WEBVTT
00:00:02.399 --> 00:00:04.799
What a wild few days in AI.
00:00:05.040 --> 00:00:10.080
I mean, we got not only one, but two new AI models from Anthropic.
00:00:10.240 --> 00:00:16.879
Although one was technically older and it's gonna be pretty much gone in like a hundred hours, but who's counting, anyways?
00:00:17.120 --> 00:00:23.280
So Fable 5 and Sonnet 5 should be on your testing list this holiday weekend.
00:00:23.440 --> 00:00:29.120
Oh, and you remember that whole open claw thing that was the most popular open source software ever?
00:00:29.359 --> 00:00:35.359
Well, they had one quiet update this week that makes it probably like five times more useful than before.
00:00:35.600 --> 00:00:40.719
And do we finally have high quality educational brain rot videos?
00:00:40.880 --> 00:00:41.920
Is that a thing?
00:00:42.159 --> 00:00:42.719
Woof.
00:00:42.960 --> 00:00:45.359
A lot happening in AI this week.
00:00:45.439 --> 00:00:48.079
That's why we have our Friday features.
00:00:48.240 --> 00:00:55.119
Well, this time coming to you on Thursday because we have the extended holiday weekend in the US.
00:00:55.280 --> 00:01:00.799
So everyday AI will be taking a break on Friday, but we'll be back on Monday.
00:01:00.960 --> 00:01:05.359
But long story short, is we have our Friday features here on Thursday.
00:01:05.519 --> 00:01:08.159
So uh on today's show, here's what you're gonna learn.
00:01:08.239 --> 00:01:13.439
You're gonna know why Anthropic's most important update this week might not actually be Fable 5.
00:01:13.599 --> 00:01:27.040
You're gonna know how your iPhone can now control unlimited AI models with multiple big releases, and you'll know the new AI video doom scrolling format that will actually help you learn things and make you smarter.
00:01:27.200 --> 00:01:30.480
Yay! Brain rot that we can all actually enjoy.
00:01:30.799 --> 00:01:32.319
All right, let's get into it.
00:01:32.400 --> 00:01:33.439
Welcome to everyday AI.
00:01:33.599 --> 00:01:36.159
If you're new here, my name's Jordan Wilson and we do this every day.
00:01:36.239 --> 00:01:46.400
It's your unedited, unscripted live stream podcast, and free daily newsletter, helping everyday business leaders like you and me not just keep up with all this AI craziness, but I tell you what matters, what doesn't.
00:01:46.480 --> 00:01:53.359
And you use that information like you're the smartest person in AI because that's what you are if you listen and read every single day.
00:01:53.439 --> 00:02:02.799
So uh make sure if you haven't already, please subscribe to the show on Apple Podcasts or on Spotify and sign up for the free daily newsletter at your everydayai.com.
00:02:02.959 --> 00:02:09.759
Each day we recap the highlights from today's episode as well as all of the other AI news and developments you need to know to stay ahead.
00:02:10.000 --> 00:02:10.319
All right.
00:02:10.400 --> 00:02:12.560
So, like I said, just FYI.
00:02:12.639 --> 00:02:15.120
Normally, this is our Friday AI features.
00:02:15.199 --> 00:02:28.560
This is where uh I tell you, here's AI you can actually use today, because uh so much of everything else that's happening in the AI world, it's you know, drama and the stock market and jobs and rumors.
00:02:28.719 --> 00:02:30.400
So on Fridays, I say nope.
00:02:30.479 --> 00:02:33.520
Here's the things you can actually go use this minute.
00:02:33.680 --> 00:02:36.719
So, yeah, normally Fridays, but Thursdays, you get it.
00:02:36.879 --> 00:02:43.599
All right, but let's get straight to it, and we'll start with Anthropic's Sonnet 5.
00:02:43.840 --> 00:02:51.120
So this is Anthropic's most agentic saw sonnet model yet, uh, which was released this week.
00:02:51.919 --> 00:02:58.560
And according to benchmarks, it performs pretty close to Opus 4.8 at a lower cost.
00:02:58.879 --> 00:03:05.680
So it's about uh on some benchmarks, it's about 63-ish percent versus you know Opus 69.
00:03:06.560 --> 00:03:13.039
Uh, so you know, pretty high uh compared to Opus in terms of benchmarks.
00:03:13.120 --> 00:03:15.520
Uh so some it's even ahead of Opus.
00:03:15.919 --> 00:03:24.879
So you could kind of think of this as like a maybe a sonnet 4.8, but it's sonnet five.
00:03:25.120 --> 00:03:51.280
Kind of confusing because now we have multiple tiers of models from Anthropic, all on different levels, because you still have Port Haiku, this fastest but least intelligent model that's on 4.5, and then you have Opus, which used to be the most powerful, but now it's second in line on 4.8, and then you have what used to be the middle model, which is now the second in Sonnet on five, and then you have Fable Five.
00:03:51.360 --> 00:03:55.199
So, in order, yeah, if you're confused, all these new naming systems, right?
00:03:55.280 --> 00:03:59.680
We have all the new stuff from OpenAI, you have Soul, Terra, Luna, right?
00:03:59.759 --> 00:04:01.199
So here we're gonna go in order.
00:04:01.280 --> 00:04:02.400
I'm gonna tell you the models.
00:04:02.479 --> 00:04:20.079
Uh, so entropic from the least powerful to the most powerful, uh, we have haiku 4.5, then we have Sonnet 5, then we have Opus 4.8, then we technically have Fable 5 slash Mythos 5, kind of the same models, one without guardrails, one with.
00:04:20.480 --> 00:04:21.759
So all right.
00:04:21.920 --> 00:04:25.680
So let's talk about what makes Sonnet 5 good.
00:04:26.000 --> 00:04:28.800
So uh 1 million context win, right?
00:04:28.879 --> 00:04:39.439
That kind of seems to be the norm now, and that includes 128,000 uh tokens of the outputs, it has adaptive thinking on by default and a new tokenizer built in.
00:04:39.600 --> 00:04:41.680
So this is available for everyone.
00:04:41.839 --> 00:04:45.759
And when I say everyone, even if you're on the free plan, which is good.
00:04:45.839 --> 00:04:53.360
Uh, so you know, a lot of times on a free plan, you're getting a less powerful model, and it's one of the most common misconceptions in AI, right?
00:04:53.439 --> 00:04:57.680
I've literally I don't see it as much in 2026.
00:04:58.079 --> 00:05:16.560
Uh, but last year I would still see like literally Fortune 500 level uh enterprise executives leading AI strategy, and they would log into a free, you know, Chat GPT or a free Gemini or a free anthropic uh account, and they would be testing things that way, which is the worst thing to do, by the way.
00:05:16.720 --> 00:05:22.800
But it's good now that Anthropic is upping their default free model to something on the five tier.
00:05:22.879 --> 00:05:24.399
It is a really good model.
00:05:24.560 --> 00:05:31.199
I've actually really enjoyed uh using Sonnet 5 more so than Opus and maybe even Fable.
00:05:31.439 --> 00:05:33.360
We'll get to that here in a little bit.
00:05:33.680 --> 00:05:35.360
So here's why it's useful.
00:05:35.519 --> 00:05:40.000
Well, you have Opus level agentic performance at Sonnet pricing.
00:05:40.160 --> 00:05:47.360
So the pricing, uh, if you are paying on the API side, uh, there's kind of this uh you know discount pricing for a couple of weeks.
00:05:47.519 --> 00:05:55.279
So it's two dollars uh per million um in, ten dollars per million out, then it goes up to three dollars and fifteen dollars.
00:05:55.600 --> 00:05:59.519
So uh if you are paying uh the bill, so to speak, right?
00:05:59.600 --> 00:06:20.879
So if you're not on a subscription, if your company is using um, you know, anthropics models on the API side, Sonnet might actually be a good like delegator model because with how freaking expensive the mythos uh tier models are with mythos 5 and fable five, you can't be using those things for everything, right?
00:06:21.040 --> 00:06:28.639
I'm even having to, you know, shift my own usage just on the subscription plan because you can't run you know fable five for everything.
00:06:28.720 --> 00:06:31.839
Yes, it's out, it's only a limited time, but you know, you can't.
00:06:32.000 --> 00:06:38.000
So you do have to make smart decisions whether you're on a subscription using anthropic or you're on the API side.
00:06:38.079 --> 00:06:40.319
And I've really liked Sonnet so far.
00:06:40.399 --> 00:06:46.800
So it's cheaper than uh, you know, Opus 4.8 with similar-ish, um, you know, similar-ish performance.
00:06:46.879 --> 00:06:54.079
So I would slot it just below Opus 4.8, uh, GPT-5.5, Gemini 3.1, Gemini 3.5, uh Flash.
00:06:54.240 --> 00:06:56.560
So it's it's kind of that tier below.
00:06:56.720 --> 00:07:03.759
Uh, but it's lower, uh, you know, anthropic says lower sycopency, uh, lower hallucination rates, all those good things.
00:07:03.920 --> 00:07:05.920
Um, so who's gonna find this valuable?
00:07:06.079 --> 00:07:13.199
Uh I mean, if you want a daily driver model you can afford to use in anthropic, it's is gonna be much better.
00:07:13.279 --> 00:07:21.279
Uh, you know, if you are on a$20 plan, a team plan, a$200 max plan that I'm on, you can't really use the big model much, right?
00:07:21.439 --> 00:07:24.959
The fable five, uh, and it's gonna be gone in a couple of days, anyways.
00:07:25.040 --> 00:07:27.680
So this is a great, I think, everyday driver.
00:07:27.839 --> 00:07:40.639
Um, and it's probably even if you are comfortable using Opus 4.8, I'm sure we'll see an Opus V in the next, you know, I don't know, three to six weeks, uh, something like that, depending on if it gets held up uh by the US government.
00:07:40.720 --> 00:07:49.519
Uh, but it might just be a good standard practice moving forward for anthropic bottles uh to get used to using Sonnet, right?
