WEBVTT
00:00:00.959 --> 00:00:04.799
This week kind of seemed like an upside down week in AI.
00:00:04.959 --> 00:00:06.799
I mean, let's look at the headlines.
00:00:07.040 --> 00:00:12.080
Microsoft is starting a new company with billions of dollars and thousands of employees.
00:00:12.400 --> 00:00:15.199
Anthropic is going to start making drugs.
00:00:15.359 --> 00:00:20.879
The world's most powerful AI model will no longer be available via subscription in a few days.
00:00:21.039 --> 00:00:30.559
And maybe even a weirder one, OpenAI is reportedly considering giving up up to 5% in equity to the federal government.
00:00:30.800 --> 00:00:39.119
Yeah, let's take some time to sort out this week's AI news and see if we can't enter this week's standing right side up.
00:00:39.359 --> 00:00:40.399
Let's get into it.
00:00:40.560 --> 00:00:42.159
Welcome to everyday AI.
00:00:42.240 --> 00:00:45.359
If you're new here, my name's Jordan Wilson, and well, we do this every day.
00:00:45.439 --> 00:00:56.799
This is your daily live stream podcast and free daily newsletter helping everyday business leaders like you and me keep up with the AI madness because, yeah, sometimes it seems like we're flipped upside down.
00:00:57.039 --> 00:01:02.799
I help you every day make sense of it and help you make smart decisions to grow your company and career.
00:01:02.960 --> 00:01:06.719
So it starts here with the unedited, unscripted live stream and podcast.
00:01:06.799 --> 00:01:15.200
But please make sure if you haven't already, to subscribe to the podcast on Apple Podcasts or Spotify and go to our website at your everydayai.com.
00:01:15.280 --> 00:01:16.799
Sign up for the free daily newsletter.
00:01:16.879 --> 00:01:21.120
We're going to be recapping the highlights from today's show and a whole lot more.
00:01:21.359 --> 00:01:22.879
So let's get into it.
00:01:23.200 --> 00:01:24.959
A ton to cover.
00:01:26.719 --> 00:01:30.159
So let's start with Microsoft.
00:01:30.400 --> 00:01:42.959
Yes, Microsoft started a new company uh as they invested$2.5 billion in a new AI client services company called New Frontier Company.
00:01:43.120 --> 00:01:43.599
Yeah.
00:01:43.760 --> 00:01:48.000
So Microsoft Frontier Company, sorry, is the exact name.
00:01:48.159 --> 00:01:59.280
So Microsoft making a major push to help businesses adopt AI with that$2.5 billion investment in the new group called Microsoft Frontier Company.
00:01:59.439 --> 00:02:13.199
So Microsoft will also embed 6,000 employees directly with clients to support AI adoption using the model now very commonly known as Forward Deployed Engineering or FDEs.
00:02:13.360 --> 00:02:26.639
We've seen OpenAI and Anthropic and Google make uh similar plays, but Microsoft taking it kind of almost a step further with a$2.5 billion commitment of its own and 6,000 employees.
00:02:26.800 --> 00:02:36.159
So the new division brings together existing Microsoft engineers, technical consultants, support staff, and salespeople with deep industry experience.
00:02:36.960 --> 00:02:50.639
So the move comes just days after Amazon announced a$1 billion initiative for similar AI support services, highlighting the fierce competition among tech giants to lead in AI business solutions.
00:02:50.879 --> 00:02:52.000
Yeah, weird.
00:02:52.479 --> 00:02:55.120
Everyone's becoming a consultant now.
00:02:55.280 --> 00:03:02.639
So Microsoft's approach aims to help clients navigate tough decisions, such as which AI models to use.
00:03:02.800 --> 00:03:06.319
Like, do you use OpenAI or Andropics models?
00:03:06.400 --> 00:03:08.879
Well, Microsoft is a big investor in each.
00:03:09.039 --> 00:03:25.439
So maybe that's a leg up that the new company, uh Microsoft Frontier Company, will pitch when they're like, hey, you should hire us to help you use AI versus you know, hiring open AI or anthropics deployment companies because we use all of their models.
00:03:25.599 --> 00:03:41.520
Uh, so the company says its platform supports more AI models and integrations than competitors like Palantir, which popularize the forward deploy engineering role in the tech world before all the AI or the large language model companies, I guess, got into the race.
00:03:41.759 --> 00:03:58.240
Microsoft has invested tens of billions of dollars in data centers for generative AI and offers obviously many AI-powered tools and solutions from you know the dozens of you know co-pilot uh offerings across their tech stack.
00:03:58.479 --> 00:04:03.759
Uh, but Microsoft so far, their stock is down a bit this year.
00:04:04.159 --> 00:04:16.079
They're actually down about 20% year to date, the biggest drop among the major tech companies, partly due to concerns that AI could disrupt established software markets.
00:04:16.319 --> 00:04:26.720
So the company did, though, generate about$2 billion in revenue from enterprise and partner services last quarter, up 2.5% from a year earlier.
00:04:26.879 --> 00:04:44.240
So, yeah, this one, not necessarily surprising that Microsoft is starting its own deployment company, but maybe surprising that they committed$2.5 billion and 6,000 employees, especially since Microsoft is in the middle of a lot of cuts, right?
00:04:44.319 --> 00:04:50.959
They've cut thousands of employees uh and kind of pushing the early retirement uh portion as well.
00:04:51.199 --> 00:05:16.160
So it's it's it's one of those things that I think that every single big company uh, you know, that is building any type of AI will have these forward-deployed engineers because what we've seen more than anything else, uh, I think especially it's become very clear over the last year that even companies that are usually somewhat agile and can adopt uh to new digital technologies.
