tier.
00:14:40.879 --> 00:14:43.360
You know, all of a sudden it is a top contender.
00:14:43.600 --> 00:14:47.039
So let's talk about what's new uh in the model.
00:14:47.200 --> 00:15:02.399
So Space XAI, yes, formerly XAI, they just jammed all the names together, but the real name is Space XAI, released Grock 4.6, its newest top-tier AI model built for coding and long-running automated tasks.
00:15:02.559 --> 00:15:18.000
So uh on a widely watched independent scorecard that is uh in a lot of the benchmarks that artificial analysis covers, it now ties open AI's GPT 5.6 soul in a lot of important benchmarks and jumped five points over the previous GROC.
00:15:18.159 --> 00:15:24.080
So, yeah, on artificial uh analysis index, it is right there neck and neck with GPT 5.6 soul.
00:15:24.399 --> 00:15:27.039
So who has access to this right now?
00:15:27.200 --> 00:15:36.480
Well, developers have it available through the XAI API, and it's also included in the coding tool cursor on all plans.
00:15:36.639 --> 00:15:43.279
So, yeah, uh cursor obviously uh space XAI uh acquired cursor.
00:15:43.519 --> 00:15:47.679
Um, so this is available if you are on a cursor plan as well.
00:15:47.840 --> 00:15:56.480
It's also now the default model in Grok build, which is you know XAI's attempt at building a codex or clawed code competitor.
00:15:56.639 --> 00:15:57.759
So, why is it useful?
00:15:57.840 --> 00:16:01.200
Well, the price didn't change from the last version of Grok.
00:16:01.600 --> 00:16:06.320
So anyone using Grok got a smarter model for the same money if you are paying per tokens.
00:16:06.399 --> 00:16:15.279
But the biggest thing is this keeps the price war alive uh because the jump from Grok 4.5 to 4.6 was huge, right?
00:16:15.440 --> 00:16:24.159
Uh, maybe some of that compute that the company has been investing in and the cursor acquisition is starting to pay off in a model that maybe now companies will start to look at.
00:16:24.320 --> 00:16:34.320
Um, I'm kind of surprised, maybe that SpaceX AI isn't just going with the cursor brand name and going with, you know, Composer 3 or something like that.
00:16:34.559 --> 00:16:37.600
I will say this, and this isn't my own personal bias.
00:16:37.679 --> 00:16:39.360
This is just literally anyone on the internet.
00:16:39.679 --> 00:16:52.320
I think because of some of the areas that um, you know, Grok has uh made a niche in, people from a business perspective, when it comes to paying, you know, tokens, they're not always looking at Grok as a competitor.
00:16:52.399 --> 00:16:56.879
They're like, oh, this is the one that's you know been doing some you know wild things on the internet.
00:16:57.279 --> 00:17:05.599
Um, so we'll see if now you know Grok 4.6 kind of teetering between that 1B and 1A tier changes that.
00:17:05.759 --> 00:17:11.359
But here is what the company says and said in its release: it said today we are releasing grok 4.6.
00:17:11.519 --> 00:17:21.119
Grock 4.6 builds on GROC 4.5 with a particularly uh with a particular focus on long-running agents and more ambitious interactive and visual work.
00:17:21.279 --> 00:17:33.839
It stays with complex tasks across many steps, whether researching a topic, analyzing information, working across a code base, or turning an idea into a polished application or work artifact.
00:17:34.000 --> 00:17:47.279
Uh, the the the other thing that's kind of related to this Grok 4.6, and we won't be covering this on today's Friday features, even though I would have liked to, but it's only available on the very expensive Grok and cursor plans.
00:17:47.359 --> 00:17:52.880
But the new Grok bot, which does look exciting, and I'm looking forward to it being released on the standard base plan.
00:17:53.119 --> 00:18:05.440
So, yeah, on the Friday show, we're usually not covering things that are only exclusively on the you know 200 or 300 plans, which is grok bot for now, but grok bot is powered by grok 4.6.
00:18:05.519 --> 00:18:12.799
So pretty big real uh pretty big week of releases from the SpaceX slash grok slash cursor teams.
00:18:13.519 --> 00:18:17.279
All right, uh, let's move on to our next one.
00:18:17.440 --> 00:18:21.599
And this one, I don't know, I'm I'm a Google Sheets dork.
00:18:21.920 --> 00:18:36.559
Um, so maybe I'm more excited about this one than the average person, but uh yeah, Canvas mode, one of my favorite modes of all time uh from Google Gemini is coming and it's here inside of Google Sheets.
00:18:36.640 --> 00:18:38.240
So let's talk about this feature.
00:18:38.319 --> 00:18:39.759
You can go use right now.
00:18:40.000 --> 00:18:47.119
Google expanded access to its Sheets Canvas, a feature that turns a plain spreadsheet into an interactive dashboard.
00:18:47.279 --> 00:18:56.079
So, yeah, any uh Google Sheet that you have inside of Google Sheets now, you can literally turn it into an app, which is really cool, right?
00:18:56.240 --> 00:19:03.359
So, this is you know, at least for me and for a lot of people, all this technically does is cut down on you know two or three little steps in between.
00:19:03.519 --> 00:19:16.079
Before I would just take my Google Sheet, you know, download it, export it, uh, or connect it inside, you know, uh Codex, Claude Code, or even Google Gemini, the chatbot side, and then go ahead and you know build a little app or dashboard.
00:19:16.160 --> 00:19:19.519
But now you can do it all inside Google Sheets without leaving.
00:19:19.680 --> 00:19:26.799
Uh so yeah, it can just build a working mini app on top of your data that stays in sync and it updates as that data changes.
00:19:26.960 --> 00:19:28.400
So who has access to this?
00:19:28.480 --> 00:19:32.240
Uh, any paid subscribers for Google Gemini, you have access to it.
00:19:32.319 --> 00:19:34.960
It is literally just inside of the right panel.
00:19:35.039 --> 00:19:41.200
So just go click that uh that Gemini uh button inside Google Sheets in plain language.
00:19:41.359 --> 00:19:44.960
Say turn this into an app that does A, B, and C and it's done for you.
00:19:45.119 --> 00:19:51.200
Uh, also, now if you have a uh Gemini work or school account, it's also available.
00:19:51.440 --> 00:19:52.960
All right, so why is this useful?
00:19:53.119 --> 00:19:57.440
Well, building a decent dashboard and sheets used to mean uh a lot, right?
00:19:57.599 --> 00:20:08.480
So uh if you're just looking at building, you know, graphs and charts, yeah, you have to kind of uh know the formulas or some formatting hacks uh or use a separate business intelligence tool.
00:20:08.559 --> 00:20:12.240
Now you just type build me a sales pipeline board and it builds one.
00:20:12.400 --> 00:20:12.640
All right.
00:20:12.720 --> 00:20:17.279
So the spreadsheet though stay uh stays as your source source of truth.
00:20:17.440 --> 00:20:20.480
And here's the kind of the unlock, at least for me personally, right?
00:20:20.640 --> 00:20:29.039
The thing, most people, if you work a lot in Google Sheets, you know this, but Google Sheets is notoriously good at connecting with all of your different data sources.
00:20:29.200 --> 00:20:37.920
So, you know, I don't know if you have you know a MailChimp list or if you have a uh, you know, ads going in, meta, whatever it is, right?
00:20:38.000 --> 00:20:41.039
So many of these platforms sync directly with Google Sheets.
00:20:41.119 --> 00:20:52.559
So when you can say if you set something up once and then you can build the app inside of Google Sheets, and then if the data is syncing live, you can build a nice little app once and never have to update it again.
00:20:52.640 --> 00:20:54.799
And you can go in and it's updated every single day.
00:20:54.880 --> 00:21:05.359
So this is kind of similar to a, I would say, a light version of like ChatGPT sites or the uh live artifacts feature inside of Claude uh code.
00:21:05.599 --> 00:21:18.799
All right, speaking of Claude, all right, they took a page out of the codex book and essentially the very good uh chat GPT uh Chrome extension that we covered here a couple of weeks ago.
00:21:18.960 --> 00:21:26.640
Now the exact same thing is available, but for Claude, essentially bringing the power of Claude Cowork into the Chrome side panel.
00:21:26.799 --> 00:21:31.039
So here's uh let's actually start uh with what Anthropic says.
00:21:31.200 --> 00:21:36.079
It says the Claude and Chrome side panel is now a Claude Cowork session.
00:21:36.240 --> 00:21:46.480
Conversations are saved to your history, your skills and connectors work in the browser, and a task you start in a tab can be finished on the Claude desktop, web, and mobile apps.
00:21:46.640 --> 00:21:51.920
It's available on Macs in Team Plans Today and is rolling out to pro users over the coming weeks.
00:21:52.079 --> 00:21:55.599
All right, so here's uh kind of the the gist.
00:21:55.759 --> 00:21:59.759
So essentially, a task that you start in a Chrome tab can now be picked up and finished.