00:07:49.600 --> 00:08:05.040
For I think for the majority of knowledge workers out there, I'm not talking about software developers, uh, you know, people running extremely heavy uh agentic workloads, but for everyday usage, this Sonnet 5, I think, could become a de facto driver.
00:08:05.199 --> 00:08:13.360
I do think in the same way, how uh, you know, we are unfortunately in this new era of AI where large language models are no longer democratized, right?
00:08:13.439 --> 00:08:20.720
It's not everyone in the world is gonna have pretty much unlimited use of the best models like we've had up until you know June 2025.
00:08:20.879 --> 00:08:24.720
So, you know, we might have to also start shifting our usage accordingly.
00:08:24.879 --> 00:08:27.519
That's why I think Sonnet 5 is a great model.
00:08:27.920 --> 00:08:30.720
Um, so uh who's gonna find it valuable?
00:08:30.800 --> 00:08:39.440
So, yeah, DevonOps teams, running agents at scale, uh, because it is good for a lot of those lower, uh lower hanging fruit agentic tasks, long-running tasks.
00:08:39.519 --> 00:08:42.720
Uh, and then also, like I said, cost conscious API buyers.
00:08:42.799 --> 00:08:49.519
Uh, so if you want that more OPIS class uh performance, but without the Opus pricing, uh, there you go.
00:08:49.600 --> 00:08:55.120
Or if you didn't win the lottery and can't afford uh Fable, then you know, Sonnet is gonna be your go-to.
00:08:55.919 --> 00:08:58.000
All right, uh, let's keep it going.
00:08:58.240 --> 00:09:01.759
Next, yeah, brain rot that could actually be good for you.
00:09:01.919 --> 00:09:03.440
All right, this isn't really brain rot.
00:09:03.519 --> 00:09:08.080
Uh and and and FYI, I am not a fan of scrolling vertical videos.
00:09:08.240 --> 00:09:12.320
Uh, you know, when open AI came out with their Sora, I'm like, no, this is terrible.
00:09:12.480 --> 00:09:14.000
Please, you know, I don't want this.
00:09:14.080 --> 00:09:15.919
You know, I'm actually glad they killed it off.
00:09:16.000 --> 00:09:22.320
Uh, this one with Notebook LM, this might be Doom scrolling that I will actually start to do.
00:09:22.480 --> 00:09:26.639
That's because Notebook LM has a new video format.
00:09:26.799 --> 00:09:33.919
Uh, it is yes, vertical video 60 second, and they're really good.
00:09:34.159 --> 00:09:45.759
All right, so this is a new studio output in Notebook LM, a 60-second vertical video generated from your uploaded sources with narrated explanations and animations.
00:09:45.919 --> 00:10:00.399
So this was just announced like a couple hours ago, uh, alongside the new and powered by uh the new Nano Banana 2 Lite, which does power the visuals inside of this new 60-second vertical video format.
00:10:00.559 --> 00:10:07.200
Uh, so it is a shorter vertical sibling to Notebook LM's existing cinematic video overview.
00:10:07.279 --> 00:10:09.039
So that's the big difference here.
00:10:09.120 --> 00:10:16.960
So you've inside of Notebook LM uh with their video generations, they've always had the video generations, which are really, really good, by the way.
00:10:17.120 --> 00:10:26.399
But they upgraded um using their uh you know, VO or I guess now omni technology, uh, you know, this cinematic level of video.
00:10:26.480 --> 00:10:30.720
So think uh the normal video overviews, great.
00:10:30.879 --> 00:10:33.840
Um, you know, but there's really not that cinematic feel.
00:10:34.000 --> 00:10:38.960
So this is the cinematic feel of the videos, but just short 60 second verticals.
00:10:39.039 --> 00:10:41.919
So obviously made for learning on the go.
00:10:42.080 --> 00:10:46.960
So I am someone I am not like on any kind of vertical video social media.
00:10:47.120 --> 00:10:49.360
I don't do it, I I refuse to.
00:10:49.519 --> 00:10:56.799
That's why, like when I'm looking at this, it's like I almost don't want to, but I've wanted something like this for a very long time.
00:10:56.960 --> 00:11:02.720
So I might actually uh finally be a short form vertical video uh consumer.
00:11:02.879 --> 00:11:05.200
So who has access right now?
00:11:05.360 --> 00:11:09.840
If you are a paid user, you should have access.
00:11:10.000 --> 00:11:11.759
So they did already start the rollout.
00:11:11.919 --> 00:11:16.480
I think I got access to this on some of my pit accounts, but not all of my pit accounts uh on Gemini.
00:11:16.720 --> 00:11:23.840
So if you are a Gemini paid user, it should you either have it now or it should be rolling out.
00:11:24.159 --> 00:11:32.399
Uh so it is starting with users in English, uh, and then it is rolling out to others in the coming weeks.
00:11:32.480 --> 00:11:36.000
So mobile and non-English rollout status isn't confirmed yet.
00:11:36.159 --> 00:11:38.159
Uh so why is it useful?
00:11:38.320 --> 00:11:50.159
Well, it turns all your long source material inside of Notebook LM into these really cinematic, fun, and engaging uh you know, short 60-second videos.
00:11:50.320 --> 00:11:54.879
So um, notebook LM, I cannot stop gushing about it, right?
00:11:54.960 --> 00:12:05.200
As much as I as I love Codex, as much as I, you know, use Claude Desktop and all these other programs, notebook LM is still one of my go-to's, right?
00:12:05.360 --> 00:12:11.840
One of my most used skills inside of Codex is having it make notebook LM generations for me.
00:12:11.919 --> 00:12:14.320
So I can just wake up, open the app on my phone.
00:12:14.480 --> 00:12:24.320
If I'm out touching grass, taking a walk, I have literally so many uh, you know, usually uh audio uh audio overviews, but sometimes video overviews as well.
00:12:24.559 --> 00:12:32.080
Just bespoke with all the most pinpoint specific information I need to know uh for my day to learn different things.
00:12:32.240 --> 00:12:32.879
So it's great.
00:12:32.960 --> 00:12:34.320
So who's gonna find this valuable?
00:12:34.480 --> 00:12:43.039
I think any obviously notebook online power user, if you're an educator, a student, a student of life, if you're trying to learn something new, uh this is great.
00:12:43.120 --> 00:12:46.240
But also, this is good content if for social teams.
00:12:46.320 --> 00:12:51.519
So if you're a marketer, uh, I mean, these cinematic videos are amazing.
00:12:51.679 --> 00:12:57.919
I'm actually surprised when they were first announced, they didn't like go more viral than they actually did.
00:12:58.080 --> 00:13:08.399
I think one of the reasons, uh, you know, which is a rollout strategy I've never quite understood, um, is they limited this to ultra subscribers when cinematic videos first rolled out.
00:13:08.480 --> 00:13:16.159
So this is cool that the 60 second versions, if you are on a normal paid Gemini plan, uh, you should have access to this.
00:13:16.399 --> 00:13:18.399
All right, let's keep it going.
00:13:19.120 --> 00:13:25.919
Uh, yeah, now you have a new financial assistant in your pocket.
00:13:26.159 --> 00:13:36.559
So our next Friday feature on Thursday is the ChatGPT finance app uh that has now rolled out to plus users.
00:13:36.720 --> 00:13:36.960
Yeah.
00:13:37.200 --> 00:13:40.159
Speaking of that rollout strategy, same thing here.
00:13:40.240 --> 00:13:45.919
This is technically something that came out to pro users on the$200 a month plan a couple of weeks ago.
00:13:46.080 --> 00:13:53.919
I don't think we really covered it much because I tend to not cover things if they're only on that, you know, pricey$200 tier.
00:13:54.080 --> 00:14:02.720
I try to stick to things that are just on the base paid plans because I think that's what the overwhelming majority of you know our listeners uh are on.
00:14:02.960 --> 00:14:06.240
So uh, what is the Chat GPT finance?
00:14:06.320 --> 00:14:19.279
So this is built into Chat GPT and it's OpenAI's personal finance feature that connects your bank and investment accounts via plaid, and then you can ask finance questions grounded in real account data.
00:14:19.440 --> 00:14:23.600
Uh so this was originally launched, like I said, mid-May, but it was pro only.
00:14:23.840 --> 00:14:29.679
But now it is available in live to anyone if you are on that$20 a month plan.
00:14:29.840 --> 00:14:32.000
You've got it now in the US.
00:14:32.240 --> 00:14:39.200
So uh the original Pro launch covered uh web and iOS.
00:14:39.279 --> 00:14:44.159
Uh, I'm not sure if Android is out, so yeah, make sure to check in the newsletter for that and we'll let you know.
00:14:44.320 --> 00:14:51.519
Uh, so uh it is US only, like I said, and here's why I think it's useful.
00:14:51.679 --> 00:14:59.679
Well, you can just make your financial use of Chat GPT way faster and way more personalized.
00:14:59.840 --> 00:15:03.120
A couple of things to keep in mind because I probably hear what a lot of people are saying.
00:15:03.279 --> 00:15:10.960
Oh my gosh, I would never connect my financial data to any AI assistant, any AI account, right?
00:15:11.039 --> 00:15:18.080
So it's connected through Plaid, and Plaid is one of the most trusted uh, you know, names in finance around.
00:15:18.240 --> 00:15:28.480
Um, so you it's not like you're signing up for an unknown uh you know platform or bringing in, you know, if let's just say you have a couple of retirement accounts, a couple of bank accounts, a couple credit cards.
00:15:28.720 --> 00:15:37.360
It's not like you're bringing in, you know, five, six, seven, twelve uh separate entities into Chat GBT and giving them all different permissions and you have to monitor them all.
00:15:37.519 --> 00:15:40.000
It's not like that, it just connects through plaid.
00:15:40.159 --> 00:15:42.799
So it's already coming in through a trusted name.