00:05:16.319 --> 00:05:29.199
Well, hardly no company out there has been able to close the capability gap between what AI models can actually do and what the majority of their employees are actually using it for, right?
00:05:29.279 --> 00:05:44.720
Which is why huge companies like OpenAI, Microsoft, Google, Amazon, uh, right, uh everyone, they're now coming up with these roles where their employees go in, work at these other companies and help them essentially use their tools.
00:05:44.800 --> 00:05:48.319
Uh, so it's something we'll probably do a deeper dive later.
00:05:48.399 --> 00:05:51.279
Like, what does an FDE role look like, right?
00:05:51.360 --> 00:05:52.959
What are these companies coming in to do?
00:05:53.120 --> 00:06:05.839
Because I know not every single uh you know uh listener out there is going to be able to at their enterprise bring in uh you know Microsoft FDE or an OpenAI or Anthropic FTE into their organization.
00:06:05.920 --> 00:06:09.839
So we'll probably uh unpack later what that actually looks like.
00:06:10.319 --> 00:06:10.879
All right.
00:06:11.120 --> 00:06:17.839
Speaking of still living in upside-down world, anthropic is gonna make drugs now.
00:06:18.480 --> 00:06:20.720
Yeah, it's apparently a thing.
00:06:20.879 --> 00:06:34.879
So Anthropic has announced Claude Science, an AI workbench designed to streamline scientific research by integrated uh by integrating fragmented tools and data sets into a single environment.
00:06:35.040 --> 00:06:48.480
So yeah, they're making waves with the new Claude Science, uh, making their try to make their splash in the pharma sectors uh as they launch into with Claude Science as they launch into drug discovery.
00:06:48.639 --> 00:06:56.000
So the new Claude Science platform generates figures and visuals aiming to help scientists accelerate the pace of discovery and healthcare innovation.
00:06:56.240 --> 00:07:02.240
Anthropic, though, is already a leading provider of coding tools and AI models, right?
00:07:02.399 --> 00:07:08.160
We all know that, with a strong presence among biotech and pharma customers.
00:07:08.720 --> 00:07:21.839
So at a recent briefing uh event called AI, uh, the briefing AI for science, anthropic revealed plans to help develop drugs faster, focusing on treatments for neglected diseases.
00:07:22.000 --> 00:07:31.040
So this marks one of the most direct attempts by a major AI company to move in drug development rather than just selling AI to other drug makers.
00:07:31.199 --> 00:07:40.879
So the announcement puts Anthropic in a unique position as it will both be a software provider and a potential competitor to its pharma clients.
00:07:40.959 --> 00:08:00.959
And yeah, you're gonna want to stick around uh for this piece later when we talk about mythos and fable five kind of coming back and the data retention policy, because I think it's gonna come into play now that we see companies like Anthropic uh spinning up completely new companies and maybe competing against others that they have access to their data.
00:08:01.120 --> 00:08:15.120
So Anthropic joins a growing race that includes AI first drug companies like Insilico, Google DeepMinds, uh uh Ice of uh Isomophoric Labs, and establish pharma giants building or buying their own AI tools.
00:08:15.199 --> 00:08:24.000
So Anthropic has not provided detailed information yet about its drug development goals or what it will do if it discovers promising candidates.
00:08:24.160 --> 00:08:32.720
But they did say, along with this uh new announcement with Claude Science, that they are going to be working in drug developments.
00:08:33.279 --> 00:08:38.559
All right, more and more, more and more big tech companies doing new things.
00:08:38.799 --> 00:08:47.679
So, yes, first Microsoft with their uh new uh Microsoft frontier company, Anthropic, gonna be trying to discover drugs.
00:08:47.840 --> 00:08:52.399
And now Meta is a neo cloud, they're selling compute, all right.
00:08:52.480 --> 00:09:06.159
So, Meta Platforms is launching their own subsidiary, uh subsidiary called Meta Compute, a new cloud infrastructure business that will sell the company's excess AI computing capacity to outside customers.
00:09:06.320 --> 00:09:23.200
So the move positions Meta as a direct competitor to Amazon with their AWS, Amazon Web Services, Microsoft's Azure, and Google Cloud, marking a major shift beyond Meta's traditional focus on consumer apps and digital advertising.
00:09:23.600 --> 00:09:32.639
So uh Meta's stock did close up about six percent late last week, but still down a little bit month over month and year over year.
00:09:32.799 --> 00:09:45.440
So the company aims to turn its significant investment in AI data centers into a new, in what they hope, a potentially high margin revenue stream by targeting enterprise technology budgets.
00:09:45.600 --> 00:09:57.039
Uh so the success of Metacompute will depend on how quickly the company can attract paying cloud customers and whether this new business offsets the heavy capital spending on AI infrastructure.
00:09:57.200 --> 00:10:10.799
So Metacompute represents the company's clearest attempt yet to monetize its multi-gigawatt AI infrastructure, which has so far mainly powered internal products like ads and recommendations.
00:10:10.879 --> 00:10:18.720
And for those customers, maybe that are using Meta's new Muse Spark model, haven't really met them yet.
00:10:18.799 --> 00:10:32.639
Uh, but entering the cloud market puts Meta in direct competition with established players and also raises questions about how existing cloud providers and AI chip suppliers will respond to a new large-scale entrant.
00:10:33.360 --> 00:10:39.840
So this one, when you look at it now, maybe not super surprising.