00:22:00.160 --> 00:22:03.920
Finished in the Claude Desktop if you're using Claude Cower.
00:22:04.160 --> 00:22:08.880
So it is a co-work to co-work sync, uh, as far as my tests show.
00:22:09.039 --> 00:22:09.920
Uh, yeah, right.
00:22:10.000 --> 00:22:17.680
Still, one of the big downsides of Claude Code, Claude Desktop in general is Claude CoWork has no idea what Claude Code is doing.
00:22:17.920 --> 00:22:22.160
Claude Code has no idea what Claude Chat on the desktop is doing, and vice versa.
00:22:22.319 --> 00:22:25.759
So there are three very different silos, uh, which is uh unfortunate.
00:22:25.920 --> 00:22:33.119
So the uh the Claude uh Chrome extension here that works with Claude Cowork is just with Claude Cower.
00:22:33.440 --> 00:22:44.079
So the exact same functionality has been out for about a month now uh with the chat GPT sidebar, and then that one is unified and it works whether you're using the chat panel or the work panel.
00:22:44.240 --> 00:22:47.599
Uh, but regardless, a big step up for Claude users.
00:22:47.759 --> 00:22:48.960
Here's why it's useful.
00:22:49.119 --> 00:22:56.160
Well, Claude can now see the page that you're on and act on it, and then obviously sync that session co-work to co-work.
00:22:56.319 --> 00:23:00.799
So everything from clicking, typing, filling forms, and using your existing logins.
00:23:01.119 --> 00:23:02.480
That's the big unlock there.
00:23:02.559 --> 00:23:18.400
So that means it can work in the tools that don't already connect to AI via direct connectors or uh MCPs like your vendor portals or just you know internal dashboards that are you know a little more antiquated and don't have you know all that uh you know fun updated AI connections.
00:23:18.559 --> 00:23:27.920
So now, you know, uh uh Claude co work, just go around, click around, and use uh computer use inside your browser via the updated Chrome extension.
00:23:28.000 --> 00:23:29.599
So who's gonna find this valuable?
00:23:29.759 --> 00:23:36.559
I mean, anyone if you're a heavy claw user um and you still have to use the internet a lot, right?
00:23:36.640 --> 00:23:37.599
That's the big thing.
00:23:37.759 --> 00:23:43.440
I always say first, always check uh directly integrated connectors or integrations.
00:23:43.599 --> 00:23:54.240
Second, uh see if whatever app that you're using has an MCP or model context protocol that's supported in Claude, Gemini, ChatGPT, uh Microsoft Copilot, etc.
00:23:54.480 --> 00:24:15.599
And if it doesn't have one of those two things, that's where something like this new uh claude in Chrome uh Claud Cowork extension really comes in handy because for all the other, you know, probably 90% of the internet that can't talk directly, you know, to your Claude cowork, this new extension will at least allow it to enter uh to understand and use the interface.
00:24:15.759 --> 00:24:28.480
So yeah, I think anyone is gonna like this if you've had those you know annoying tools or you know, that you're like, oh man, you know, this this app that I use that's a big part of my job, there's no AI integration.
00:24:28.640 --> 00:24:33.119
Well, now uh at least Claude Cowork uh in Chrome can use it.
00:24:33.359 --> 00:24:38.160
All right, we have two more, and both of these are from OpenAI.
00:24:38.240 --> 00:24:41.119
So this one has been talked about for a while.
00:24:41.200 --> 00:24:43.519
It was supposed to be released in July.
00:24:43.599 --> 00:24:47.039
So an extra two weeks uh of waiting.
00:24:47.200 --> 00:24:57.920
And if you are a heavy API user, if your company is using the new GPT-5-6 soul via the API, this one could be a legit game changer.
00:24:58.079 --> 00:25:09.359
This is the first of its kind across any model, uh, literally in the history of AI, that you can run a frontier AI model at legit blazing speeds.
00:25:09.519 --> 00:25:11.200
All right, so let's talk about what's new.
00:25:11.359 --> 00:25:23.200
So OpenAI announced ultra fast uh its new speed tier that runs its most capable model, GBD56 Sol, at up to 14 times faster than normal.
00:25:23.359 --> 00:25:25.039
So, how the heck is this possible?
00:25:25.279 --> 00:25:43.599
Well, it is a specific uh partnership with the now public company, the chipmaker Cerebrus, and generates up to 750 tokens per second, with OpenAI and Cerebrus saying it's the exact same intelligence, just legit lightning speed faster.
00:25:43.759 --> 00:25:45.920
So, who has access right now?
00:25:46.160 --> 00:25:53.839
So, if you are in this is uh this one is a little more select, but I think for open AI API customers, it's a lot of our listeners.
00:25:53.920 --> 00:25:55.279
That's why I still included this one.
00:25:55.440 --> 00:26:02.240
So it is in limited preview, but this is y'all, especially for our larger enterprises.
00:26:02.400 --> 00:26:09.200
This has been one of the most looking, uh, the most highly anticipated features that I've seen in a while.
00:26:09.359 --> 00:26:12.079
Uh, right, people have been talking about this for many months.
00:26:12.240 --> 00:26:18.799
So essentially, uh Open AI did have a Cerebrus model before, but it was an older model.
00:26:18.880 --> 00:26:24.880
It was GPT 5.3 Codex, so it was a codex-specific model, and it was an older model.
00:26:25.119 --> 00:26:28.240
I literally just used this like yesterday still.
00:26:28.400 --> 00:26:33.759
Uh, so it's still a really good model, but it's older, uh, but it's so blazing fast, right?
00:26:33.839 --> 00:26:38.960
I when it first came out, I was just using it for computer use because it could use a computer, right?
00:26:39.039 --> 00:26:46.640
Like as an example, not to knock on you know Claude, but you know, Claude's computer use, as powerful as it is, it is painfully slow, right?
00:26:46.799 --> 00:26:51.599
Codex's default computer use is in my testing five to seven times faster.
00:26:51.759 --> 00:27:02.960
So when you uh use a Cerebrus model with it, it is like as fast as a human or faster than a human can use a computer, which for me is an exciting unlock.
00:27:03.119 --> 00:27:04.160
So, why is it useful?
00:27:04.240 --> 00:27:10.240
Well, until now, if you wanted real-time speed, you had to settle for a smaller, not as capable model.
00:27:10.880 --> 00:27:17.759
So this kind of removes that trade-off, it's top-tier intelligence at speed that are fast enough for live products.
00:27:17.839 --> 00:27:23.039
So, you know, as an example, we talked earlier about the new uh Gemini 3.7 flash.
00:27:23.119 --> 00:27:27.200
So this is at speed, uh, I think technically a little faster.
00:27:27.359 --> 00:27:28.079
We'll have to wait.
00:27:28.160 --> 00:27:30.880
Uh, this has only been out for you know less than 24 hours now.
00:27:30.960 --> 00:27:35.839
So we'll have to see the third-party uh speeds, but presumably this is going to be faster.
00:27:36.000 --> 00:27:43.599
So this is going to be delivering a level of intelligence via the API that literally the world has never seen.
00:27:43.759 --> 00:27:49.839
So I haven't even quite wrapped my brain yet around what is going to be made possible uh with this.
00:27:50.079 --> 00:27:52.880
But uh, you know, speed just matters for business.
00:27:52.960 --> 00:28:03.519
So, like, as an example, if you are using this to power your agents, you know, uh a 40-step task as an example, it's gonna pay that speed tax on every single step.
00:28:03.680 --> 00:28:08.880
So, faster steps turn a workflow that maybe took minutes into one that takes seconds.
00:28:09.039 --> 00:28:16.720
So you can see results faster, you can make improvements faster, and you can find a good internal use case fit faster.
00:28:16.880 --> 00:28:18.960
Whereas before it just might take longer.
00:28:19.200 --> 00:28:20.640
So, who's gonna find this valuable?
00:28:20.799 --> 00:28:31.119
Any large enterprise company that's building customer-facing AI products where people are sitting there waiting for an answer or sitting there waiting for an agent to complete its tasks.
00:28:31.279 --> 00:28:40.079
So, uh teams, uh, you know, teams that are running uh long multi-step AI agents are gonna find this uh extremely valuable as well.
00:28:40.880 --> 00:28:45.519
All right, uh, let's go to uh our last one here.
00:28:45.759 --> 00:28:53.200
Uh, this one, another kind of niche one, but is super, super exciting for me.
00:28:53.440 --> 00:28:53.839
All right.
00:28:53.920 --> 00:29:01.839
So if you remember, like a year and a half ago, Microsoft had its recall product that was kind of recalled, uh, right?
00:29:02.000 --> 00:29:03.519
Not really, but technically, right?
00:29:03.599 --> 00:29:04.319
They released it.
00:29:04.559 --> 00:29:09.039
Privacy advocates are like, you can't release this, and then it got pulled and back and forth, right?