00:15:42.960 --> 00:15:47.519
Um, you know, according to OpenAI, there's you know data they just can't see, right?
00:15:47.600 --> 00:15:49.039
They can't see your full account numbers.
00:15:49.200 --> 00:15:51.039
There is no right access.
00:15:51.120 --> 00:15:52.159
That's the big thing, right?
00:15:52.240 --> 00:15:54.000
So it's not gonna make any changes.
00:15:54.159 --> 00:16:01.440
So it can only read the information, it doesn't know your full account access, and it is all secure through Plaid.
00:16:02.000 --> 00:16:04.960
Uh, so here is what OpenAI said about it.
00:16:05.120 --> 00:16:16.399
They said today we're releasing, and this is from uh, you know, the original release, but they said today we're releasing a preview of a new personal finance experience in Chat GPT to pro users, now out to everyone.
00:16:16.559 --> 00:16:29.759
Uh, now you can securely connect your financial accounts, see a dashboard of where your money is going, and ask chat GPT questions grounded in your financial context, all while staying in control of your data.
00:16:30.000 --> 00:16:39.679
Um, so they say money, uh money touches nearly every part of life, where we live, what decisions we make, how we care for our loved ones, what future we might imagine.
00:16:39.759 --> 00:16:48.480
But managing finances today often means piecing together accounts, apps, cards, loans, and spreadsheets just to know where things stand.
00:16:48.639 --> 00:17:07.200
Uh, then they say, with your financial apps uh accounts connected, ChatGPT can combine that reasoning with your financial, with your real financial context and what you've shared about your goals, lifestyle, and priorities, helping you spot patterns, understand trade-offs, and plan for big decisions in a way that feels more personal and complete.
00:17:07.359 --> 00:17:20.559
Uh, so yeah, whether you're trying to uh plan you know goals on your spending, uh travel spend analysis, your spending insights, scenario planning, investment risk, subscription reviews, whatever it is.
00:17:20.640 --> 00:17:27.519
Uh so anything that you do give Chad GPT access to inside of PLAD, it can pull all of that information in.
00:17:27.599 --> 00:17:35.200
So not only can you interact with all of your finances in natural language, which is a big cheat code in and of itself, right?
00:17:35.359 --> 00:17:43.519
Because a lot of these uh applications do not have, I mean, many of them do, but not all of them have reasoning AI, right?
00:17:43.599 --> 00:17:49.599
You might be able to pull a certain number and ask it, you know, what was my balance last month in this account, right?
00:17:49.839 --> 00:17:54.720
But many of them do not have AI that can reason over all of this data.
00:17:54.880 --> 00:18:04.000
So not only that, the ability to have a powerful reasoning model like GPT 5.5, but also that it can marry all of your other context, right?
00:18:04.079 --> 00:18:07.119
So your memory inside of ChatGPT and your chat history.
00:18:07.200 --> 00:18:25.759
So when you can throw those things, those three things together, a reasoning model, well, four things, uh, you know, uh a reasoning model, uh, you know, your um bank accounts, your uh, you know, all your financial uh accounts, then your chat history, memory.
00:18:25.839 --> 00:18:27.759
That's that's a huge unlock.
00:18:27.839 --> 00:18:29.200
So who's gonna find this valuable?
00:18:29.359 --> 00:18:32.640
Well, I mean, I think this is big consumer play, uh, right.
00:18:32.720 --> 00:18:36.720
If if you've ever been a user of like a mint uh.com, right?
00:18:36.799 --> 00:18:42.559
There's these popular uh you know personal finance apps, but like let me just call it out.
00:18:42.720 --> 00:18:45.119
A lot of them cost money, right?
00:18:45.200 --> 00:18:55.359
I think that open AI is going to start winning and retaining a lot of consumer users just by replacing other subscriptions that people just used to pay for.
00:18:55.519 --> 00:19:02.240
You know, as an example, I used to pay for mint, and then it got to the point I'm like, you know, why am I still paying for this?
00:19:02.400 --> 00:19:07.279
Uh, you know, and now it's like, okay, well, this is going to be included in my subscription, anyways.
00:19:07.359 --> 00:19:14.079
And it does bring in all of the other personal context that you know ChatGPT has on me, which I'm gonna love.
00:19:14.240 --> 00:19:16.000
All right, our next one.
00:19:16.400 --> 00:19:26.319
This one might sound small, but I want you to think about what it unlocks, especially if you are a power uh Twitter user like me, right?
00:19:26.400 --> 00:19:29.279
This is where I get a lot of my you know AI information.
00:19:29.440 --> 00:19:38.400
I follow the discourse of what's happening, you know, online, you know, in the AI space, mainly on Twitter, because that's where a lot of people are conversing about these things.
00:19:38.559 --> 00:19:53.279
So X, formerly known as Twitter, I still call it Twitter, uh, launched their official MCP model context protocol server, letting all different AI tools connect directly to the X API.
00:19:53.440 --> 00:19:55.279
Uh, so this is big.
00:19:55.519 --> 00:19:57.519
So this is live right now.
00:19:57.680 --> 00:20:12.240
So, you know, whether you want to uh you know go through Twitter if you have a bajillion bookmarks like I do now, uh you you do have to load money uh into the uh into the X account.
00:20:12.319 --> 00:20:17.279
So this isn't something that's free, but it looks extremely affordable and cheap.
00:20:17.440 --> 00:20:20.640
Uh, and I'm gonna tell you the uh the reason why I think that.
00:20:20.720 --> 00:20:22.799
So you do have to, you know, go through that step.
00:20:22.880 --> 00:20:29.119
So it's like, oh bummer, you know, you might spend a couple dollars a month if you're a heavy uh user.
00:20:29.279 --> 00:20:35.119
So you do have to go through, sign up for the X developer portal, uh, you know, go through all that hoops and all that.
00:20:35.200 --> 00:20:53.279
But then at that point, I mean, whether you're a Claude user, cursor, you know, if you're one of the 10 people using GrocBuild, you know, Codecs, whatever, you know, anywhere that MCPs are supported, which is essentially everywhere, you can now scroll X and grab important information uh from there.
00:20:53.440 --> 00:20:59.839
So here is um what was said in a tech crunch article on this.
00:21:00.079 --> 00:21:07.759
So it says this allows developers to save time spent on integration work to focus on whatever they're actually building.
00:21:07.920 --> 00:21:11.599
So, yeah, obviously, if you're trying to build something, this is helpful as well.
00:21:11.759 --> 00:21:20.799
So it says developers have long been able to search X, read posts, look up users, analyze conversations and trends, and do more using the platform's API.
00:21:20.960 --> 00:21:27.599
The hosted MCP doesn't add new capabilities on that front, it just makes them easier to expose to AI applications.
00:21:27.839 --> 00:21:37.039
So by doing so, X can position itself as an information network filled with real-time data to retrieve and analyze rather than just a social hangout.
00:21:37.359 --> 00:21:38.960
So uh for me, this is big.
00:21:39.119 --> 00:21:40.240
So why is it useful?
00:21:40.400 --> 00:21:50.079
Well, it removes the work that developers used to have to do themselves because you used to have to build a custom MCP server, host it, wire it to the X API, all that stuff.
00:21:50.240 --> 00:21:53.440
Now you can set up an MCP server in like a minute, right?
00:21:53.519 --> 00:21:55.440
You can literally just give the URL.
00:21:55.519 --> 00:21:59.680
Uh, you know, once you go through and sign up for the developer account and authorize it, you just give the URL.
00:22:00.079 --> 00:22:03.680
to codecs or you know clawed code and it does it all for you.
00:22:04.240 --> 00:22:10.079
So if you're a develop, if if you're a developer, you know, uh you're gonna find this valuable.
00:22:10.160 --> 00:22:19.839
If you're building social listening tools, competitive monitoring, uh research tools, you know, security researchers, threat intelligence analysts, right?
00:22:20.079 --> 00:22:25.279
Essentially if you spend time for your job um on Twitter, right?
00:22:25.440 --> 00:22:30.880
Following different conversations, what's happening in your industry, how are people reacting to certain things?
00:22:31.119 --> 00:22:38.000
And you are also a user of something like you know Claude on the desktop, uh Codex, etc.
00:22:38.319 --> 00:22:39.759
Bring those two things together.
00:22:39.839 --> 00:22:43.039
Uh I do spend a lot of time on Twitter every single day.
00:22:43.119 --> 00:22:44.720
I'm bookmarking so many things.
00:22:44.880 --> 00:22:53.039
So this is one I'm actually excited for because I'll spend way less time on Twitter and I'm just gonna have Codex go through and do all those things for me.
00:22:53.359 --> 00:22:55.039
So pretty exciting one there.
00:22:56.000 --> 00:23:01.119
Speaking of things you can hand off to others, well now OpenClaw.
00:23:01.200 --> 00:23:08.799
Yes, you can talk with your open claws without having to go onto Discord or Telegram or whatever it is.
00:23:08.960 --> 00:23:13.759
So yeah now your open claw is accessible on your phone.
00:23:13.920 --> 00:23:31.839
That's because OpenClaw the uh open source personal AI agent project launched its first official native mobile app for iOS and Android uh so it is free on the Apple App Store and Google Play.
00:23:31.920 --> 00:23:51.440
Uh you obviously FYI this is just a bridge this is a remote right so it's not like you can skip the whole open claw setup which can be a little cumbersome and you it is a lot of maintenance to keep it up and working I barely use open claw anymore because in my opinion codecs can do just about everything.
00:23:51.599 --> 00:24:03.839
I even text my codecs uh so anyways uh but if you are a heavy open claw user whether it's personally or your organization uh I think this is a great new feature.
00:24:04.000 --> 00:24:42.079
Uh so uh so it is like I said it's not a standalone it's just a companion node that pairs with your posted open claw gateway wherever that is right so most people are using an actual dedicated machine or if you're using a cloud server whatever it is and pairing uh it's pretty easy it's just a QR code and then you set it up once and then the gateway is running so uh this is great because then you can monitor your open claw activities without having to sit down in front of a computer or you know I think what a lot of people were doing is they would just download like you know Discord and then hook it up through Discord or all these other you know third party messaging platforms.