00:10:39.919 --> 00:10:46.480
Uh, right, meta has been one of the leaders in building out infrastructure for AI, right?
00:10:46.559 --> 00:10:52.720
They've uh started along with probably uh, you know, X, formerly known as Twitter, right?
00:10:52.799 --> 00:10:56.960
So all of well, now I think they're technically known as SpaceX, right?
00:10:57.039 --> 00:11:00.720
It's all of Elon Musk's uh companies combining into one.
00:11:00.799 --> 00:11:08.159
Uh, I think they've been the leader uh in getting new uh, you know, these gigawatt compute factories up and running.
00:11:08.240 --> 00:11:11.039
But meta is definitely one of the big names, right?
00:11:11.120 --> 00:11:14.799
And being able to essentially churn out these large data centers.
00:11:14.879 --> 00:11:20.720
They obviously uh were using them before to power uh, you know, the meta networks.
00:11:20.879 --> 00:11:35.120
So now when they're trying to use this for their own models, well, uh kind of the chatter and some of the anonymous reporting is like, well, yeah, this makes sense for Meta because they built out all of this infrastructure, right?
00:11:35.200 --> 00:11:45.279
All of this uh tremendous cap X spend, uh, right, billions of dollars buying all these, you know, buying and building all these factories and bringing all this compute online.
00:11:45.440 --> 00:11:49.759
And well, people are saying, well, yeah, Meta's selling it now because they're not using it.
00:11:49.919 --> 00:11:50.480
Uh, right.
00:11:50.559 --> 00:11:58.720
That's the uh the allegations, anyways, is that Meta is getting into this business is well, because they built up all of this.
00:11:58.799 --> 00:12:01.519
Uh, you know, it's kind of like the supply and the demand.
00:12:01.759 --> 00:12:10.720
They built the supply, assuming that their models, right, whether uh externally people using the new meta muse model, which I actually think is a pretty good model.
00:12:10.879 --> 00:12:13.120
I know people kind of have poo-pooed on it.
00:12:13.200 --> 00:12:20.240
I think it's you know pretty good, and we'll see what Alexander Wang and uh Zuckerberg and crew bring out with the next meta models.
00:12:20.399 --> 00:12:26.159
But regardless, there's not many companies that are sitting there saying, hey, we have a ton of extra compute, let's build it.
00:12:26.320 --> 00:12:27.919
Surprisingly, or let's sell it.
00:12:28.080 --> 00:12:29.519
Surprisingly enough, right?
00:12:29.679 --> 00:12:37.600
The other company that did it successfully, albeit is you know, SpaceX slash X slash Rock slash Elon Musk's companies, right?
00:12:37.679 --> 00:12:41.840
And you can make the same argument for them that they didn't really have the demand.
00:12:42.000 --> 00:12:44.320
They thought everyone's gonna be using groc.
00:12:44.480 --> 00:12:49.279
Turns out not a lot of people are, but they've successfully, very successfully, by the way.
00:12:49.360 --> 00:12:56.799
SpaceX has uh been putting out all of these uh you know longer term uh partnerships where they're selling compute.
00:12:56.879 --> 00:13:13.279
So now Meta looks like is hoping to take a page out of SpaceX's or X's book when maybe they were preparing for more demand for their AI models, it wasn't there, and then they quickly sold it uh or rented it out to companies, well, like Anthropic.
00:13:13.919 --> 00:13:31.919
All right, so here's an interesting one that I didn't think was gonna make a splash on this week's AI news, but looking at its potential long-term ramifications, I think this new move from Cloudflare definitely uh, you know, requires us to spend a little bit of time on it.
00:13:32.080 --> 00:13:46.159
So here's what's new Cloudflare has announced that they are blocking some AI crawlers on ad-supported pages, which could reshape how well humans and AI bots access the web.
00:13:46.399 --> 00:13:57.600
So starting in September, Cloudflare will default to blocking AI training and agent crawlers on ad-supported pages while still allowing search crawlers.
00:13:57.759 --> 00:14:06.480
And that's a shift that could affect a large share of the web because well, cloudflare essentially runs like half of the internet.
00:14:06.879 --> 00:14:22.960
So the key change is that Cloudflare will judge crawlers by what they do, not just who they are, meaning that mixed use crawlers that handle both search and AI training could be blocked from training access if publishers choose that setting.
00:14:23.200 --> 00:14:24.720
Here's why that's important.
00:14:24.799 --> 00:14:32.080
Uh, well, Cloudflare says that these mixed-use crawlers account for about 36% of crawler activities.
00:14:32.320 --> 00:14:34.480
So you're like, okay, what's these crawlers?
00:14:34.639 --> 00:14:35.759
Why do they impact me?
00:14:35.919 --> 00:14:44.320
Well, the big one is Googlebot, you know how your company probably has gotten over the years a lot of business, a lot of visibility from Google search.
00:14:44.559 --> 00:14:44.960
Yeah.
00:14:45.120 --> 00:14:50.879
So now a lot of publishers may just go click that block button, and well, maybe it's a good thing.
00:14:50.960 --> 00:14:52.080
It might be a bad thing.
00:14:52.159 --> 00:14:56.480
That's not for us to determine here uh on this news recap show.
00:14:56.639 --> 00:15:07.120
But um, but those bots tied to companies like Google, Apple, and Microsoft uh may now face new limits when their search and training functions overlap.
00:15:07.440 --> 00:15:22.080
So for publishers, the move creates new leverage because public content is no longer treated as automatically free training material for these big AI companies, especially on pages designed to earn revenue from human readers through ads.