00:29:09.119 --> 00:29:18.559
Uh now I think you probably have a much better version of this in the new update that OpenAI just dropped called Computer History.
00:29:18.720 --> 00:29:21.119
So we had talked about Chronicle before.
00:29:21.279 --> 00:29:24.240
So this is just an extension or Chronicle made better.
00:29:24.319 --> 00:29:25.680
Let me break it down in simple terms.
00:29:25.839 --> 00:29:28.240
So OpenAI just released computer history.
00:29:28.559 --> 00:29:33.440
This is a new feature that watches your activity across apps and websites on your Mac.
00:29:33.680 --> 00:29:34.799
It is Mac only right now.
00:29:34.960 --> 00:29:36.240
Sorry, Windows and Linux.
00:29:36.480 --> 00:29:42.559
Uh, and yes, uh, OpenAI did this week uh release chat GPT workslash codex for Linux as well.
00:29:42.720 --> 00:29:53.039
But this um essentially watches all your activity across apps and websites on your Mac and turns it into memories in a timeline that ChatGPT and Codecs can use.
00:29:53.279 --> 00:30:05.119
So you can literally scroll through every single thing that you've done on the new Chat GPT slash codex uh desktop app, and you can ask things like, what was I working on before my break?
00:30:05.200 --> 00:30:10.799
Or where's that proposal doc from this morning, or summarize yesterday for stand up, right?
00:30:10.960 --> 00:30:18.880
Anything that you work on inside of Codex or ChatGPT work, which for me is quite literally everything, it can notice repeated workflows, right?
00:30:19.039 --> 00:30:24.640
So it can find things that you started and didn't stop, and then it can finish it itself.
00:30:24.799 --> 00:30:26.079
That is wild.
00:30:26.240 --> 00:30:34.079
So it can not only finish work that you were doing and just forgot, or you know, maybe you ran into a roadblock, got busy, whatever.
00:30:34.240 --> 00:30:51.119
So this new computer history can not only finish those tasks that you started and just stopped or abandoned, but it can actually suggest skills as you go along, which for me, I talked about this uh on the show the other day in my uh the next 12 months of AI.
00:30:51.200 --> 00:30:52.720
Uh so make sure you go check that out.
00:30:52.799 --> 00:30:54.319
That was uh yesterday's show.
00:30:54.400 --> 00:30:58.960
So Thursday's show, uh, went over the 19 predictions every business leader needs to hear.
00:30:59.119 --> 00:31:03.279
And one of the things I said was, you know, lines of code is not an important metric.
00:31:03.359 --> 00:31:05.680
Uh tokens burned is not an important metric.
00:31:05.839 --> 00:31:12.960
One of the most important metrics, I think, for the rest of 2026 and early 2027 is how many times you are reusing skills, right?
00:31:13.039 --> 00:31:28.240
Because you are saving not only a ton of time and sharing those across your team, uh, but also you're saving tokens technically, because you are only running the most optimized version of a workflow inside of your AI operating system that you know is going to bring a certain level of output or results.
00:31:28.559 --> 00:31:43.920
So now, with this new computer history enabled, uh, it is going to complete things for you, which my gosh, I need because normally I have like, you know, eight to 20 different uh, you know, codecs threads going at once, and I forget these things a lot.
00:31:44.000 --> 00:31:46.240
Uh, but then also to suggest skills.
00:31:46.400 --> 00:31:47.519
This is really cool.
00:31:47.680 --> 00:31:59.279
So it essentially allows ChatGPT to observe and learn your activity on your computer, helping it understand your workflow, complete ongoing tasks that you might have forgotten, and then recommend those skills and automations tailored to how you use your device.
00:31:59.440 --> 00:32:00.720
So, who has this right now?
00:32:00.960 --> 00:32:07.279
Any paid um business plan has it right now, as well as ChatGPT Pro.
00:32:07.440 --> 00:32:10.960
Uh, so you have to go into the Chat GPT desktop app.
00:32:11.119 --> 00:32:14.319
It is Mac only and it is off by default.
00:32:14.400 --> 00:32:15.440
So you have to turn it on.
00:32:15.519 --> 00:32:19.599
And if you are using this across your team, every person has to turn it on themselves.
00:32:19.759 --> 00:32:24.079
And obviously, this is not yet available in the EU, UK, or Switzerland.
00:32:24.160 --> 00:32:29.039
So, all of those, you know, uh strict privacy first companies, you're not gonna have it.
00:32:29.119 --> 00:32:30.319
So, who's gonna find this useful?
00:32:30.480 --> 00:32:36.960
Well, this is the AI that finally knows what you did yesterday, not just what you typed into a chat box, right?
00:32:37.119 --> 00:32:44.160
So the context switching and like the where was I moments, I think are some of the biggest silent time wasters right now in knowledge work.
00:32:44.240 --> 00:32:58.880
Uh so a real example, you know, at the end of the day, asking for a summary of everything you worked on, or start the morning by asking ChatGPT to pick up where you stopped, or you know, looking at the skills that it suggested based on all the work that you did yesterday.
00:32:59.119 --> 00:33:07.680
So if you're just juggling a ton of different projects like me, if you're losing time, if you're a heavy AI user, you're really gonna like that.
00:33:07.759 --> 00:33:16.960
Uh, it is a little uh you know, so this computer history builds on the Chronicle research preview uh with reduced token usage and more privacy control.
00:33:17.039 --> 00:33:18.720
So, like I said, this is fresh.
00:33:18.880 --> 00:33:27.759
I haven't even used it yet, uh, because it literally I had so many uh codex threads running, I didn't want to, you know, download the update uh just yet.
00:33:27.920 --> 00:33:33.039
Uh so you know, Chronicle did take a little bit of space on your hard drive.
00:33:33.200 --> 00:33:37.200
So, but this specifically the computer history is the next iteration of that.
00:33:37.279 --> 00:33:41.440
And they did say it is a little faster in reduced token usage.
00:33:41.599 --> 00:33:48.160
All right, so that is a wrap for the seven AI updates that you should use today, y'all.
00:33:48.319 --> 00:33:53.599
If you're not paying attention every single day, that's why this Friday show is extremely useful.
00:33:53.839 --> 00:33:56.799
All right, and a reminder, y'all.
00:33:57.039 --> 00:34:00.640
The live stream and the newsletter are taking a little break.
00:34:00.799 --> 00:34:03.440
Quick announcement it is time to go back to school.
00:34:03.680 --> 00:34:07.839
So uh we're gonna be doing some learning together on everyday AI.
00:34:08.000 --> 00:34:14.880
So our start here series that we kicked off this year has been some of our most popular shows ever.
00:34:14.960 --> 00:34:19.760
Uh and we've gotten a lot of comments of people like, hey, uh, you should just run these all.
00:34:19.920 --> 00:34:21.119
So that's what we're gonna do.
00:34:21.199 --> 00:34:26.079
We're gonna be running back the entire series, 30 episodes, 30 days.
00:34:26.239 --> 00:34:26.880
Commit to this.
00:34:26.960 --> 00:34:32.559
So we're gonna be running that series back starting on Monday, starting with the very first episode.
00:34:32.639 --> 00:34:36.000
So, yeah, a little break from our normal Monday to Friday schedule.
00:34:36.079 --> 00:34:45.519
But I guarantee if you stick with this series from volume one through volume 30 and put the work in, you'll know more about AI than like 99% of the people at your company.
00:34:45.679 --> 00:34:49.679
So it's made for beginners and advanced users alike.
00:34:49.840 --> 00:34:53.519
No jargon, no fluff, no computer science degree required.
00:34:53.760 --> 00:34:55.360
Class starts Monday.
00:34:55.519 --> 00:34:56.400
Don't be late.
00:34:56.480 --> 00:34:57.920
All right, that's a wrap.
00:34:58.079 --> 00:35:01.679
Make sure if you haven't already, go to your everydayai.com.
00:35:01.840 --> 00:35:02.639
Thanks for tuning in.
00:35:02.719 --> 00:35:05.119
We'll see you back next time for more everyday AI.
00:35:05.199 --> 00:35:06.239
Thanks, y'all.
00:00:02.560 --> 00:00:10.240
AI just got faster, more accessible, and cheaper all in the same week, but there's a good chance you might have missed all of it.
00:00:10.480 --> 00:00:13.359
I mean, let's see what happened just recently.
00:00:13.599 --> 00:00:22.000
Google dropped its new Gemini 3.7 flash at half the price of the version that they just released three weeks ago.
00:00:22.160 --> 00:00:27.120
Then Elon Musk's team shipped a smarter Grok at the exact same price.
00:00:27.280 --> 00:00:34.880
And OpenAI answered with a new mode that runs its most powerful model 14 times faster.
00:00:35.119 --> 00:00:38.079
I mean, I still don't even know how that's even possible.
00:00:38.240 --> 00:00:50.000
But here's why you should care, even if you've never touched an API, because price and speed are the two things standing between AI demos and AI that actually is running inside of your business.