00:24:42.240 --> 00:24:47.440
So now you don't have to do this anymore and you have a dedicated app that can do that for you.
00:24:47.599 --> 00:24:48.880
So who's going to find this valuable?
00:24:49.039 --> 00:25:32.640
Well like I said if you're already an open claw user or if you were on the open claw fence and one of the biggest setbacks is well you wanted a way to be able to uh you know check in on something locally because that's the reality right open claw is great and the fact that you can say oh I have a you know computer running somewhere 247 right like I do a lot of my work in my home office but I have an actual office in Chicago as well and that's where I had my open claw and if I'm being honest I'm like well if I'm only in the office once or twice a week it's like do I really want to be you know going in and you know using Discord which I don't really use anymore right since Midjourney no longer required it like a year or so ago I don't really use Discord.
00:25:32.799 --> 00:25:58.000
So it's like do I want to be using another app just to be messaging uh you know it's another communication channel that I just didn't want to take on so maybe this might you know get me to uh you know re-engage my open claw probably not because I'm plenty happy right now with codex um and with claude code but if this is something that you want to see more coverage on let me know in the comments and maybe we'll do something a little more dedicated.
00:25:58.480 --> 00:26:09.039
Uh and the thing that a lot of people are wondering is like how the heck did this make it onto the Apple App Store uh because they are extremely strict on allowing anything agentic.
00:26:09.119 --> 00:26:14.559
But the reason why is because well none of this is actually stored on your iPhone.
00:26:14.720 --> 00:26:35.839
The iPhone is just a bridge and all of the data lives on your actual machine so yeah no uh according to uh open claw and a lot of the reporting uh you know none of the data actually lives on your phone so it is literally just a bridge or remote control to see what's going on and to be able to communicate with your claws in real time.
00:26:36.079 --> 00:26:40.240
All right two more and we'll make these ones quick even though they're bigger ones.
00:26:40.480 --> 00:26:45.039
All right next we have Cursor released an iOS app.
00:26:45.119 --> 00:26:48.480
Yeah like I said the phones were getting all the love this week.
00:26:48.559 --> 00:26:57.599
So here's what's new Cursor launched its first native iPhone and iPad app in public beta and it is available now.
00:26:57.839 --> 00:27:10.880
So this lets developers launch and monitor AI coding agents from their phone uh as well as their new popular always on cloud agents or just remote control of agents already running on their new computer.
00:27:11.119 --> 00:27:19.279
So who has access well it's live now it's in public beta on the app store and it is gated right now obviously to paid cursor plan.
00:27:19.519 --> 00:27:21.839
So you have to have a paid cursor plan for this to work.
00:27:22.400 --> 00:27:23.759
So why is it useful?
00:27:23.839 --> 00:27:52.240
Well you can kick off or check agent work without keeping your laptop open uh right so no more walking around with your laptop half ajar like I used to do back in February uh with uh codex uh so now you can just have it all running from your phone uh so this uses the iOS live activities and push notifications so you're pulled back in when you want to when an agent finish hits a blocker or if it has a PR ready.
00:27:52.400 --> 00:28:00.240
It does also support voice input screenshot annotation uh as well as full diff review and PR merge from your phone.
00:28:00.400 --> 00:28:03.119
So here's who I think is going to find it useful and valuable.
00:28:03.200 --> 00:28:21.759
I think software engineers obviously teams who are already using cursor uh if you're an on-call person who needs to triage or fix issues away from your primary machine if you want to go outside and touch grass right or teams wanting a lightweight way uh to review and merge AI generated PRs without being at the desk.
00:28:21.839 --> 00:28:31.839
Uh so here is what cursor says cursor in their blog post said cursor is now available as a native iOS app in public beta so you can build from anywhere.
00:28:32.000 --> 00:29:03.359
Until now developers have worked around the limits of their local machines keeping laptops half open and caffeinated everywhere they go yeah that one's fun because there's obviously a popular program called caffeine that keeps your uh laptop awake even when it's closed that's the joke there uh with cursor for iOS you can launch always on agents in the cloud or control agents running on your computer from your phone kick them off when ideas strike get notified when work is ready for review and merge PRs on the go.
00:29:03.680 --> 00:29:27.279
Whether your agents are running on your machine or in the cloud you can move work forward from wherever you are so this really thrusts cursor I think into serious consideration in the super app race right I think we're gonna be talking about the super apps a lot more come like September uh we should be hearing from openai any day on their super app.
00:29:27.359 --> 00:31:06.160
Uh you know Microsoft you know alluded to their super app at the build conference in San Francisco a few weeks ago uh so yes cursor I think uh between the uh SpaceX XAI acquisition so the fact that they essentially have like unlimited compute now number one the fact that their last model they made composer 2.5 was actually really good and cost efficient so you throw in those two things with now uh I think they have a top two harness next to codecs so the combination of a very good desktop app and now maybe the best iOS control pane right uh chet inside of Chad GPT has a great remote control for codecs but I think the cursor might be a little bit ahead mainly because you can start um you know new threads then pick them up on your computer start new cloud agents so there are a couple of dedicated uh a couple of uh small advantages right now I would say that cursor has over codecs slash chat gbt when it comes to managing your local machine uh from your phone all right and then last but not least the technically the biggest story of I don't know the last month or so well anthropics Fable 5 is back yay right probably not all right so uh I would say the overwhelming majority of our listeners are people who are on subscriptions so don't get too excited on this because it's gonna be gone in like a hundred hours uh and then you're gonna have to pay for it out of pocket.
00:31:06.319 --> 00:32:16.799
So uh but we've covered uh Fable five plenty but let me just give you the update and well you can use it now and you absolutely should uh because it is going to be gone in about five days uh so you should be using your entire week limits uh immediately to see if this is something that you are going to pay those actual bills for uh so here is the newest update so uh it said as of today June 30th the export controls on Fable 5 and Mythos 5 have been listed uh fable five will be available starting tomorrow so this was from yesterday's uh no yeah yesterday's blog post uh so available no two days ago uh it said fable five will be available starting Wednesday july 1st to users globally on the Claude platform claude.ai claud code and claude co-work uh and then it says for pro max team and select enterprise plans fable five will be included for up to 50% of weekly usage limits through July 7th after which it will be available via usage credits.
00:32:17.039 --> 00:34:04.960
All right and yeah so uh essentially here let me just tell you what happened um right there's all this this drama that's been going on for the last two and a half weeks uh you know mytho uh anthropic has been hyping uh you know mythos and the project glasswing project since April saying it was a bioweapon you know essentially begging the federal government uh to um you know control ai they said we need to pause AI uh so they got what they wanted uh you know and now frontier models are going through a delayed process to essentially get approved from the government uh right openai had a limited rollout of their GPT 5.6 uh new models that are only available to trusted partners right now uh so that's what started with mythos uh and fable five so fable five is essentially mythos five with added guardrails uh so we got it for three days it was pulled for two weeks now it's back but for the majority of our listeners if you're on a clawed subscription plan you're only going to have it for about five days so that's a big bummer uh and there's essentially this thing called classifiers um and this is something that they put in guardrails uh to satisfy some of the US government's requests uh that this wouldn't be used in nefarious ways right um but they are extremely overbearing right anthropic even said uh in their post to their credit uh that the uh these are gonna take a bunch of uh prompts to Fable and they're gonna say nope you can't do it and they're gonna send it to Opus 4.8 uh so so far it seems like a lot of people are sour on Fable five I'll say just use it for what you can use it while it's there um and see if it's gonna be worth the price of admission.
00:34:05.119 --> 00:34:41.840
All right so for just uh comparison here let me just throw out a a number so yes I'm a heavy ish AI user uh right so inside of codex um I use about two and a half billion tokens a week so we'll say that's about 10 billion tokens a month okay and there's people who use way more right because I'm not a coder yes I'm building projects I have automations and loops and all that good stuff right but I there's plenty of people in your organization that you're listening to who go through way more than 10 billion tokens.
00:34:42.000 --> 00:36:38.960
So if I'm paying for that via Fable five right I'm paying$200 a month to do my you know roughly 10 billion tokens uh like I said give or take um on codecs GPT 5.5 high if I'm paying for that on Fable because in five days that's the only way you can use it that's roughly depending on the blend of you know input output it's about$200,000 uh a month are people gonna pay that yes is everyone gonna pay that probably not so like I said I think you're gonna have to get used to uh the new sonnet five model which I really like uh I'm sure that we'll see an Opus five model pretty soon but go play with opus you know for large organizations you know that have monopoly money uh maybe Fable 5 is something you'll be able to use in the long run but I think for most people go have fun with it you got five days see what happens all right so that is a wrap on our AI features on Friday on Thursday uh I hope this one was helpful as a quick rundown here's what it is Claude Sonnet 5 is out it's good I like it notebook LM short video overviews uh cinematic brain rot that helps you learn not actually brain rot it's really good then you have the new chat GPT finance for all plus users you have the X slash Twitter hosted MCP server uh so essentially I would have my uh agents doing computer use for that which took way too long and a lot of tokens uh so maybe my token count's gonna go down uh then we had open claw coming out with apps for iOS and Android uh cursor coming out with their iOS app and then last but not least Claude Fable 5 is back but not for long.
00:36:39.199 --> 00:36:53.760
All right if this was helpful do me a favor click that subscribe or follow button on Apple Podcast Spotify I'd appreciate that make sure you go to your everydayai.com as a reminder the show and the newsletter will be off tomorrow.
00:36:53.920 --> 00:37:04.159
Happy early 4th of July yeah we have the uh observed holiday this Friday so I'll see you back Monday thanks for tuning in we'll see you back then and every day for more everyday AI.