00:15:22.240 --> 00:15:39.039
So, according to reports, Cloudflare is moving from a pay-per-crawl marketplace toward a pay-per-use model using the new X402 standard, an open payment protocol reportedly backed by Cloudflare, Stripe, Amazon, Google, and Microsoft.
00:15:39.200 --> 00:15:55.919
So the risk is that AI answers could start favoring large publishers with these licensing deals, making smaller blogs, local news sites, and niche experts harder to find, which matters for anyone building a career or company around online visibility.
00:15:56.559 --> 00:16:00.559
So this one, another one to keep an eye on, right?
00:16:00.720 --> 00:16:04.720
A lot of times we're not paying much attention to the plumbing, right?
00:16:04.879 --> 00:16:09.759
To the AI plumbing, you know, the data centers, but also things like this, right?
00:16:09.919 --> 00:16:17.679
I think we become overly, maybe overly reliant on search inside of AI, right?
00:16:17.759 --> 00:16:24.879
So when you go and you use uh Chad GBT or Codecs or Claude or Claude Code or Google Gemini, etc., right?
00:16:25.039 --> 00:16:30.480
Almost every single query, whether you see it or not, is pulling from the web.
00:16:30.799 --> 00:16:45.759
So there's obviously been a lot of uproar in the publishing community over the past two to three years as more and more of that direct traffic is now just kind of ingested inside of these answers that come inside of these chatbots.
00:16:45.919 --> 00:16:48.000
And well, what does that actually mean?
00:16:48.240 --> 00:16:54.639
Well, it means a lot of things because chances are you might want someone to land on your page and become your customer.
00:16:54.879 --> 00:17:08.720
Or in the case in this example of what Cloudflare is kind of uh uh talking about here, well, maybe you get uh a little penny from Google AdSense or something like that, you know, for every certain number of users that end up on your page.
00:17:08.880 --> 00:17:18.000
So for those people who are more, you know, trying to draw, you know, readers in the masses to come to their website and they just get paid by putting ads on their website, right?
00:17:18.160 --> 00:17:35.680
They've obviously been hit very hard as well by AI, uh by big AI companies, essentially consuming all of this information and then just more or less kind of displaying a um, I guess, collection of the answers, uh, right, inside of the AI chatbot.
00:17:35.839 --> 00:17:39.440
So uh this one from Cloudflare, definitely one to keep an eye on.
00:17:39.680 --> 00:17:44.319
All right, let's take a quick break for a word from our partners.
00:17:44.960 --> 00:17:50.000
Here's what most AI tools still can't do work outside their old little box.
00:17:50.160 --> 00:17:52.480
The all new Slack bot, just change that.
00:17:52.559 --> 00:17:59.440
It's your AI teammate inside Slack, and now it can read, write, and act across the other apps your team already uses.
00:17:59.599 --> 00:18:04.960
No more tool switching, no more re-explaining yourself every time you open a new tab.
00:18:05.119 --> 00:18:11.599
One ops team at engine says the summary feature alone saves them 15 to 20 minutes of use.
00:18:11.759 --> 00:18:15.039
See what Slackbot can do at Slack.com.
00:18:16.640 --> 00:18:23.519
All right, now we've got a lot of news when it comes to AI models and the US government.
00:18:23.680 --> 00:18:31.680
Yeah, we're gonna get to Fable and Mythos and even something with OpenAI, maybe partnering with the federal government.
00:18:31.920 --> 00:18:43.519
But first, we have some new news and developments when it comes to this new uh voluntary model standards as part of President Trump's executive orders on AI models.
00:18:43.680 --> 00:18:55.680
So the US government is moving closer to announcing new voluntary standards for AI models, a step aimed at tightening oversight and addressing national uh security concerns.
00:18:56.000 --> 00:19:09.839
So, uh, according to reports, the US is in advanced talks with leading AI companies to create voluntary standards for releasing new AI models, with an announcement expected as soon as this week.
00:19:10.079 --> 00:19:22.240
So the new standards would set clear benchmarks for advanced AI models, establish release timelines, and clarify who can access these technologies both in the US and internationally.
00:19:22.480 --> 00:19:36.160
Uh, national security is a driving concern, with Washington seeking to prevent misuse of powerful American AI by military intelligence services in China, Russia, and other countries of concern.
00:19:36.400 --> 00:19:50.880
So in June, President Donald Trump issued an executive order requiring federal agencies to work with top AI developers and AI labs to test advanced models before public release and to draft new standards for their deployment.
00:19:51.200 --> 00:20:01.119
Google has been actively involved in discussions with the government, according to reports, particularly regarding the release of advanced coding models with more sophisticated capabilities.
00:20:01.279 --> 00:20:13.839
So the US Commerce Department recently lifted export controls on anthropics, uh Fable and Mythos models just weeks after suspending them due to national security concerns.
00:20:14.079 --> 00:20:24.960
Then we heard from OpenAI when they announced their uh GPT 5.6 uh about a week and a half ago, said that they were doing this tiered release.
00:20:25.039 --> 00:20:36.559
They weren't releasing it to the public, and they were only releasing it to trusted uh testers that had been essentially vetted by the government or approved between OpenAI, the testers, and the government.
00:20:36.720 --> 00:20:50.160
So both OpenAI and Anthropic are preparing for IPOs, which makes this news with a government extremely interesting and one to watch, making government oversight and standards especially relevant for their business strategies.
00:20:50.400 --> 00:21:02.000
And then these voluntary standards could impact how quickly new AI tools reach businesses, developers, and the public, potentially slowing releases but increasing safety and transparency.