00:00:50.159 --> 00:00:56.960
Because every time one of those two things drops, a use case that didn't make sense last month suddenly makes a ton of sense.
00:00:57.119 --> 00:00:58.560
And that was just the model news.
00:00:58.719 --> 00:01:10.079
I mean, Claude started working inside your browser tabs, your spreadsheets can now turn themselves into live dashboards, and ChatGPT quietly shipped a feature that watches everything you do on your Mac.
00:01:10.319 --> 00:01:11.519
And I'm here for it.
00:01:11.680 --> 00:01:24.319
So that's why on Fridays we bring you Friday features, our weekly update of the latest AI features that you can actually use today, not announcements, teasers.
00:01:24.480 --> 00:01:27.519
Yeah, you can go use all these things right now.
00:01:27.680 --> 00:01:29.280
All right, let's get into it.
00:01:29.439 --> 00:01:31.200
Welcome to Everyday AI.
00:01:31.359 --> 00:01:34.159
If you're new here, my name's Jordan Wilson, and we do this every day.
00:01:34.239 --> 00:01:45.120
It's a daily, unedited, unscripted live stream, podcast, and free dealing newsletter helping business leaders like you and me not just keep up with everything that's happening, but I tell you what matters, what doesn't, how to use it.
00:01:45.200 --> 00:01:47.680
And you take that information to grow your company and your career.
00:01:47.840 --> 00:01:50.640
All of a sudden it's like, whoa, you're the smartest person in AI.
00:01:50.719 --> 00:01:52.560
Everyone's looking at you like, what's your secret?
00:01:52.719 --> 00:01:57.519
Well, your secret is this, but also our website, your everydayai.com.
00:01:57.680 --> 00:02:00.799
So if you didn't know, uh you can go check that out.
00:02:00.959 --> 00:02:07.120
We have literally 830 plus episodes you can go watch, listen to, and read about all for free.
00:02:07.200 --> 00:02:10.479
It is a free generative AI university on our website.
00:02:10.639 --> 00:02:13.840
So uh let's get into it on today's show.
00:02:14.000 --> 00:02:14.879
Here's what you're gonna learn.
00:02:14.960 --> 00:02:23.039
You're gonna know the new Mac feature that watches everything you do at work, then tells you what you did all day and maybe finishes the work that you forgot to do.
00:02:23.280 --> 00:02:29.360
You're gonna know why the smartest cheap model in AI just got 50% cheaper.
00:02:29.599 --> 00:02:32.159
And the catch that's hiding in the fine prints.
00:02:32.400 --> 00:02:38.560
And you're gonna know how Claude can now do real work inside your browser tabs, login and all.
00:02:38.800 --> 00:02:40.879
All right, so let's take a look.
00:02:41.039 --> 00:02:43.840
Actually, live stream audience, haven't shouted you out in a bit.
00:02:44.000 --> 00:02:45.520
Big bogey face, good to see you.
00:02:45.680 --> 00:02:54.159
Brent joining us from London, Brian from Minnesota, Jose joining us from Santiago, uh, Miko joining us from Tokyo.
00:02:54.319 --> 00:02:56.719
We're we're all over the place this morning.
00:02:56.879 --> 00:02:57.759
Uh, love to see it.
00:02:57.840 --> 00:02:59.840
Michelle, good to good to see you join us.
00:03:00.000 --> 00:03:02.560
All right, a lot of new AI features, y'all.
00:03:02.639 --> 00:03:09.599
So on Fridays, it's one of my favorite shows, and it's actually been one of our uh more listened to shows because yeah, it's hard to keep up.
00:03:09.680 --> 00:03:14.800
And some of these things literally, I I had the show planned last night, woke up this morning.
00:03:14.879 --> 00:03:16.479
I'm like, all right, there's already new things.
00:03:16.560 --> 00:03:18.719
You know, some things had to get the cut, right?
00:03:18.800 --> 00:03:23.280
It's kind of crazy because there's some big updates this week this week that didn't even make our top seven.
00:03:23.360 --> 00:03:27.280
I mean, uh what let me just read these before we get into the ones we're actually going to cover.
00:03:27.439 --> 00:03:32.639
I mean, the Microsoft Copilot started unifying its co-pilot offerings into the super app.
00:03:32.719 --> 00:03:35.360
So that's actually happening, wasn't even a top seven thing.
00:03:35.439 --> 00:03:41.039
Uh, ChatGPT upgraded its voice mode to include projects and file uploads, which is pretty big.
00:03:41.120 --> 00:03:42.879
I've been using that one a lot since it came out.
00:03:43.039 --> 00:03:51.599
Uh, Google Gemini released 13 new connectors, and Google also released a slew of new AI features for Google Ads and Google Analytics.
00:03:51.759 --> 00:03:55.599
So let's get into the ones that we're actually going to be covering today.
00:03:55.759 --> 00:04:05.840
And let's start with well, if you care specifically about the fastest, cheapest model that is still the most intelligent.
00:04:06.080 --> 00:04:11.599
Technically, if you look at things like artificial analysis, they actually have a nice little uh model filter.
00:04:11.759 --> 00:04:18.079
And if you say those are the three things that are most important for you, the new Gemini 3.7 flash is the recommended model.
00:04:18.240 --> 00:04:19.839
So it is brand new.
00:04:19.920 --> 00:04:25.040
And if you're wondering, like, wait, Jordan, didn't you just cover this new Gemini Flash like very recently?
00:04:25.120 --> 00:04:26.560
Yeah, three weeks ago.
00:04:26.639 --> 00:04:32.959
That's how long it took Google to go from Gemini 3.6 Flash to Gemini 3.7 Flash.
00:04:33.040 --> 00:04:34.480
So uh let's see what's new.
00:04:34.639 --> 00:04:44.000
Google released Gemini 3.7 Flash, its new AI model that it calls the most intelligent workhorse model uh for coding and agents.
00:04:44.079 --> 00:04:51.279
And it's a big step up from the version released just three weeks ago with major gains in coding, debugging, and building web apps.
00:04:51.439 --> 00:04:52.959
So who has access to this?
00:04:53.120 --> 00:04:56.560
If you're a paid subscriber, it's available already.
00:04:56.720 --> 00:05:01.519
So uh interestingly enough, it's not yet available inside of Google Gemini, right?
00:05:01.600 --> 00:05:11.759
So Gemini.google.com, but it is available in its Gemini Spark model, which is available, uh, the uh kind of agentic offering for Google, which is available in Gemini.
00:05:12.000 --> 00:05:13.199
I know that's a little confusing.
00:05:13.360 --> 00:05:21.839
Uh, but if you go into Google's AI studio, it's in there, as well as Android Studio and the Gemini API and Gemini Enterprise platform.
00:05:21.920 --> 00:05:31.680
So yeah, if you're a normal paid Gemini user, which is I think the majority of our listeners, and you go in there, you're not gonna find it there, but it is powering the Gemini Spark.
00:05:31.759 --> 00:05:38.560
And if you're using Google Anti-Gravity, so many different ways that you can access Google's uh uh AI.
00:05:38.639 --> 00:05:40.720
It's like you need a roadmap sometimes.
00:05:40.879 --> 00:05:50.879
All right, but here's the key thing API pricing is 75 cents per million input tokens and$3.75 per million output tokens.
00:05:51.199 --> 00:05:56.560
All right, so let's read just quick uh kind of what Google said about their newest release.
00:05:56.720 --> 00:05:58.240
Let me zoom in on my browser here.
00:05:58.399 --> 00:06:00.399
Apparently, I'm old and can barely see.
00:06:00.560 --> 00:06:11.839
All right, so Google says today we're building on the progress of our widely used Flash series by introducing Gemini 3.7 Flash, our most intelligent workhorse model yet for coding and agents.
00:06:12.000 --> 00:06:22.399
The release comes just three weeks after Gemini 3.6 Flash and is a direct result of developer feedback and algal uh algorithmic uh innovations that we look forward to bringing to future models.
00:06:22.560 --> 00:06:34.800
3.7 Flash delivers substantial improvements across software engineering, knowledge work, and web development workflows with an introductory price of half of the original uh 3.6 flash cost per million tokens.
00:06:35.040 --> 00:06:42.560
So if you recall uh when Gemini 3.6 Flash came out, I said I don't understand this model.
00:06:42.720 --> 00:06:50.560
Uh, because the whole point of the Flash series, all along from Google, has been to be extremely fast and extremely cheap.
00:06:50.720 --> 00:06:56.720
And Gemini 3.6 Flash, although it was fast-ish, it was not cheap, right?
00:06:56.800 --> 00:07:07.040
It was not its normal, you know, you would look at Gemini's Flash series and be like, all right, this is, you know, not like free, but this is a very low-cost model that you can put out at scale.
00:07:07.199 --> 00:07:12.399
So uh unfortunately, I think a lot of people were upset at Gemini 3.6 Flash because it wasn't really that.