00:37:04.239 --> 00:37:05.199
Thanks y'all
00:00:02.399 --> 00:00:04.799
What a wild few days in AI.
00:00:05.040 --> 00:00:10.080
I mean, we got not only one, but two new AI models from Anthropic.
00:00:10.240 --> 00:00:16.879
Although one was technically older and it's gonna be pretty much gone in like a hundred hours, but who's counting, anyways?
00:00:17.120 --> 00:00:23.280
So Fable 5 and Sonnet 5 should be on your testing list this holiday weekend.
00:00:23.440 --> 00:00:29.120
Oh, and you remember that whole open claw thing that was the most popular open source software ever?
00:00:29.359 --> 00:00:35.359
Well, they had one quiet update this week that makes it probably like five times more useful than before.
00:00:35.600 --> 00:00:40.719
And do we finally have high quality educational brain rot videos?
00:00:40.880 --> 00:00:41.920
Is that a thing?
00:00:42.159 --> 00:00:42.719
Woof.
00:00:42.960 --> 00:00:45.359
A lot happening in AI this week.
00:00:45.439 --> 00:00:48.079
That's why we have our Friday features.
00:00:48.240 --> 00:00:55.119
Well, this time coming to you on Thursday because we have the extended holiday weekend in the US.
00:00:55.280 --> 00:01:00.799
So everyday AI will be taking a break on Friday, but we'll be back on Monday.
00:01:00.960 --> 00:01:05.359
But long story short, is we have our Friday features here on Thursday.
00:01:05.519 --> 00:01:08.159
So uh on today's show, here's what you're gonna learn.
00:01:08.239 --> 00:01:13.439
You're gonna know why Anthropic's most important update this week might not actually be Fable 5.
00:01:13.599 --> 00:01:27.040
You're gonna know how your iPhone can now control unlimited AI models with multiple big releases, and you'll know the new AI video doom scrolling format that will actually help you learn things and make you smarter.
00:01:27.200 --> 00:01:30.480
Yay! Brain rot that we can all actually enjoy.
00:01:30.799 --> 00:01:32.319
All right, let's get into it.
00:01:32.400 --> 00:01:33.439
Welcome to everyday AI.
00:01:33.599 --> 00:01:36.159
If you're new here, my name's Jordan Wilson and we do this every day.
00:01:36.239 --> 00:01:46.400
It's your unedited, unscripted live stream podcast, and free daily newsletter, helping everyday business leaders like you and me not just keep up with all this AI craziness, but I tell you what matters, what doesn't.
00:01:46.480 --> 00:01:53.359
And you use that information like you're the smartest person in AI because that's what you are if you listen and read every single day.
00:01:53.439 --> 00:02:02.799
So uh make sure if you haven't already, please subscribe to the show on Apple Podcasts or on Spotify and sign up for the free daily newsletter at your everydayai.com.
00:02:02.959 --> 00:02:09.759
Each day we recap the highlights from today's episode as well as all of the other AI news and developments you need to know to stay ahead.
00:02:10.000 --> 00:02:10.319
All right.
00:02:10.400 --> 00:02:12.560
So, like I said, just FYI.
00:02:12.639 --> 00:02:15.120
Normally, this is our Friday AI features.
00:02:15.199 --> 00:02:28.560
This is where uh I tell you, here's AI you can actually use today, because uh so much of everything else that's happening in the AI world, it's you know, drama and the stock market and jobs and rumors.
00:02:28.719 --> 00:02:30.400
So on Fridays, I say nope.
00:02:30.479 --> 00:02:33.520
Here's the things you can actually go use this minute.
00:02:33.680 --> 00:02:36.719
So, yeah, normally Fridays, but Thursdays, you get it.
00:02:36.879 --> 00:02:43.599
All right, but let's get straight to it, and we'll start with Anthropic's Sonnet 5.
00:02:43.840 --> 00:02:51.120
So this is Anthropic's most agentic saw sonnet model yet, uh, which was released this week.
00:02:51.919 --> 00:02:58.560
And according to benchmarks, it performs pretty close to Opus 4.8 at a lower cost.
00:02:58.879 --> 00:03:05.680
So it's about uh on some benchmarks, it's about 63-ish percent versus you know Opus 69.
00:03:06.560 --> 00:03:13.039
Uh, so you know, pretty high uh compared to Opus in terms of benchmarks.
00:03:13.120 --> 00:03:15.520
Uh so some it's even ahead of Opus.
00:03:15.919 --> 00:03:24.879
So you could kind of think of this as like a maybe a sonnet 4.8, but it's sonnet five.
00:03:25.120 --> 00:03:51.280
Kind of confusing because now we have multiple tiers of models from Anthropic, all on different levels, because you still have Port Haiku, this fastest but least intelligent model that's on 4.5, and then you have Opus, which used to be the most powerful, but now it's second in line on 4.8, and then you have what used to be the middle model, which is now the second in Sonnet on five, and then you have Fable Five.
00:03:51.360 --> 00:03:55.199
So, in order, yeah, if you're confused, all these new naming systems, right?
00:03:55.280 --> 00:03:59.680
We have all the new stuff from OpenAI, you have Soul, Terra, Luna, right?
00:03:59.759 --> 00:04:01.199
So here we're gonna go in order.
00:04:01.280 --> 00:04:02.400
I'm gonna tell you the models.
00:04:02.479 --> 00:04:20.079
Uh, so entropic from the least powerful to the most powerful, uh, we have haiku 4.5, then we have Sonnet 5, then we have Opus 4.8, then we technically have Fable 5 slash Mythos 5, kind of the same models, one without guardrails, one with.
00:04:20.480 --> 00:04:21.759
So all right.
00:04:21.920 --> 00:04:25.680
So let's talk about what makes Sonnet 5 good.
00:04:26.000 --> 00:04:28.800
So uh 1 million context win, right?
00:04:28.879 --> 00:04:39.439
That kind of seems to be the norm now, and that includes 128,000 uh tokens of the outputs, it has adaptive thinking on by default and a new tokenizer built in.
00:04:39.600 --> 00:04:41.680
So this is available for everyone.
00:04:41.839 --> 00:04:45.759
And when I say everyone, even if you're on the free plan, which is good.
00:04:45.839 --> 00:04:53.360
Uh, so you know, a lot of times on a free plan, you're getting a less powerful model, and it's one of the most common misconceptions in AI, right?
00:04:53.439 --> 00:04:57.680
I've literally I don't see it as much in 2026.
00:04:58.079 --> 00:05:16.560
Uh, but last year I would still see like literally Fortune 500 level uh enterprise executives leading AI strategy, and they would log into a free, you know, Chat GPT or a free Gemini or a free anthropic uh account, and they would be testing things that way, which is the worst thing to do, by the way.
00:05:16.720 --> 00:05:22.800
But it's good now that Anthropic is upping their default free model to something on the five tier.
00:05:22.879 --> 00:05:24.399
It is a really good model.
00:05:24.560 --> 00:05:31.199
I've actually really enjoyed uh using Sonnet 5 more so than Opus and maybe even Fable.
00:05:31.439 --> 00:05:33.360
We'll get to that here in a little bit.
00:05:33.680 --> 00:05:35.360
So here's why it's useful.
00:05:35.519 --> 00:05:40.000
Well, you have Opus level agentic performance at Sonnet pricing.
00:05:40.160 --> 00:05:47.360
So the pricing, uh, if you are paying on the API side, uh, there's kind of this uh you know discount pricing for a couple of weeks.
00:05:47.519 --> 00:05:55.279
So it's two dollars uh per million um in, ten dollars per million out, then it goes up to three dollars and fifteen dollars.
00:05:55.600 --> 00:05:59.519
So uh if you are paying uh the bill, so to speak, right?
00:05:59.600 --> 00:06:20.879
So if you're not on a subscription, if your company is using um, you know, anthropics models on the API side, Sonnet might actually be a good like delegator model because with how freaking expensive the mythos uh tier models are with mythos 5 and fable five, you can't be using those things for everything, right?
00:06:21.040 --> 00:06:28.639
I'm even having to, you know, shift my own usage just on the subscription plan because you can't run you know fable five for everything.
00:06:28.720 --> 00:06:31.839
Yes, it's out, it's only a limited time, but you know, you can't.
00:06:32.000 --> 00:06:38.000
So you do have to make smart decisions whether you're on a subscription using anthropic or you're on the API side.
00:06:38.079 --> 00:06:40.319
And I've really liked Sonnet so far.
00:06:40.399 --> 00:06:46.800
So it's cheaper than uh, you know, Opus 4.8 with similar-ish, um, you know, similar-ish performance.
00:06:46.879 --> 00:06:54.079
So I would slot it just below Opus 4.8, uh, GPT-5.5, Gemini 3.1, Gemini 3.5, uh Flash.
00:06:54.240 --> 00:06:56.560
So it's it's kind of that tier below.
00:06:56.720 --> 00:07:03.759
Uh, but it's lower, uh, you know, anthropic says lower sycopency, uh, lower hallucination rates, all those good things.
00:07:03.920 --> 00:07:05.920
Um, so who's gonna find this valuable?
00:07:06.079 --> 00:07:13.199
Uh I mean, if you want a daily driver model you can afford to use in anthropic, it's is gonna be much better.
00:07:13.279 --> 00:07:21.279
Uh, you know, if you are on a$20 plan, a team plan, a$200 max plan that I'm on, you can't really use the big model much, right?
00:07:21.439 --> 00:07:24.959
The fable five, uh, and it's gonna be gone in a couple of days, anyways.
00:07:25.040 --> 00:07:27.680
So this is a great, I think, everyday driver.
00:07:27.839 --> 00:07:40.639
Um, and it's probably even if you are comfortable using Opus 4.8, I'm sure we'll see an Opus V in the next, you know, I don't know, three to six weeks, uh, something like that, depending on if it gets held up uh by the US government.
00:07:40.720 --> 00:07:49.519
Uh, but it might just be a good standard practice moving forward for anthropic bottles uh to get used to using Sonnet, right?