00:21:02.960 --> 00:21:08.240
So this is what we all knew was going to be coming, right?
00:21:08.400 --> 00:21:14.799
Uh, the debacle between Anthropic and the US government obviously was pre-right.
00:21:14.960 --> 00:21:16.799
We knew something like that was gonna happen.
00:21:16.880 --> 00:21:23.519
Uh Anthropic and the US government had been going uh tit for tat since you like, I don't know, February.
00:21:23.680 --> 00:21:38.480
So we knew when uh you know Anthropic spent a couple of months hyping up how dangerous uh you know their mythos model was and it was a cyber weapon, and they were begging for uh you know uh government involvement, essentially, right?
00:21:38.559 --> 00:21:40.000
We knew that that was gonna happen.
00:21:40.160 --> 00:22:03.440
But this is a new development that we see that the uh the companies and the US government might be close to announcing something this week on a set of actual standards because one thing that Anthropic responded to, uh, you know, when they announced uh and released Mythos and Fable on a Friday and then By the uh weekend, it was essentially pulled, right?
00:22:03.519 --> 00:22:08.000
So this model was quote unquote out in the public for like two or three days.
00:22:08.160 --> 00:22:21.119
It was pulled, and you know, it seemed like there was not going to be a good uh way to have these set of rules um and safeguards that every single company would have to play by, right?
00:22:21.359 --> 00:22:37.039
According to reports, essentially, we heard it was Amazon CEO uh that ultimately flagged this to the US government and said, Hey, you know, our team here at Amazon found this this jailbreak or this vulnerability in the mythos model.
00:22:37.119 --> 00:22:42.000
They told that to the US government, and then the US government essentially said, All right, well, you got to pull it immediately.
00:22:42.160 --> 00:22:43.519
Uh, Anthropic, right?
00:22:43.599 --> 00:22:48.640
Uh, and Anthropic didn't really have a way to do this, they didn't have a process to go through to appeal it.
00:22:48.720 --> 00:22:54.960
Uh, so this is good news, I think, for all AI companies, especially here in the US.
00:22:55.119 --> 00:23:03.119
Uh, yes, we were supposed to have you know new models from uh open AI and Google that were supposed to land in June.
00:23:03.279 --> 00:23:19.519
Things have been delayed, but I think ultimately uh a couple of weeks of delays uh for ultimately a more secure, uh, more um, you know, strictly vetted AI is probably for the benefit of all of us.
00:23:19.599 --> 00:23:20.720
So, this is one of those things.
00:23:20.880 --> 00:23:34.480
Hey, a couple of weeks of patience and getting something set up with the government, uh, you know, even though you don't want to go into the state of sovereign AI, it's probably for the best versus just having all of these one-off reactions.
00:23:34.559 --> 00:23:45.200
Because if your business, as an example, let's just say your business shut down for a day and said, All right, we're reworking our entire organization around mythos, and then mythos drops, right?
00:23:45.359 --> 00:23:48.000
And it was gone for 19 days, right?
00:23:48.160 --> 00:23:49.119
19 days.
00:23:49.359 --> 00:23:53.359
If your company did that, you may have been kind of screwed, right?
00:23:53.519 --> 00:24:06.079
Uh, I'm sure most companies didn't do that, but that's just an example on why I think we need this set of uh streamlined safeguards and accountability between the US government and the AI companies.
00:24:06.559 --> 00:24:07.039
All right.
00:24:07.200 --> 00:24:12.799
Uh, speaking of that, yeah, the other news is well, Fable V is now back.
00:24:13.039 --> 00:24:14.160
It's actually back.
00:24:14.319 --> 00:24:22.640
So Anthropic has reinstated its advanced Fable V model after a 19-day global suspension.
00:24:22.880 --> 00:24:33.839
And they did add some kind of aggressive new cybersecurity classifiers to address the US government concerns about that jailbreak exploit that I just talked about.
00:24:34.000 --> 00:24:53.440
So Anthropic says the new classifier blocks over 99% of the previously reported exploit attempts, but it is also so strict that it frequently misidentifies harmless software development, coding, and even simple personal biology-related prompts as security risks.
00:24:53.680 --> 00:25:07.759
So when a prompt triggers a new safety filter, uh Anthropic says the system automatically reroutes the task to Claude Opus 4.8 instead of Fable 5 or Mythos 5, whatever you may be using, and then notifies the user.
00:25:07.920 --> 00:25:14.799
That's good, it actually notifies them, leading to a noticeable increase in false positives for regular users.
00:25:15.039 --> 00:25:28.880
So that is, I guess, the good news is that uh after a 19-day delay, uh Mythos 5 and Fable 5, which is essentially mythos five with guardrails, are back.
00:25:29.119 --> 00:25:33.839
But here's the downside it's not truly back for most people.
00:25:34.240 --> 00:25:40.319
Uh so yeah, look at your watch, it's gonna be gone for most users tomorrow again.
00:25:40.480 --> 00:25:54.400
That's because anthropic is not going to include uh Fable Five and Mythos 5 wasn't available anyways, but Fable 5 had been available if you were a paid subscriber, so that's no longer the case.
00:25:54.640 --> 00:26:06.720
So uh after Tuesday, tomorrow, uh, Anthropic says that Fable Five access will only be if you are paying those very expensive API prices.
00:26:06.880 --> 00:26:16.400
They did hint and say, uh, maybe we'll bring it back in the future if we have enough compute, but the only promise is that it's gonna be gone in a matter of hours.