00:07:12.639 --> 00:07:16.800
Gemini 3.7 Flash gets back to exactly that.
00:07:17.040 --> 00:07:20.800
So uh who's gonna find this useful?
00:07:20.959 --> 00:07:26.800
Well, it's useful because it got smarter, faster, and the price cut in half.
00:07:26.879 --> 00:07:33.120
Again, that's if you're paying on the uh API side, but the real thing is inside Google Gemini Spark.
00:07:33.199 --> 00:07:40.160
So if you haven't used Spark, that's kind of their uh new ish autonomous coding agent that is inside of Google Gemini.
00:07:40.319 --> 00:07:44.000
So it's kind of a mode, it's an agentic mode inside of Google Gemini.
00:07:44.240 --> 00:07:46.879
So for that, it's gonna make, I think, a big difference.
00:07:47.040 --> 00:07:49.199
So I tested out Gemini Spark.
00:07:49.360 --> 00:07:59.519
I wasn't super impressed, uh, you know, when it first came out to paid plans, mainly just because I've been using, you know, codex and you know, claude code, claude co-work for so long.
00:07:59.600 --> 00:08:07.519
And I'm like, yeah, I'm not really seeing anything that I can't do inside of those, but now with a better model, I think some people are gonna find it much more valuable.
00:08:07.759 --> 00:08:09.600
So who's gonna find this valuable?
00:08:09.759 --> 00:08:19.519
Well, if you're a business using AI on high via the API, uh, this is might be your new workhorse model, right?
00:08:19.600 --> 00:08:23.680
Uh, people don't really understand from a serving tokens perspective.
00:08:23.839 --> 00:08:28.240
Uh, Google is and has always been a leader in the space, right?
00:08:28.319 --> 00:08:37.919
Even though, even if you don't see a lot of people talking about the Gemini chatbot, when it comes to serving up tokens, uh, right, Google has been a leader and will continue to be.
00:08:38.159 --> 00:08:38.720
All right.
00:08:39.360 --> 00:08:43.039
Uh, so oh, Steven Leeds says uh he's been waiting for this one.
00:08:43.120 --> 00:08:43.440
All right.
00:08:43.519 --> 00:08:47.120
Hopefully you enjoy uh Gemini 3.7 flash.
00:08:47.279 --> 00:08:49.279
All right, let's get into our next one.
00:08:49.360 --> 00:08:55.039
And this is the one that just came out like kind of like an hour, a couple hours ago.
00:08:55.200 --> 00:09:00.320
So, yeah, we may have a new King of the Open Source Hill.
00:09:00.480 --> 00:09:09.919
Yeah, we don't even have full benchmarks for it, so we don't know if it's gonna truly be the best open weights model in the world, but it might be.
00:09:10.080 --> 00:09:15.840
That's because we have a new version of ZAI's GLM5.
00:09:16.639 --> 00:09:23.600
So a new model it's calling, it's best yet for coding, and the strongest open weights coding models that it has measured.
00:09:23.759 --> 00:09:32.480
So uh an interesting wrinkle, it's the same underlying model as GLM 5.2, but every single gain just comes from better post-training.
00:09:32.559 --> 00:09:35.519
So, yeah, not like a completely new model per se.
00:09:35.600 --> 00:09:40.080
This is more of a step change that's very um, very normal.
00:09:40.240 --> 00:09:48.879
But the big surprise here in the big jump, at least according to ZAI's benchmarks, like I said, it's too fresh to have any third-party uh benchmarks yet.
00:09:49.039 --> 00:09:50.399
Big jump is cybersecurity.
00:09:50.480 --> 00:10:02.720
So ZAI says the model got better at finding and chaining software vulnerabilities than the company intended, and it has already found over a thousand critical security flaws in real software like Linux and Apple's web kits.
00:10:02.799 --> 00:10:12.799
Yeah, ones that have already uh, you know, according to ZAI, withstood, you know, the fable and the GPT-5.6 uh, you know, vulnerabilities that everyone has found.
00:10:12.960 --> 00:10:19.679
The thing that's interesting here, you know, so for our live stream audience, I have uh kind of their blog post announcement up on my screen.
00:10:19.759 --> 00:10:27.600
You can always watch the video version of our podcast on uh your everydayai.com, as well as you can go catch that link in the show notes.
00:10:27.759 --> 00:10:32.559
But the interesting thing was, yeah, uh some of these benches, right?
00:10:32.720 --> 00:10:44.639
I mean, they're they're definitely um they're definitely competing with the closed source proprietary models, specifically, you know, Fable 5 from Anthropic and GBD56 Soul.
00:10:44.879 --> 00:10:47.919
So the one that really stood out is Cyber Gym.
00:10:48.080 --> 00:10:48.399
All right.
00:10:48.480 --> 00:10:53.440
So this is kind of uh a metric on the cybersecurity side, which is very important.
00:10:53.600 --> 00:10:56.000
I talked about that uh, you know, this week.
00:10:56.159 --> 00:11:01.120
Was that this week that I had my uh yeah, my my rogue AI agent show?
00:11:01.200 --> 00:11:02.320
It's all blending together now.
00:11:02.399 --> 00:11:09.759
But you know, let's go back to when Anthropic teased uh its mythos and fable models, and then they sat on it, didn't release it for like three months.
00:11:10.000 --> 00:11:15.120
The main uh the main benchmark that it led with was Cyber Gym.
00:11:15.360 --> 00:11:15.919
All right.
00:11:16.000 --> 00:11:22.000
Uh and right now, you know, it is a pretty important uh benchmark, you know, Cyber Gym and Exploit Gym.
00:11:22.080 --> 00:11:28.559
Uh, but for Cyber Gym, this new open model is the best in the world, right?
00:11:29.120 --> 00:11:30.799
Which is crazy, right?
00:11:31.200 --> 00:11:34.879
Yeah, an open, uh, an open model uh that is the best.
00:11:35.039 --> 00:11:46.639
So it scored an 84.5 on Cyber Gym, uh ahead of the other two models uh that were kind of tied there for first place previously, which is Fable 5 and GPT 5.6 soul.
00:11:46.879 --> 00:11:51.679
So a little bit more about GLM5.3 and who has access.
00:11:51.840 --> 00:11:53.759
Well, it's available to everyone now.
00:11:53.840 --> 00:12:06.960
So if you are on ZAI's GLM coding plan, or if you have their Z code installed, which Z code is pretty much just a pixel per pixel copy of Codex, anyways.
00:12:07.039 --> 00:12:13.279
Well, if you use Codex and you're like, Oh, I want to try this new GLM5.2, you can try Z code and it's pretty much the same thing as Codex.
00:12:13.519 --> 00:12:17.039
Uh, but it's been rolled out to all uh subscribers.
00:12:17.200 --> 00:12:20.000
So the open weights are not out yet.
00:12:20.159 --> 00:12:25.039
All right, so ZAI says it will release them in about two weeks after further safety testing.
00:12:25.120 --> 00:12:28.559
Uh, so calling it open source today is a little premature.
00:12:28.639 --> 00:12:36.080
It will be open weights, and uh, we at least know that GLM uh sorry, ZAI has a positive track record of doing this.
00:12:36.159 --> 00:12:47.440
This is what happened with GLM 5.2, this is what happened with Kimi K3, with Quen uh, you know, 3.8, all the Chinese open models they've released, and then they usually will release the weights, you know, in about two weeks.
00:12:47.519 --> 00:12:53.840
So technically, the weights are not available today if you're super dork and care about those things, uh, but they will be available soon.
00:12:53.919 --> 00:12:55.120
So, why is this useful?
00:12:55.279 --> 00:12:58.159
Uh, well, it's been a budget favorite.
00:12:58.240 --> 00:13:07.679
So the GLM coding plan has been a budget favorite for AI coding, and this makes the budget option meaningfully better at complex long-running tasks.
00:13:07.840 --> 00:13:13.759
So, uh, the company said on its own tests, it's roughly 50% better at coding than the last version, which is crazy.
00:13:13.919 --> 00:13:30.080
And if you recall back a couple of months when GLM 5.2 was released, this was kind of the first open model that people started to look at and say, wait, has China kind of closed the gap on US uh frontier proprietary closed models?
00:13:30.159 --> 00:13:31.759
And it was like kind of right.
00:13:31.840 --> 00:13:39.200
Um, I think before it was always, you know, US models were tier one and Chinese open models were tier two.
00:13:39.519 --> 00:13:48.000
So the US models are still ahead, but it's now it's one A and one B with the Chinese open models.
00:13:48.080 --> 00:14:05.759
So GLM, Kimi K3, uh and Quen 3.8 are probably in that 1B tier uh with you know Anthropics, uh Opus V and Fable 5, along with um OpenAI's G uh GPT 5.6 soul are kind of still in that 1A tier by themselves.
00:14:06.000 --> 00:14:09.120
So uh pretty, pretty interesting now.