00:07:49.600 --> 00:08:05.040
For I think for the majority of knowledge workers out there, I'm not talking about software developers, uh, you know, people running extremely heavy uh agentic workloads, but for everyday usage, this Sonnet 5, I think, could become a de facto driver.
00:08:05.199 --> 00:08:13.360
I do think in the same way, how uh, you know, we are unfortunately in this new era of AI where large language models are no longer democratized, right?
00:08:13.439 --> 00:08:20.720
It's not everyone in the world is gonna have pretty much unlimited use of the best models like we've had up until you know June 2025.
00:08:20.879 --> 00:08:24.720
So, you know, we might have to also start shifting our usage accordingly.
00:08:24.879 --> 00:08:27.519
That's why I think Sonnet 5 is a great model.
00:08:27.920 --> 00:08:30.720
Um, so uh who's gonna find it valuable?
00:08:30.800 --> 00:08:39.440
So, yeah, DevonOps teams, running agents at scale, uh, because it is good for a lot of those lower, uh lower hanging fruit agentic tasks, long-running tasks.
00:08:39.519 --> 00:08:42.720
Uh, and then also, like I said, cost conscious API buyers.
00:08:42.799 --> 00:08:49.519
Uh, so if you want that more OPIS class uh performance, but without the Opus pricing, uh, there you go.
00:08:49.600 --> 00:08:55.120
Or if you didn't win the lottery and can't afford uh Fable, then you know, Sonnet is gonna be your go-to.
00:08:55.919 --> 00:08:58.000
All right, uh, let's keep it going.
00:08:58.240 --> 00:09:01.759
Next, yeah, brain rot that could actually be good for you.
00:09:01.919 --> 00:09:03.440
All right, this isn't really brain rot.
00:09:03.519 --> 00:09:08.080
Uh and and and FYI, I am not a fan of scrolling vertical videos.
00:09:08.240 --> 00:09:12.320
Uh, you know, when open AI came out with their Sora, I'm like, no, this is terrible.
00:09:12.480 --> 00:09:14.000
Please, you know, I don't want this.
00:09:14.080 --> 00:09:15.919
You know, I'm actually glad they killed it off.
00:09:16.000 --> 00:09:22.320
Uh, this one with Notebook LM, this might be Doom scrolling that I will actually start to do.
00:09:22.480 --> 00:09:26.639
That's because Notebook LM has a new video format.
00:09:26.799 --> 00:09:33.919
Uh, it is yes, vertical video 60 second, and they're really good.
00:09:34.159 --> 00:09:45.759
All right, so this is a new studio output in Notebook LM, a 60-second vertical video generated from your uploaded sources with narrated explanations and animations.
00:09:45.919 --> 00:10:00.399
So this was just announced like a couple hours ago, uh, alongside the new and powered by uh the new Nano Banana 2 Lite, which does power the visuals inside of this new 60-second vertical video format.
00:10:00.559 --> 00:10:07.200
Uh, so it is a shorter vertical sibling to Notebook LM's existing cinematic video overview.
00:10:07.279 --> 00:10:09.039
So that's the big difference here.
00:10:09.120 --> 00:10:16.960
So you've inside of Notebook LM uh with their video generations, they've always had the video generations, which are really, really good, by the way.
00:10:17.120 --> 00:10:26.399
But they upgraded um using their uh you know, VO or I guess now omni technology, uh, you know, this cinematic level of video.
00:10:26.480 --> 00:10:30.720
So think uh the normal video overviews, great.
00:10:30.879 --> 00:10:33.840
Um, you know, but there's really not that cinematic feel.
00:10:34.000 --> 00:10:38.960
So this is the cinematic feel of the videos, but just short 60 second verticals.
00:10:39.039 --> 00:10:41.919
So obviously made for learning on the go.
00:10:42.080 --> 00:10:46.960
So I am someone I am not like on any kind of vertical video social media.
00:10:47.120 --> 00:10:49.360
I don't do it, I I refuse to.
00:10:49.519 --> 00:10:56.799
That's why, like when I'm looking at this, it's like I almost don't want to, but I've wanted something like this for a very long time.
00:10:56.960 --> 00:11:02.720
So I might actually uh finally be a short form vertical video uh consumer.
00:11:02.879 --> 00:11:05.200
So who has access right now?
00:11:05.360 --> 00:11:09.840
If you are a paid user, you should have access.
00:11:10.000 --> 00:11:11.759
So they did already start the rollout.
00:11:11.919 --> 00:11:16.480
I think I got access to this on some of my pit accounts, but not all of my pit accounts uh on Gemini.
00:11:16.720 --> 00:11:23.840
So if you are a Gemini paid user, it should you either have it now or it should be rolling out.
00:11:24.159 --> 00:11:32.399
Uh so it is starting with users in English, uh, and then it is rolling out to others in the coming weeks.
00:11:32.480 --> 00:11:36.000
So mobile and non-English rollout status isn't confirmed yet.
00:11:36.159 --> 00:11:38.159
Uh so why is it useful?
00:11:38.320 --> 00:11:50.159
Well, it turns all your long source material inside of Notebook LM into these really cinematic, fun, and engaging uh you know, short 60-second videos.
00:11:50.320 --> 00:11:54.879
So um, notebook LM, I cannot stop gushing about it, right?
00:11:54.960 --> 00:12:05.200
As much as I as I love Codex, as much as I, you know, use Claude Desktop and all these other programs, notebook LM is still one of my go-to's, right?
00:12:05.360 --> 00:12:11.840
One of my most used skills inside of Codex is having it make notebook LM generations for me.
00:12:11.919 --> 00:12:14.320
So I can just wake up, open the app on my phone.
00:12:14.480 --> 00:12:24.320
If I'm out touching grass, taking a walk, I have literally so many uh, you know, usually uh audio uh audio overviews, but sometimes video overviews as well.
00:12:24.559 --> 00:12:32.080
Just bespoke with all the most pinpoint specific information I need to know uh for my day to learn different things.
00:12:32.240 --> 00:12:32.879
So it's great.
00:12:32.960 --> 00:12:34.320
So who's gonna find this valuable?
00:12:34.480 --> 00:12:43.039
I think any obviously notebook online power user, if you're an educator, a student, a student of life, if you're trying to learn something new, uh this is great.
00:12:43.120 --> 00:12:46.240
But also, this is good content if for social teams.
00:12:46.320 --> 00:12:51.519
So if you're a marketer, uh, I mean, these cinematic videos are amazing.
00:12:51.679 --> 00:12:57.919
I'm actually surprised when they were first announced, they didn't like go more viral than they actually did.
00:12:58.080 --> 00:13:08.399
I think one of the reasons, uh, you know, which is a rollout strategy I've never quite understood, um, is they limited this to ultra subscribers when cinematic videos first rolled out.
00:13:08.480 --> 00:13:16.159
So this is cool that the 60 second versions, if you are on a normal paid Gemini plan, uh, you should have access to this.
00:13:16.399 --> 00:13:18.399
All right, let's keep it going.
00:13:19.120 --> 00:13:25.919
Uh, yeah, now you have a new financial assistant in your pocket.
00:13:26.159 --> 00:13:36.559
So our next Friday feature on Thursday is the ChatGPT finance app uh that has now rolled out to plus users.
00:13:36.720 --> 00:13:36.960
Yeah.
00:13:37.200 --> 00:13:40.159
Speaking of that rollout strategy, same thing here.
00:13:40.240 --> 00:13:45.919
This is technically something that came out to pro users on the$200 a month plan a couple of weeks ago.
00:13:46.080 --> 00:13:53.919
I don't think we really covered it much because I tend to not cover things if they're only on that, you know, pricey$200 tier.
00:13:54.080 --> 00:14:02.720
I try to stick to things that are just on the base paid plans because I think that's what the overwhelming majority of you know our listeners uh are on.
00:14:02.960 --> 00:14:06.240
So uh, what is the Chat GPT finance?
00:14:06.320 --> 00:14:19.279
So this is built into Chat GPT and it's OpenAI's personal finance feature that connects your bank and investment accounts via plaid, and then you can ask finance questions grounded in real account data.
00:14:19.440 --> 00:14:23.600
Uh so this was originally launched, like I said, mid-May, but it was pro only.
00:14:23.840 --> 00:14:29.679
But now it is available in live to anyone if you are on that$20 a month plan.
00:14:29.840 --> 00:14:32.000
You've got it now in the US.
00:14:32.240 --> 00:14:39.200
So uh the original Pro launch covered uh web and iOS.
00:14:39.279 --> 00:14:44.159
Uh, I'm not sure if Android is out, so yeah, make sure to check in the newsletter for that and we'll let you know.
00:14:44.320 --> 00:14:51.519
Uh, so uh it is US only, like I said, and here's why I think it's useful.
00:14:51.679 --> 00:14:59.679
Well, you can just make your financial use of Chat GPT way faster and way more personalized.
00:14:59.840 --> 00:15:03.120
A couple of things to keep in mind because I probably hear what a lot of people are saying.
00:15:03.279 --> 00:15:10.960
Oh my gosh, I would never connect my financial data to any AI assistant, any AI account, right?
00:15:11.039 --> 00:15:18.080
So it's connected through Plaid, and Plaid is one of the most trusted uh, you know, names in finance around.
00:15:18.240 --> 00:15:28.480
Um, so you it's not like you're signing up for an unknown uh you know platform or bringing in, you know, if let's just say you have a couple of retirement accounts, a couple of bank accounts, a couple credit cards.
00:15:28.720 --> 00:15:37.360
It's not like you're bringing in, you know, five, six, seven, twelve uh separate entities into Chat GBT and giving them all different permissions and you have to monitor them all.
00:15:37.519 --> 00:15:40.000
It's not like that, it just connects through plaid.
00:15:40.159 --> 00:15:42.799
So it's already coming in through a trusted name.