00:26:16.640 --> 00:26:29.599
So Anthropic says the restriction is due to the model's high computational demands and the current infrastructure limits, and they do hope that they might be able to restore broader access once capacity improves.
00:26:30.160 --> 00:26:46.160
So before you go and rush out and you know, have your company use all the limits you can in the next day or two, it is important to remind people that the updated Fable 5 and Mythos 5 explicitly overrides zero data retention.
00:26:46.319 --> 00:26:52.960
So, yeah, you know that whole ZDR thing that makes uh anything in enterprise possible, yeah, that's out the window.
00:26:53.119 --> 00:27:05.440
Because all out all inputs and outputs in Fable 5 and Mythos 5 are stored for at least 30 days, even for those enterprise clients who previously had strict privacy guarantees.
00:27:05.599 --> 00:27:25.599
So if a user's prompt is flagged as a potential policy violation, which we've seen can be literally anything, the data can be held for up to two years for human review, and then automated safety scoring data may be stored for up to seven years to monitor abuse patterns, according to well, anthropic's privacy policy.
00:27:25.839 --> 00:27:30.720
So that's the reason that big enterprises like Microsoft banned internal use.
00:27:30.799 --> 00:27:46.880
Yes, Microsoft, the company that is one of Anthropics' largest investors and makes a ton of money selling anthropics AI through their systems, banned internal use of mythos and fable because you can't get away with them taking your data.
00:27:46.960 --> 00:27:48.160
There's no way around it.
00:27:48.319 --> 00:27:55.279
All right, and yes, that is a first for enterprise AI, has never happened before.
00:27:55.440 --> 00:28:03.440
So they became the first frontier AI provider to essentially say, Hey, yeah, even if you're an enterprise company paying us millions of dollars, yeah, we're gonna keep your data.
00:28:03.599 --> 00:28:10.480
So just keep that in mind before you say, Oh my gosh, you know, we have this fable thing for another 30 some hours.
00:28:10.640 --> 00:28:14.319
Let's go out and dump our company's entire code base.
00:28:14.400 --> 00:28:20.640
Or remember how anthropic is well trying to uh create, go down the drug discovery route.
00:28:20.799 --> 00:28:28.640
Maybe you're a pharma company and saying, Hey, we pay all this money for you know enterprise subscriptions, subscriptions to anthropic.
00:28:28.720 --> 00:28:33.200
We should all go, you know, use our our Fable Five and see what we can cook up.
00:28:33.359 --> 00:28:41.200
Yeah, you might want to pause on that, at least talk to your legal and your IT, and I don't know, maybe take a page out of Microsoft's book and don't do it.
00:28:41.440 --> 00:28:42.000
All right.
00:28:42.319 --> 00:28:47.759
Last but not least, maybe the ultimate curveball in our upside down AI week.
00:28:47.920 --> 00:28:54.640
OpenAI is reportedly considering giving the US government a 5% stake in the company.
00:28:54.960 --> 00:29:04.400
So, according to reports, open AI is in early discussions to offer the US government a 5% stake in the company.
00:29:04.559 --> 00:29:16.000
So the talks are part of efforts by major AI firms to improve relations with the Trump administration as regulatory pressure on the industry increases, according to reports.
00:29:16.160 --> 00:29:27.200
So open AI CEO Sam Altman is advocating for the US public to have a financial stake in the company, aiming to share the economic benefits of AI development more broadly.
00:29:27.440 --> 00:29:39.519
So the proposal could see other leading US companies, including Anthropic, Google, and Meta, asked to give similar stakes to the government, though the participation is not confirmed.
00:29:39.599 --> 00:29:49.680
And yeah, reports said that at least Anthropic and the US government had not yet had these similar uh talks as OpenAI and the US government have.
00:29:49.839 --> 00:30:00.400
So the plan is actually inspired by the Alaska Permanent Fund, which invests oil revenue for public benefits and could require an act of Congress to implement.
00:30:00.640 --> 00:30:10.559
So the discussions are right now being described as consent uh conceptual and at an early stage with no formal agreement yet in place.
00:30:10.720 --> 00:30:23.200
So the move comes as Washington intensifies oversight of AI companies, highlighted by the obvious uh anthropic recent suspension of its latest model due to those national security concerns.
00:30:23.359 --> 00:30:31.920
So both OpenAI and Anthropic have previously floated the idea of a public or sovereign wealth fund to distribute AI-generated wealth to citizens.
00:30:32.079 --> 00:30:45.200
OpenAI and Anthropic are also both obviously preparing to go public, and it is estimated that both companies are gonna come in valued at more than a trillion dollars at their market cap.
00:30:45.680 --> 00:31:05.279
So Altman has reportedly discussed the idea with President Trump, Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessence, and Senator Bernie Sanders, who uh Bernie Sanders, who had proposed a sovereign wealth fund financed by a one-time 50% tax on leading AI companies.
00:31:05.519 --> 00:31:23.200
So, yeah, definitely all the big companies this week, either starting new companies going in completely different directions, or apparently talking with the federal government to give the American people a small slice of the AI pie.
00:31:23.599 --> 00:31:25.359
I don't know, we'll have to find out and see.
00:31:25.599 --> 00:31:27.359
But that was a lot.
00:31:27.680 --> 00:31:33.759
But we always have our little roundup at the end where we go over what's new and what's next.
00:31:33.920 --> 00:31:40.640
So some of these are maybe big stories that any other week might have made our top AI uh news list.
00:31:40.799 --> 00:31:44.319
Others are some rumors, uh, some things around the corner.