00:14:09.200 --> 00:14:15.440
Yeah, Josh here on YouTube, good comment said actually OpenAI now has new cybersecurity daybreak blue and red.
00:14:15.600 --> 00:14:16.320
That is correct.
00:14:16.399 --> 00:14:17.679
They rolled that out earlier this week.
00:14:17.759 --> 00:14:20.320
If you read our newsletter, you caught that as well.
00:14:20.480 --> 00:14:22.960
All right, more new models.
00:14:23.200 --> 00:14:24.799
Do you see a pattern here?
00:14:24.960 --> 00:14:26.320
A lot of new models this week.
00:14:26.399 --> 00:14:28.559
They're coming out faster and faster than ever.
00:14:28.720 --> 00:14:36.559
And this one, I was kind of surprised with Grock 4.6 is actually a pretty good model, right?
00:14:36.639 --> 00:14:40.799
It's probably joined that one that you know, 1B uh tier.
00:14:40.879 --> 00:14:43.360
You know, all of a sudden it is a top contender.
00:14:43.600 --> 00:14:47.039
So let's talk about what's new uh in the model.
00:14:47.200 --> 00:15:02.399
So Space XAI, yes, formerly XAI, they just jammed all the names together, but the real name is Space XAI, released Grock 4.6, its newest top-tier AI model built for coding and long-running automated tasks.
00:15:02.559 --> 00:15:18.000
So uh on a widely watched independent scorecard that is uh in a lot of the benchmarks that artificial analysis covers, it now ties open AI's GPT 5.6 soul in a lot of important benchmarks and jumped five points over the previous GROC.
00:15:18.159 --> 00:15:24.080
So, yeah, on artificial uh analysis index, it is right there neck and neck with GPT 5.6 soul.
00:15:24.399 --> 00:15:27.039
So who has access to this right now?
00:15:27.200 --> 00:15:36.480
Well, developers have it available through the XAI API, and it's also included in the coding tool cursor on all plans.
00:15:36.639 --> 00:15:43.279
So, yeah, uh cursor obviously uh space XAI uh acquired cursor.
00:15:43.519 --> 00:15:47.679
Um, so this is available if you are on a cursor plan as well.
00:15:47.840 --> 00:15:56.480
It's also now the default model in Grok build, which is you know XAI's attempt at building a codex or clawed code competitor.
00:15:56.639 --> 00:15:57.759
So, why is it useful?
00:15:57.840 --> 00:16:01.200
Well, the price didn't change from the last version of Grok.
00:16:01.600 --> 00:16:06.320
So anyone using Grok got a smarter model for the same money if you are paying per tokens.
00:16:06.399 --> 00:16:15.279
But the biggest thing is this keeps the price war alive uh because the jump from Grok 4.5 to 4.6 was huge, right?
00:16:15.440 --> 00:16:24.159
Uh, maybe some of that compute that the company has been investing in and the cursor acquisition is starting to pay off in a model that maybe now companies will start to look at.
00:16:24.320 --> 00:16:34.320
Um, I'm kind of surprised, maybe that SpaceX AI isn't just going with the cursor brand name and going with, you know, Composer 3 or something like that.
00:16:34.559 --> 00:16:37.600
I will say this, and this isn't my own personal bias.
00:16:37.679 --> 00:16:39.360
This is just literally anyone on the internet.
00:16:39.679 --> 00:16:52.320
I think because of some of the areas that um, you know, Grok has uh made a niche in, people from a business perspective, when it comes to paying, you know, tokens, they're not always looking at Grok as a competitor.
00:16:52.399 --> 00:16:56.879
They're like, oh, this is the one that's you know been doing some you know wild things on the internet.
00:16:57.279 --> 00:17:05.599
Um, so we'll see if now you know Grok 4.6 kind of teetering between that 1B and 1A tier changes that.
00:17:05.759 --> 00:17:11.359
But here is what the company says and said in its release: it said today we are releasing grok 4.6.
00:17:11.519 --> 00:17:21.119
Grock 4.6 builds on GROC 4.5 with a particularly uh with a particular focus on long-running agents and more ambitious interactive and visual work.
00:17:21.279 --> 00:17:33.839
It stays with complex tasks across many steps, whether researching a topic, analyzing information, working across a code base, or turning an idea into a polished application or work artifact.
00:17:34.000 --> 00:17:47.279
Uh, the the the other thing that's kind of related to this Grok 4.6, and we won't be covering this on today's Friday features, even though I would have liked to, but it's only available on the very expensive Grok and cursor plans.
00:17:47.359 --> 00:17:52.880
But the new Grok bot, which does look exciting, and I'm looking forward to it being released on the standard base plan.
00:17:53.119 --> 00:18:05.440
So, yeah, on the Friday show, we're usually not covering things that are only exclusively on the you know 200 or 300 plans, which is grok bot for now, but grok bot is powered by grok 4.6.
00:18:05.519 --> 00:18:12.799
So pretty big real uh pretty big week of releases from the SpaceX slash grok slash cursor teams.
00:18:13.519 --> 00:18:17.279
All right, uh, let's move on to our next one.
00:18:17.440 --> 00:18:21.599
And this one, I don't know, I'm I'm a Google Sheets dork.
00:18:21.920 --> 00:18:36.559
Um, so maybe I'm more excited about this one than the average person, but uh yeah, Canvas mode, one of my favorite modes of all time uh from Google Gemini is coming and it's here inside of Google Sheets.
00:18:36.640 --> 00:18:38.240
So let's talk about this feature.
00:18:38.319 --> 00:18:39.759
You can go use right now.
00:18:40.000 --> 00:18:47.119
Google expanded access to its Sheets Canvas, a feature that turns a plain spreadsheet into an interactive dashboard.
00:18:47.279 --> 00:18:56.079
So, yeah, any uh Google Sheet that you have inside of Google Sheets now, you can literally turn it into an app, which is really cool, right?
00:18:56.240 --> 00:19:03.359
So, this is you know, at least for me and for a lot of people, all this technically does is cut down on you know two or three little steps in between.
00:19:03.519 --> 00:19:16.079
Before I would just take my Google Sheet, you know, download it, export it, uh, or connect it inside, you know, uh Codex, Claude Code, or even Google Gemini, the chatbot side, and then go ahead and you know build a little app or dashboard.
00:19:16.160 --> 00:19:19.519
But now you can do it all inside Google Sheets without leaving.
00:19:19.680 --> 00:19:26.799
Uh so yeah, it can just build a working mini app on top of your data that stays in sync and it updates as that data changes.
00:19:26.960 --> 00:19:28.400
So who has access to this?
00:19:28.480 --> 00:19:32.240
Uh, any paid subscribers for Google Gemini, you have access to it.
00:19:32.319 --> 00:19:34.960
It is literally just inside of the right panel.
00:19:35.039 --> 00:19:41.200
So just go click that uh that Gemini uh button inside Google Sheets in plain language.
00:19:41.359 --> 00:19:44.960
Say turn this into an app that does A, B, and C and it's done for you.
00:19:45.119 --> 00:19:51.200
Uh, also, now if you have a uh Gemini work or school account, it's also available.
00:19:51.440 --> 00:19:52.960
All right, so why is this useful?
00:19:53.119 --> 00:19:57.440
Well, building a decent dashboard and sheets used to mean uh a lot, right?
00:19:57.599 --> 00:20:08.480
So uh if you're just looking at building, you know, graphs and charts, yeah, you have to kind of uh know the formulas or some formatting hacks uh or use a separate business intelligence tool.
00:20:08.559 --> 00:20:12.240
Now you just type build me a sales pipeline board and it builds one.
00:20:12.400 --> 00:20:12.640
All right.
00:20:12.720 --> 00:20:17.279
So the spreadsheet though stay uh stays as your source source of truth.
00:20:17.440 --> 00:20:20.480
And here's the kind of the unlock, at least for me personally, right?
00:20:20.640 --> 00:20:29.039
The thing, most people, if you work a lot in Google Sheets, you know this, but Google Sheets is notoriously good at connecting with all of your different data sources.
00:20:29.200 --> 00:20:37.920
So, you know, I don't know if you have you know a MailChimp list or if you have a uh, you know, ads going in, meta, whatever it is, right?
00:20:38.000 --> 00:20:41.039
So many of these platforms sync directly with Google Sheets.
00:20:41.119 --> 00:20:52.559
So when you can say if you set something up once and then you can build the app inside of Google Sheets, and then if the data is syncing live, you can build a nice little app once and never have to update it again.
00:20:52.640 --> 00:20:54.799
And you can go in and it's updated every single day.
00:20:54.880 --> 00:21:05.359
So this is kind of similar to a, I would say, a light version of like ChatGPT sites or the uh live artifacts feature inside of Claude uh code.
00:21:05.599 --> 00:21:18.799
All right, speaking of Claude, all right, they took a page out of the codex book and essentially the very good uh chat GPT uh Chrome extension that we covered here a couple of weeks ago.