00:15:42.960 --> 00:15:47.519
Um, you know, according to OpenAI, there's you know data they just can't see, right?
00:15:47.600 --> 00:15:49.039
They can't see your full account numbers.
00:15:49.200 --> 00:15:51.039
There is no right access.
00:15:51.120 --> 00:15:52.159
That's the big thing, right?
00:15:52.240 --> 00:15:54.000
So it's not gonna make any changes.
00:15:54.159 --> 00:16:01.440
So it can only read the information, it doesn't know your full account access, and it is all secure through Plaid.
00:16:02.000 --> 00:16:04.960
Uh, so here is what OpenAI said about it.
00:16:05.120 --> 00:16:16.399
They said today we're releasing, and this is from uh, you know, the original release, but they said today we're releasing a preview of a new personal finance experience in Chat GPT to pro users, now out to everyone.
00:16:16.559 --> 00:16:29.759
Uh, now you can securely connect your financial accounts, see a dashboard of where your money is going, and ask chat GPT questions grounded in your financial context, all while staying in control of your data.
00:16:30.000 --> 00:16:39.679
Um, so they say money, uh money touches nearly every part of life, where we live, what decisions we make, how we care for our loved ones, what future we might imagine.
00:16:39.759 --> 00:16:48.480
But managing finances today often means piecing together accounts, apps, cards, loans, and spreadsheets just to know where things stand.
00:16:48.639 --> 00:17:07.200
Uh, then they say, with your financial apps uh accounts connected, ChatGPT can combine that reasoning with your financial, with your real financial context and what you've shared about your goals, lifestyle, and priorities, helping you spot patterns, understand trade-offs, and plan for big decisions in a way that feels more personal and complete.
00:17:07.359 --> 00:17:20.559
Uh, so yeah, whether you're trying to uh plan you know goals on your spending, uh travel spend analysis, your spending insights, scenario planning, investment risk, subscription reviews, whatever it is.
00:17:20.640 --> 00:17:27.519
Uh so anything that you do give Chad GPT access to inside of PLAD, it can pull all of that information in.
00:17:27.599 --> 00:17:35.200
So not only can you interact with all of your finances in natural language, which is a big cheat code in and of itself, right?
00:17:35.359 --> 00:17:43.519
Because a lot of these uh applications do not have, I mean, many of them do, but not all of them have reasoning AI, right?
00:17:43.599 --> 00:17:49.599
You might be able to pull a certain number and ask it, you know, what was my balance last month in this account, right?
00:17:49.839 --> 00:17:54.720
But many of them do not have AI that can reason over all of this data.
00:17:54.880 --> 00:18:04.000
So not only that, the ability to have a powerful reasoning model like GPT 5.5, but also that it can marry all of your other context, right?
00:18:04.079 --> 00:18:07.119
So your memory inside of ChatGPT and your chat history.
00:18:07.200 --> 00:18:25.759
So when you can throw those things, those three things together, a reasoning model, well, four things, uh, you know, uh a reasoning model, uh, you know, your um bank accounts, your uh, you know, all your financial uh accounts, then your chat history, memory.
00:18:25.839 --> 00:18:27.759
That's that's a huge unlock.
00:18:27.839 --> 00:18:29.200
So who's gonna find this valuable?
00:18:29.359 --> 00:18:32.640
Well, I mean, I think this is big consumer play, uh, right.
00:18:32.720 --> 00:18:36.720
If if you've ever been a user of like a mint uh.com, right?
00:18:36.799 --> 00:18:42.559
There's these popular uh you know personal finance apps, but like let me just call it out.
00:18:42.720 --> 00:18:45.119
A lot of them cost money, right?
00:18:45.200 --> 00:18:55.359
I think that open AI is going to start winning and retaining a lot of consumer users just by replacing other subscriptions that people just used to pay for.
00:18:55.519 --> 00:19:02.240
You know, as an example, I used to pay for mint, and then it got to the point I'm like, you know, why am I still paying for this?
00:19:02.400 --> 00:19:07.279
Uh, you know, and now it's like, okay, well, this is going to be included in my subscription, anyways.
00:19:07.359 --> 00:19:14.079
And it does bring in all of the other personal context that you know ChatGPT has on me, which I'm gonna love.
00:19:14.240 --> 00:19:16.000
All right, our next one.
00:19:16.400 --> 00:19:26.319
This one might sound small, but I want you to think about what it unlocks, especially if you are a power uh Twitter user like me, right?
00:19:26.400 --> 00:19:29.279
This is where I get a lot of my you know AI information.
00:19:29.440 --> 00:19:38.400
I follow the discourse of what's happening, you know, online, you know, in the AI space, mainly on Twitter, because that's where a lot of people are conversing about these things.
00:19:38.559 --> 00:19:53.279
So X, formerly known as Twitter, I still call it Twitter, uh, launched their official MCP model context protocol server, letting all different AI tools connect directly to the X API.
00:19:53.440 --> 00:19:55.279
Uh, so this is big.
00:19:55.519 --> 00:19:57.519
So this is live right now.
00:19:57.680 --> 00:20:12.240
So, you know, whether you want to uh you know go through Twitter if you have a bajillion bookmarks like I do now, uh you you do have to load money uh into the uh into the X account.
00:20:12.319 --> 00:20:17.279
So this isn't something that's free, but it looks extremely affordable and cheap.
00:20:17.440 --> 00:20:20.640
Uh, and I'm gonna tell you the uh the reason why I think that.
00:20:20.720 --> 00:20:22.799
So you do have to, you know, go through that step.
00:20:22.880 --> 00:20:29.119
So it's like, oh bummer, you know, you might spend a couple dollars a month if you're a heavy uh user.
00:20:29.279 --> 00:20:35.119
So you do have to go through, sign up for the X developer portal, uh, you know, go through all that hoops and all that.
00:20:35.200 --> 00:20:53.279
But then at that point, I mean, whether you're a Claude user, cursor, you know, if you're one of the 10 people using GrocBuild, you know, Codecs, whatever, you know, anywhere that MCPs are supported, which is essentially everywhere, you can now scroll X and grab important information uh from there.
00:20:53.440 --> 00:20:59.839
So here is um what was said in a tech crunch article on this.
00:21:00.079 --> 00:21:07.759
So it says this allows developers to save time spent on integration work to focus on whatever they're actually building.
00:21:07.920 --> 00:21:11.599
So, yeah, obviously, if you're trying to build something, this is helpful as well.
00:21:11.759 --> 00:21:20.799
So it says developers have long been able to search X, read posts, look up users, analyze conversations and trends, and do more using the platform's API.
00:21:20.960 --> 00:21:27.599
The hosted MCP doesn't add new capabilities on that front, it just makes them easier to expose to AI applications.
00:21:27.839 --> 00:21:37.039
So by doing so, X can position itself as an information network filled with real-time data to retrieve and analyze rather than just a social hangout.
00:21:37.359 --> 00:21:38.960
So uh for me, this is big.
00:21:39.119 --> 00:21:40.240
So why is it useful?
00:21:40.400 --> 00:21:50.079
Well, it removes the work that developers used to have to do themselves because you used to have to build a custom MCP server, host it, wire it to the X API, all that stuff.
00:21:50.240 --> 00:21:53.440
Now you can set up an MCP server in like a minute, right?
00:21:53.519 --> 00:21:55.440
You can literally just give the URL.
00:21:55.519 --> 00:21:59.680
Uh, you know, once you go through and sign up for the developer account and authorize it, you just give the URL.
00:22:00.079 --> 00:22:03.680
to codecs or you know clawed code and it does it all for you.
00:22:04.240 --> 00:22:10.079
So if you're a develop, if if you're a developer, you know, uh you're gonna find this valuable.
00:22:10.160 --> 00:22:19.839
If you're building social listening tools, competitive monitoring, uh research tools, you know, security researchers, threat intelligence analysts, right?
00:22:20.079 --> 00:22:25.279
Essentially if you spend time for your job um on Twitter, right?
00:22:25.440 --> 00:22:30.880
Following different conversations, what's happening in your industry, how are people reacting to certain things?
00:22:31.119 --> 00:22:38.000
And you are also a user of something like you know Claude on the desktop, uh Codex, etc.
00:22:38.319 --> 00:22:39.759
Bring those two things together.
00:22:39.839 --> 00:22:43.039
Uh I do spend a lot of time on Twitter every single day.
00:22:43.119 --> 00:22:44.720
I'm bookmarking so many things.
00:22:44.880 --> 00:22:53.039
So this is one I'm actually excited for because I'll spend way less time on Twitter and I'm just gonna have Codex go through and do all those things for me.
00:22:53.359 --> 00:22:55.039
So pretty exciting one there.
00:22:56.000 --> 00:23:01.119
Speaking of things you can hand off to others, well now OpenClaw.
00:23:01.200 --> 00:23:08.799
Yes, you can talk with your open claws without having to go onto Discord or Telegram or whatever it is.
00:23:08.960 --> 00:23:13.759
So yeah now your open claw is accessible on your phone.
00:23:13.920 --> 00:23:31.839
That's because OpenClaw the uh open source personal AI agent project launched its first official native mobile app for iOS and Android uh so it is free on the Apple App Store and Google Play.
00:23:31.920 --> 00:23:51.440
Uh you obviously FYI this is just a bridge this is a remote right so it's not like you can skip the whole open claw setup which can be a little cumbersome and you it is a lot of maintenance to keep it up and working I barely use open claw anymore because in my opinion codecs can do just about everything.
00:23:51.599 --> 00:24:03.839
I even text my codecs uh so anyways uh but if you are a heavy open claw user whether it's personally or your organization uh I think this is a great new feature.
00:24:04.000 --> 00:24:42.079
Uh so uh so it is like I said it's not a standalone it's just a companion node that pairs with your posted open claw gateway wherever that is right so most people are using an actual dedicated machine or if you're using a cloud server whatever it is and pairing uh it's pretty easy it's just a QR code and then you set it up once and then the gateway is running so uh this is great because then you can monitor your open claw activities without having to sit down in front of a computer or you know I think what a lot of people were doing is they would just download like you know Discord and then hook it up through Discord or all these other you know third party messaging platforms.