00:31:44.480 --> 00:31:45.440
But let's get to it.
00:31:45.519 --> 00:31:48.880
Here we go, our bullet point, what's new and what's next.
00:31:49.039 --> 00:31:56.720
So anthropic released sonnet five as the default clawed model, offering near opus performance for all users.
00:31:56.880 --> 00:32:01.759
Yeah, this didn't even make our top story because Sonnet 5, I personally like it.
00:32:02.000 --> 00:32:07.119
Um, but it's not really even a top three anthropic model right now.
00:32:07.200 --> 00:32:11.519
So I don't even think it's actually that big of a story, although I think it's actually a really good model.
00:32:11.599 --> 00:32:15.839
Uh, next, open claw has landed on iPhone and Android.
00:32:16.000 --> 00:32:20.240
Uh, a new report says OpenAI has found a way to cut inference costs in half.
00:32:20.400 --> 00:32:23.839
And if that's true, that is a monumental deal.
00:32:24.160 --> 00:32:31.920
Apple has added MCP server to Safari, enable a direct AI agent web debugging and preview.
00:32:32.079 --> 00:32:32.880
That's big.
00:32:32.960 --> 00:32:46.000
Uh Meta is reportedly cooking up a new model, which they've codenamed Watermelon, and it allegedly matches GPT 5.5 uh results, and it uses 10 times of Muse's compute.
00:32:46.079 --> 00:32:50.720
So, yeah, their last model was Meta Muse, previously codenamed Avocado.
00:32:50.880 --> 00:32:57.680
So we'll see if this is their medium size or maybe their large size muse model in watermelon.
00:32:57.839 --> 00:33:06.319
Uh, next, a report says that Alibaba banned the use of clawed code and ordered employees to remove clawed models over security concerns.
00:33:06.400 --> 00:33:08.960
We're not sure if that is data retention or something else.
00:33:09.119 --> 00:33:19.119
Uh, a new report says Anthropic is working with Samsung to make its own custom AI chips after OpenAI and Broadcom launched the Jalapeno chip.
00:33:19.200 --> 00:33:24.400
Uh, the White House is set to announce a new AI model and standards, which we just talked about.
00:33:24.480 --> 00:33:31.839
Uh, OpenAI teased a codex-related keyboard device with popular hardware uh maker work louder.
00:33:32.000 --> 00:33:39.839
Uh, the US House passed the Kids Act, adding a new online safety rules for social media and AI.
00:33:39.920 --> 00:33:43.759
It's not a law yet because it still needs the Senate in presidential approval.
00:33:44.000 --> 00:33:51.119
California signed the largest US Gov AI deployment with Anthropic giving Claude at 50% off.
00:33:51.279 --> 00:33:55.599
Google launched Nano Banana 2 Lite and Gemini OmniFlash.
00:33:55.759 --> 00:33:59.279
Anthropic launched Claude Science for drug discovery and biology.
00:33:59.519 --> 00:34:03.359
Anthropic released enhanced admin controls for Claude Enterprise.
00:34:03.519 --> 00:34:11.199
A new report says Zuckerberg uh talked about Meta's AI agents and said that they're progressing slower than expected.
00:34:11.440 --> 00:34:18.320
A new report said Ford brought human engineers back after AI missed quality check issues.
00:34:18.480 --> 00:34:24.719
Uh Grock 4.5 has reportedly entered private beta at SpaceX and Tesla.
00:34:24.800 --> 00:34:29.039
So we'll see what is cooking with SpaceX's next model.
00:34:29.199 --> 00:34:35.519
Uh OpenAI confirmed that GPT 5.6 Soul Ultra will be included in Codex.
00:34:35.679 --> 00:34:48.480
I know that's a small detail, but right now, unfortunately, you don't have GPT 5.5 Pro in Codex, but OpenAI did confirm that GPT 5.6 Soul Ultra, the big bad one, will be included in Codex.
00:34:48.719 --> 00:34:49.360
That's good.
00:34:49.679 --> 00:34:58.320
And then we have a couple things we went over in our Friday feature show, which was actually on Thursday to the uh uh due to the observed federal holiday.
00:34:58.480 --> 00:35:09.360
So uh we talked about Notebook LM uh letting paid users generate 60 second vertical cinematic videos, so essentially 60 second vertical videos now available inside Notebook LM.
00:35:09.440 --> 00:35:14.239
Uh ChatGPD Plus users in the US now have the new finance app.
00:35:14.400 --> 00:35:17.840
Uh X launched a hosted MCP server.
00:35:17.920 --> 00:35:24.639
So now you can get all your Twitter data without having to go through the API, a little easier to send your agents to an MCP server.
00:35:24.800 --> 00:35:33.199
Uh, like I talked about, OpenClaw now has their free iOS and Android app and cursor released in iOS app as well for paid users.
00:35:33.360 --> 00:35:37.280
Uh and it looks pretty powerful by all accounts so far.
00:35:37.519 --> 00:35:40.239
All right, so that is a wrap.
00:35:40.480 --> 00:35:43.679
A ton of big AI news and development.
00:35:43.840 --> 00:35:47.519
So I understand because I do this every single day.
00:35:47.840 --> 00:35:56.079
But if you constantly find yourself trying to keep up with what's happening in the world of AI, that's literally what everyday AI is here for.
00:35:56.239 --> 00:35:58.000
So on Mondays, we do what we just did.
00:35:58.079 --> 00:36:01.920
It's the AI news that matters, mostly facts, a little bit of my commentary.