00:21:18.960 --> 00:21:26.640
Now the exact same thing is available, but for Claude, essentially bringing the power of Claude Cowork into the Chrome side panel.
00:21:26.799 --> 00:21:31.039
So here's uh let's actually start uh with what Anthropic says.
00:21:31.200 --> 00:21:36.079
It says the Claude and Chrome side panel is now a Claude Cowork session.
00:21:36.240 --> 00:21:46.480
Conversations are saved to your history, your skills and connectors work in the browser, and a task you start in a tab can be finished on the Claude desktop, web, and mobile apps.
00:21:46.640 --> 00:21:51.920
It's available on Macs in Team Plans Today and is rolling out to pro users over the coming weeks.
00:21:52.079 --> 00:21:55.599
All right, so here's uh kind of the the gist.
00:21:55.759 --> 00:21:59.759
So essentially, a task that you start in a Chrome tab can now be picked up and finished.
00:22:00.160 --> 00:22:03.920
Finished in the Claude Desktop if you're using Claude Cower.
00:22:04.160 --> 00:22:08.880
So it is a co-work to co-work sync, uh, as far as my tests show.
00:22:09.039 --> 00:22:09.920
Uh, yeah, right.
00:22:10.000 --> 00:22:17.680
Still, one of the big downsides of Claude Code, Claude Desktop in general is Claude CoWork has no idea what Claude Code is doing.
00:22:17.920 --> 00:22:22.160
Claude Code has no idea what Claude Chat on the desktop is doing, and vice versa.
00:22:22.319 --> 00:22:25.759
So there are three very different silos, uh, which is uh unfortunate.
00:22:25.920 --> 00:22:33.119
So the uh the Claude uh Chrome extension here that works with Claude Cowork is just with Claude Cower.
00:22:33.440 --> 00:22:44.079
So the exact same functionality has been out for about a month now uh with the chat GPT sidebar, and then that one is unified and it works whether you're using the chat panel or the work panel.
00:22:44.240 --> 00:22:47.599
Uh, but regardless, a big step up for Claude users.
00:22:47.759 --> 00:22:48.960
Here's why it's useful.
00:22:49.119 --> 00:22:56.160
Well, Claude can now see the page that you're on and act on it, and then obviously sync that session co-work to co-work.
00:22:56.319 --> 00:23:00.799
So everything from clicking, typing, filling forms, and using your existing logins.
00:23:01.119 --> 00:23:02.480
That's the big unlock there.
00:23:02.559 --> 00:23:18.400
So that means it can work in the tools that don't already connect to AI via direct connectors or uh MCPs like your vendor portals or just you know internal dashboards that are you know a little more antiquated and don't have you know all that uh you know fun updated AI connections.
00:23:18.559 --> 00:23:27.920
So now, you know, uh uh Claude co work, just go around, click around, and use uh computer use inside your browser via the updated Chrome extension.
00:23:28.000 --> 00:23:29.599
So who's gonna find this valuable?
00:23:29.759 --> 00:23:36.559
I mean, anyone if you're a heavy claw user um and you still have to use the internet a lot, right?
00:23:36.640 --> 00:23:37.599
That's the big thing.
00:23:37.759 --> 00:23:43.440
I always say first, always check uh directly integrated connectors or integrations.
00:23:43.599 --> 00:23:54.240
Second, uh see if whatever app that you're using has an MCP or model context protocol that's supported in Claude, Gemini, ChatGPT, uh Microsoft Copilot, etc.
00:23:54.480 --> 00:24:15.599
And if it doesn't have one of those two things, that's where something like this new uh claude in Chrome uh Claud Cowork extension really comes in handy because for all the other, you know, probably 90% of the internet that can't talk directly, you know, to your Claude cowork, this new extension will at least allow it to enter uh to understand and use the interface.
00:24:15.759 --> 00:24:28.480
So yeah, I think anyone is gonna like this if you've had those you know annoying tools or you know, that you're like, oh man, you know, this this app that I use that's a big part of my job, there's no AI integration.
00:24:28.640 --> 00:24:33.119
Well, now uh at least Claude Cowork uh in Chrome can use it.
00:24:33.359 --> 00:24:38.160
All right, we have two more, and both of these are from OpenAI.
00:24:38.240 --> 00:24:41.119
So this one has been talked about for a while.
00:24:41.200 --> 00:24:43.519
It was supposed to be released in July.
00:24:43.599 --> 00:24:47.039
So an extra two weeks uh of waiting.
00:24:47.200 --> 00:24:57.920
And if you are a heavy API user, if your company is using the new GPT-5-6 soul via the API, this one could be a legit game changer.
00:24:58.079 --> 00:25:09.359
This is the first of its kind across any model, uh, literally in the history of AI, that you can run a frontier AI model at legit blazing speeds.
00:25:09.519 --> 00:25:11.200
All right, so let's talk about what's new.
00:25:11.359 --> 00:25:23.200
So OpenAI announced ultra fast uh its new speed tier that runs its most capable model, GBD56 Sol, at up to 14 times faster than normal.
00:25:23.359 --> 00:25:25.039
So, how the heck is this possible?
00:25:25.279 --> 00:25:43.599
Well, it is a specific uh partnership with the now public company, the chipmaker Cerebrus, and generates up to 750 tokens per second, with OpenAI and Cerebrus saying it's the exact same intelligence, just legit lightning speed faster.
00:25:43.759 --> 00:25:45.920
So, who has access right now?
00:25:46.160 --> 00:25:53.839
So, if you are in this is uh this one is a little more select, but I think for open AI API customers, it's a lot of our listeners.
00:25:53.920 --> 00:25:55.279
That's why I still included this one.
00:25:55.440 --> 00:26:02.240
So it is in limited preview, but this is y'all, especially for our larger enterprises.
00:26:02.400 --> 00:26:09.200
This has been one of the most looking, uh, the most highly anticipated features that I've seen in a while.
00:26:09.359 --> 00:26:12.079
Uh, right, people have been talking about this for many months.
00:26:12.240 --> 00:26:18.799
So essentially, uh Open AI did have a Cerebrus model before, but it was an older model.
00:26:18.880 --> 00:26:24.880
It was GPT 5.3 Codex, so it was a codex-specific model, and it was an older model.
00:26:25.119 --> 00:26:28.240
I literally just used this like yesterday still.
00:26:28.400 --> 00:26:33.759
Uh, so it's still a really good model, but it's older, uh, but it's so blazing fast, right?
00:26:33.839 --> 00:26:38.960
I when it first came out, I was just using it for computer use because it could use a computer, right?
00:26:39.039 --> 00:26:46.640
Like as an example, not to knock on you know Claude, but you know, Claude's computer use, as powerful as it is, it is painfully slow, right?
00:26:46.799 --> 00:26:51.599
Codex's default computer use is in my testing five to seven times faster.
00:26:51.759 --> 00:27:02.960
So when you uh use a Cerebrus model with it, it is like as fast as a human or faster than a human can use a computer, which for me is an exciting unlock.
00:27:03.119 --> 00:27:04.160
So, why is it useful?
00:27:04.240 --> 00:27:10.240
Well, until now, if you wanted real-time speed, you had to settle for a smaller, not as capable model.
00:27:10.880 --> 00:27:17.759
So this kind of removes that trade-off, it's top-tier intelligence at speed that are fast enough for live products.
00:27:17.839 --> 00:27:23.039
So, you know, as an example, we talked earlier about the new uh Gemini 3.7 flash.
00:27:23.119 --> 00:27:27.200
So this is at speed, uh, I think technically a little faster.
00:27:27.359 --> 00:27:28.079
We'll have to wait.
00:27:28.160 --> 00:27:30.880
Uh, this has only been out for you know less than 24 hours now.
00:27:30.960 --> 00:27:35.839
So we'll have to see the third-party uh speeds, but presumably this is going to be faster.
00:27:36.000 --> 00:27:43.599
So this is going to be delivering a level of intelligence via the API that literally the world has never seen.
00:27:43.759 --> 00:27:49.839
So I haven't even quite wrapped my brain yet around what is going to be made possible uh with this.
00:27:50.079 --> 00:27:52.880
But uh, you know, speed just matters for business.
00:27:52.960 --> 00:28:03.519
So, like, as an example, if you are using this to power your agents, you know, uh a 40-step task as an example, it's gonna pay that speed tax on every single step.
00:28:03.680 --> 00:28:08.880
So, faster steps turn a workflow that maybe took minutes into one that takes seconds.
00:28:09.039 --> 00:28:16.720
So you can see results faster, you can make improvements faster, and you can find a good internal use case fit faster.
00:28:16.880 --> 00:28:18.960
Whereas before it just might take longer.
00:28:19.200 --> 00:28:20.640
So, who's gonna find this valuable?
00:28:20.799 --> 00:28:31.119
Any large enterprise company that's building customer-facing AI products where people are sitting there waiting for an answer or sitting there waiting for an agent to complete its tasks.