00:24:42.240 --> 00:24:47.440
So now you don't have to do this anymore and you have a dedicated app that can do that for you.
00:24:47.599 --> 00:24:48.880
So who's going to find this valuable?
00:24:49.039 --> 00:25:32.640
Well like I said if you're already an open claw user or if you were on the open claw fence and one of the biggest setbacks is well you wanted a way to be able to uh you know check in on something locally because that's the reality right open claw is great and the fact that you can say oh I have a you know computer running somewhere 247 right like I do a lot of my work in my home office but I have an actual office in Chicago as well and that's where I had my open claw and if I'm being honest I'm like well if I'm only in the office once or twice a week it's like do I really want to be you know going in and you know using Discord which I don't really use anymore right since Midjourney no longer required it like a year or so ago I don't really use Discord.
00:25:32.799 --> 00:25:58.000
So it's like do I want to be using another app just to be messaging uh you know it's another communication channel that I just didn't want to take on so maybe this might you know get me to uh you know re-engage my open claw probably not because I'm plenty happy right now with codex um and with claude code but if this is something that you want to see more coverage on let me know in the comments and maybe we'll do something a little more dedicated.
00:25:58.480 --> 00:26:09.039
Uh and the thing that a lot of people are wondering is like how the heck did this make it onto the Apple App Store uh because they are extremely strict on allowing anything agentic.
00:26:09.119 --> 00:26:14.559
But the reason why is because well none of this is actually stored on your iPhone.
00:26:14.720 --> 00:26:35.839
The iPhone is just a bridge and all of the data lives on your actual machine so yeah no uh according to uh open claw and a lot of the reporting uh you know none of the data actually lives on your phone so it is literally just a bridge or remote control to see what's going on and to be able to communicate with your claws in real time.
00:26:36.079 --> 00:26:40.240
All right two more and we'll make these ones quick even though they're bigger ones.
00:26:40.480 --> 00:26:45.039
All right next we have Cursor released an iOS app.
00:26:45.119 --> 00:26:48.480
Yeah like I said the phones were getting all the love this week.
00:26:48.559 --> 00:26:57.599
So here's what's new Cursor launched its first native iPhone and iPad app in public beta and it is available now.
00:26:57.839 --> 00:27:10.880
So this lets developers launch and monitor AI coding agents from their phone uh as well as their new popular always on cloud agents or just remote control of agents already running on their new computer.
00:27:11.119 --> 00:27:19.279
So who has access well it's live now it's in public beta on the app store and it is gated right now obviously to paid cursor plan.
00:27:19.519 --> 00:27:21.839
So you have to have a paid cursor plan for this to work.
00:27:22.400 --> 00:27:23.759
So why is it useful?
00:27:23.839 --> 00:27:52.240
Well you can kick off or check agent work without keeping your laptop open uh right so no more walking around with your laptop half ajar like I used to do back in February uh with uh codex uh so now you can just have it all running from your phone uh so this uses the iOS live activities and push notifications so you're pulled back in when you want to when an agent finish hits a blocker or if it has a PR ready.
00:27:52.400 --> 00:28:00.240
It does also support voice input screenshot annotation uh as well as full diff review and PR merge from your phone.
00:28:00.400 --> 00:28:03.119
So here's who I think is going to find it useful and valuable.
00:28:03.200 --> 00:28:21.759
I think software engineers obviously teams who are already using cursor uh if you're an on-call person who needs to triage or fix issues away from your primary machine if you want to go outside and touch grass right or teams wanting a lightweight way uh to review and merge AI generated PRs without being at the desk.
00:28:21.839 --> 00:28:31.839
Uh so here is what cursor says cursor in their blog post said cursor is now available as a native iOS app in public beta so you can build from anywhere.
00:28:32.000 --> 00:29:03.359
Until now developers have worked around the limits of their local machines keeping laptops half open and caffeinated everywhere they go yeah that one's fun because there's obviously a popular program called caffeine that keeps your uh laptop awake even when it's closed that's the joke there uh with cursor for iOS you can launch always on agents in the cloud or control agents running on your computer from your phone kick them off when ideas strike get notified when work is ready for review and merge PRs on the go.
00:29:03.680 --> 00:29:27.279
Whether your agents are running on your machine or in the cloud you can move work forward from wherever you are so this really thrusts cursor I think into serious consideration in the super app race right I think we're gonna be talking about the super apps a lot more come like September uh we should be hearing from openai any day on their super app.
00:29:27.359 --> 00:31:06.160
Uh you know Microsoft you know alluded to their super app at the build conference in San Francisco a few weeks ago uh so yes cursor I think uh between the uh SpaceX XAI acquisition so the fact that they essentially have like unlimited compute now number one the fact that their last model they made composer 2.5 was actually really good and cost efficient so you throw in those two things with now uh I think they have a top two harness next to codecs so the combination of a very good desktop app and now maybe the best iOS control pane right uh chet inside of Chad GPT has a great remote control for codecs but I think the cursor might be a little bit ahead mainly because you can start um you know new threads then pick them up on your computer start new cloud agents so there are a couple of dedicated uh a couple of uh small advantages right now I would say that cursor has over codecs slash chat gbt when it comes to managing your local machine uh from your phone all right and then last but not least the technically the biggest story of I don't know the last month or so well anthropics Fable 5 is back yay right probably not all right so uh I would say the overwhelming majority of our listeners are people who are on subscriptions so don't get too excited on this because it's gonna be gone in like a hundred hours uh and then you're gonna have to pay for it out of pocket.
00:31:06.319 --> 00:32:16.799
So uh but we've covered uh Fable five plenty but let me just give you the update and well you can use it now and you absolutely should uh because it is going to be gone in about five days uh so you should be using your entire week limits uh immediately to see if this is something that you are going to pay those actual bills for uh so here is the newest update so uh it said as of today June 30th the export controls on Fable 5 and Mythos 5 have been listed uh fable five will be available starting tomorrow so this was from yesterday's uh no yeah yesterday's blog post uh so available no two days ago uh it said fable five will be available starting Wednesday july 1st to users globally on the Claude platform claude.ai claud code and claude co-work uh and then it says for pro max team and select enterprise plans fable five will be included for up to 50% of weekly usage limits through July 7th after which it will be available via usage credits.
00:32:17.039 --> 00:34:04.960
All right and yeah so uh essentially here let me just tell you what happened um right there's all this this drama that's been going on for the last two and a half weeks uh you know mytho uh anthropic has been hyping uh you know mythos and the project glasswing project since April saying it was a bioweapon you know essentially begging the federal government uh to um you know control ai they said we need to pause AI uh so they got what they wanted uh you know and now frontier models are going through a delayed process to essentially get approved from the government uh right openai had a limited rollout of their GPT 5.6 uh new models that are only available to trusted partners right now uh so that's what started with mythos uh and fable five so fable five is essentially mythos five with added guardrails uh so we got it for three days it was pulled for two weeks now it's back but for the majority of our listeners if you're on a clawed subscription plan you're only going to have it for about five days so that's a big bummer uh and there's essentially this thing called classifiers um and this is something that they put in guardrails uh to satisfy some of the US government's requests uh that this wouldn't be used in nefarious ways right um but they are extremely overbearing right anthropic even said uh in their post to their credit uh that the uh these are gonna take a bunch of uh prompts to Fable and they're gonna say nope you can't do it and they're gonna send it to Opus 4.8 uh so so far it seems like a lot of people are sour on Fable five I'll say just use it for what you can use it while it's there um and see if it's gonna be worth the price of admission.
00:34:05.119 --> 00:34:41.840
All right so for just uh comparison here let me just throw out a a number so yes I'm a heavy ish AI user uh right so inside of codex um I use about two and a half billion tokens a week so we'll say that's about 10 billion tokens a month okay and there's people who use way more right because I'm not a coder yes I'm building projects I have automations and loops and all that good stuff right but I there's plenty of people in your organization that you're listening to who go through way more than 10 billion tokens.
00:34:42.000 --> 00:36:38.960
So if I'm paying for that via Fable five right I'm paying$200 a month to do my you know roughly 10 billion tokens uh like I said give or take um on codecs GPT 5.5 high if I'm paying for that on Fable because in five days that's the only way you can use it that's roughly depending on the blend of you know input output it's about$200,000 uh a month are people gonna pay that yes is everyone gonna pay that probably not so like I said I think you're gonna have to get used to uh the new sonnet five model which I really like uh I'm sure that we'll see an Opus five model pretty soon but go play with opus you know for large organizations you know that have monopoly money uh maybe Fable 5 is something you'll be able to use in the long run but I think for most people go have fun with it you got five days see what happens all right so that is a wrap on our AI features on Friday on Thursday uh I hope this one was helpful as a quick rundown here's what it is Claude Sonnet 5 is out it's good I like it notebook LM short video overviews uh cinematic brain rot that helps you learn not actually brain rot it's really good then you have the new chat GPT finance for all plus users you have the X slash Twitter hosted MCP server uh so essentially I would have my uh agents doing computer use for that which took way too long and a lot of tokens uh so maybe my token count's gonna go down uh then we had open claw coming out with apps for iOS and Android uh cursor coming out with their iOS app and then last but not least Claude Fable 5 is back but not for long.
00:36:39.199 --> 00:36:53.760
All right if this was helpful do me a favor click that subscribe or follow button on Apple Podcast Spotify I'd appreciate that make sure you go to your everydayai.com as a reminder the show and the newsletter will be off tomorrow.
00:36:53.920 --> 00:37:04.159
Happy early 4th of July yeah we have the uh observed holiday this Friday so I'll see you back Monday thanks for tuning in we'll see you back then and every day for more everyday AI.
00:37:04.239 --> 00:37:05.199
Thanks y'all