00:28:31.279 --> 00:28:40.079
So, uh teams, uh, you know, teams that are running uh long multi-step AI agents are gonna find this uh extremely valuable as well.
00:28:40.880 --> 00:28:45.519
All right, uh, let's go to uh our last one here.
00:28:45.759 --> 00:28:53.200
Uh, this one, another kind of niche one, but is super, super exciting for me.
00:28:53.440 --> 00:28:53.839
All right.
00:28:53.920 --> 00:29:01.839
So if you remember, like a year and a half ago, Microsoft had its recall product that was kind of recalled, uh, right?
00:29:02.000 --> 00:29:03.519
Not really, but technically, right?
00:29:03.599 --> 00:29:04.319
They released it.
00:29:04.559 --> 00:29:09.039
Privacy advocates are like, you can't release this, and then it got pulled and back and forth, right?
00:29:09.119 --> 00:29:18.559
Uh now I think you probably have a much better version of this in the new update that OpenAI just dropped called Computer History.
00:29:18.720 --> 00:29:21.119
So we had talked about Chronicle before.
00:29:21.279 --> 00:29:24.240
So this is just an extension or Chronicle made better.
00:29:24.319 --> 00:29:25.680
Let me break it down in simple terms.
00:29:25.839 --> 00:29:28.240
So OpenAI just released computer history.
00:29:28.559 --> 00:29:33.440
This is a new feature that watches your activity across apps and websites on your Mac.
00:29:33.680 --> 00:29:34.799
It is Mac only right now.
00:29:34.960 --> 00:29:36.240
Sorry, Windows and Linux.
00:29:36.480 --> 00:29:42.559
Uh, and yes, uh, OpenAI did this week uh release chat GPT workslash codex for Linux as well.
00:29:42.720 --> 00:29:53.039
But this um essentially watches all your activity across apps and websites on your Mac and turns it into memories in a timeline that ChatGPT and Codecs can use.
00:29:53.279 --> 00:30:05.119
So you can literally scroll through every single thing that you've done on the new Chat GPT slash codex uh desktop app, and you can ask things like, what was I working on before my break?
00:30:05.200 --> 00:30:10.799
Or where's that proposal doc from this morning, or summarize yesterday for stand up, right?
00:30:10.960 --> 00:30:18.880
Anything that you work on inside of Codex or ChatGPT work, which for me is quite literally everything, it can notice repeated workflows, right?
00:30:19.039 --> 00:30:24.640
So it can find things that you started and didn't stop, and then it can finish it itself.
00:30:24.799 --> 00:30:26.079
That is wild.
00:30:26.240 --> 00:30:34.079
So it can not only finish work that you were doing and just forgot, or you know, maybe you ran into a roadblock, got busy, whatever.
00:30:34.240 --> 00:30:51.119
So this new computer history can not only finish those tasks that you started and just stopped or abandoned, but it can actually suggest skills as you go along, which for me, I talked about this uh on the show the other day in my uh the next 12 months of AI.
00:30:51.200 --> 00:30:52.720
Uh so make sure you go check that out.
00:30:52.799 --> 00:30:54.319
That was uh yesterday's show.
00:30:54.400 --> 00:30:58.960
So Thursday's show, uh, went over the 19 predictions every business leader needs to hear.
00:30:59.119 --> 00:31:03.279
And one of the things I said was, you know, lines of code is not an important metric.
00:31:03.359 --> 00:31:05.680
Uh tokens burned is not an important metric.
00:31:05.839 --> 00:31:12.960
One of the most important metrics, I think, for the rest of 2026 and early 2027 is how many times you are reusing skills, right?
00:31:13.039 --> 00:31:28.240
Because you are saving not only a ton of time and sharing those across your team, uh, but also you're saving tokens technically, because you are only running the most optimized version of a workflow inside of your AI operating system that you know is going to bring a certain level of output or results.
00:31:28.559 --> 00:31:43.920
So now, with this new computer history enabled, uh, it is going to complete things for you, which my gosh, I need because normally I have like, you know, eight to 20 different uh, you know, codecs threads going at once, and I forget these things a lot.
00:31:44.000 --> 00:31:46.240
Uh, but then also to suggest skills.
00:31:46.400 --> 00:31:47.519
This is really cool.
00:31:47.680 --> 00:31:59.279
So it essentially allows ChatGPT to observe and learn your activity on your computer, helping it understand your workflow, complete ongoing tasks that you might have forgotten, and then recommend those skills and automations tailored to how you use your device.
00:31:59.440 --> 00:32:00.720
So, who has this right now?
00:32:00.960 --> 00:32:07.279
Any paid um business plan has it right now, as well as ChatGPT Pro.
00:32:07.440 --> 00:32:10.960
Uh, so you have to go into the Chat GPT desktop app.
00:32:11.119 --> 00:32:14.319
It is Mac only and it is off by default.
00:32:14.400 --> 00:32:15.440
So you have to turn it on.
00:32:15.519 --> 00:32:19.599
And if you are using this across your team, every person has to turn it on themselves.
00:32:19.759 --> 00:32:24.079
And obviously, this is not yet available in the EU, UK, or Switzerland.
00:32:24.160 --> 00:32:29.039
So, all of those, you know, uh strict privacy first companies, you're not gonna have it.
00:32:29.119 --> 00:32:30.319
So, who's gonna find this useful?
00:32:30.480 --> 00:32:36.960
Well, this is the AI that finally knows what you did yesterday, not just what you typed into a chat box, right?
00:32:37.119 --> 00:32:44.160
So the context switching and like the where was I moments, I think are some of the biggest silent time wasters right now in knowledge work.
00:32:44.240 --> 00:32:58.880
Uh so a real example, you know, at the end of the day, asking for a summary of everything you worked on, or start the morning by asking ChatGPT to pick up where you stopped, or you know, looking at the skills that it suggested based on all the work that you did yesterday.
00:32:59.119 --> 00:33:07.680
So if you're just juggling a ton of different projects like me, if you're losing time, if you're a heavy AI user, you're really gonna like that.
00:33:07.759 --> 00:33:16.960
Uh, it is a little uh you know, so this computer history builds on the Chronicle research preview uh with reduced token usage and more privacy control.
00:33:17.039 --> 00:33:18.720
So, like I said, this is fresh.
00:33:18.880 --> 00:33:27.759
I haven't even used it yet, uh, because it literally I had so many uh codex threads running, I didn't want to, you know, download the update uh just yet.
00:33:27.920 --> 00:33:33.039
Uh so you know, Chronicle did take a little bit of space on your hard drive.
00:33:33.200 --> 00:33:37.200
So, but this specifically the computer history is the next iteration of that.
00:33:37.279 --> 00:33:41.440
And they did say it is a little faster in reduced token usage.
00:33:41.599 --> 00:33:48.160
All right, so that is a wrap for the seven AI updates that you should use today, y'all.
00:33:48.319 --> 00:33:53.599
If you're not paying attention every single day, that's why this Friday show is extremely useful.
00:33:53.839 --> 00:33:56.799
All right, and a reminder, y'all.
00:33:57.039 --> 00:34:00.640
The live stream and the newsletter are taking a little break.
00:34:00.799 --> 00:34:03.440
Quick announcement it is time to go back to school.
00:34:03.680 --> 00:34:07.839
So uh we're gonna be doing some learning together on everyday AI.
00:34:08.000 --> 00:34:14.880
So our start here series that we kicked off this year has been some of our most popular shows ever.
00:34:14.960 --> 00:34:19.760
Uh and we've gotten a lot of comments of people like, hey, uh, you should just run these all.
00:34:19.920 --> 00:34:21.119
So that's what we're gonna do.
00:34:21.199 --> 00:34:26.079
We're gonna be running back the entire series, 30 episodes, 30 days.
00:34:26.239 --> 00:34:26.880
Commit to this.
00:34:26.960 --> 00:34:32.559
So we're gonna be running that series back starting on Monday, starting with the very first episode.
00:34:32.639 --> 00:34:36.000
So, yeah, a little break from our normal Monday to Friday schedule.
00:34:36.079 --> 00:34:45.519
But I guarantee if you stick with this series from volume one through volume 30 and put the work in, you'll know more about AI than like 99% of the people at your company.
00:34:45.679 --> 00:34:49.679
So it's made for beginners and advanced users alike.
00:34:49.840 --> 00:34:53.519
No jargon, no fluff, no computer science degree required.
00:34:53.760 --> 00:34:55.360
Class starts Monday.
00:34:55.519 --> 00:34:56.400
Don't be late.
00:34:56.480 --> 00:34:57.920
All right, that's a wrap.
00:34:58.079 --> 00:35:01.679
Make sure if you haven't already, go to your everydayai.com.
00:35:01.840 --> 00:35:02.639
Thanks for tuning in.
00:35:02.719 --> 00:35:05.119
We'll see you back next time for more everyday AI.
00:35:05.199 --> 00:35:06.239
Thanks, y'all.