00:36:02.000 --> 00:36:06.639
Uh, sometimes I accidentally throw in some hot takes, uh, but we do that on Mondays.
00:36:06.800 --> 00:36:14.719
On Wednesdays, we normally do a uh AI working Wednesdays, we go hands-on demo style uh with a popular new release.
00:36:14.800 --> 00:36:18.400
We'll either have a poll in today's newsletter or tomorrow's newsletter.
00:36:18.480 --> 00:36:20.559
So I just do whatever you tell me to do for that.
00:36:20.719 --> 00:36:27.599
And then on Fridays, we do our uh Friday features, which is going over new features that are actually available today.
00:36:27.760 --> 00:36:30.880
So uh, and then Tuesdays, Thursdays, we switch up the shows.
00:36:31.039 --> 00:36:32.559
So I hope this is helpful.
00:36:32.719 --> 00:36:43.519
If so, I'd appreciate it if you could take 30 seconds out of your day uh to go subscribe to the podcast on Spotify and Apple Podcasts, then go to our website at your everydayai.com.
00:36:43.599 --> 00:36:44.559
So thank you for tuning in.
00:36:44.719 --> 00:36:47.840
Hope to see you back tomorrow and every day for more everyday AI.
00:36:47.920 --> 00:36:48.960
Thanks, y'all.
00:00:00.959 --> 00:00:04.799
This week kind of seemed like an upside down week in AI.
00:00:04.959 --> 00:00:06.799
I mean, let's look at the headlines.
00:00:07.040 --> 00:00:12.080
Microsoft is starting a new company with billions of dollars and thousands of employees.
00:00:12.400 --> 00:00:15.199
Anthropic is going to start making drugs.
00:00:15.359 --> 00:00:20.879
The world's most powerful AI model will no longer be available via subscription in a few days.
00:00:21.039 --> 00:00:30.559
And maybe even a weirder one, OpenAI is reportedly considering giving up up to 5% in equity to the federal government.
00:00:30.800 --> 00:00:39.119
Yeah, let's take some time to sort out this week's AI news and see if we can't enter this week's standing right side up.
00:00:39.359 --> 00:00:40.399
Let's get into it.
00:00:40.560 --> 00:00:42.159
Welcome to everyday AI.
00:00:42.240 --> 00:00:45.359
If you're new here, my name's Jordan Wilson, and well, we do this every day.
00:00:45.439 --> 00:00:56.799
This is your daily live stream podcast and free daily newsletter helping everyday business leaders like you and me keep up with the AI madness because, yeah, sometimes it seems like we're flipped upside down.
00:00:57.039 --> 00:01:02.799
I help you every day make sense of it and help you make smart decisions to grow your company and career.
00:01:02.960 --> 00:01:06.719
So it starts here with the unedited, unscripted live stream and podcast.
00:01:06.799 --> 00:01:15.200
But please make sure if you haven't already, to subscribe to the podcast on Apple Podcasts or Spotify and go to our website at your everydayai.com.
00:01:15.280 --> 00:01:16.799
Sign up for the free daily newsletter.
00:01:16.879 --> 00:01:21.120
We're going to be recapping the highlights from today's show and a whole lot more.
00:01:21.359 --> 00:01:22.879
So let's get into it.
00:01:23.200 --> 00:01:24.959
A ton to cover.
00:01:26.719 --> 00:01:30.159
So let's start with Microsoft.
00:01:30.400 --> 00:01:42.959
Yes, Microsoft started a new company uh as they invested$2.5 billion in a new AI client services company called New Frontier Company.
00:01:43.120 --> 00:01:43.599
Yeah.
00:01:43.760 --> 00:01:48.000
So Microsoft Frontier Company, sorry, is the exact name.
00:01:48.159 --> 00:01:59.280
So Microsoft making a major push to help businesses adopt AI with that$2.5 billion investment in the new group called Microsoft Frontier Company.
00:01:59.439 --> 00:02:13.199
So Microsoft will also embed 6,000 employees directly with clients to support AI adoption using the model now very commonly known as Forward Deployed Engineering or FDEs.
00:02:13.360 --> 00:02:26.639
We've seen OpenAI and Anthropic and Google make uh similar plays, but Microsoft taking it kind of almost a step further with a$2.5 billion commitment of its own and 6,000 employees.
00:02:26.800 --> 00:02:36.159
So the new division brings together existing Microsoft engineers, technical consultants, support staff, and salespeople with deep industry experience.
00:02:36.960 --> 00:02:50.639
So the move comes just days after Amazon announced a$1 billion initiative for similar AI support services, highlighting the fierce competition among tech giants to lead in AI business solutions.
00:02:50.879 --> 00:02:52.000
Yeah, weird.
00:02:52.479 --> 00:02:55.120
Everyone's becoming a consultant now.
00:02:55.280 --> 00:03:02.639
So Microsoft's approach aims to help clients navigate tough decisions, such as which AI models to use.
00:03:02.800 --> 00:03:06.319
Like, do you use OpenAI or Andropics models?
00:03:06.400 --> 00:03:08.879
Well, Microsoft is a big investor in each.
00:03:09.039 --> 00:03:25.439
So maybe that's a leg up that the new company, uh Microsoft Frontier Company, will pitch when they're like, hey, you should hire us to help you use AI versus you know, hiring open AI or anthropics deployment companies because we use all of their models.
00:03:25.599 --> 00:03:41.520
Uh, so the company says its platform supports more AI models and integrations than competitors like Palantir, which popularize the forward deploy engineering role in the tech world before all the AI or the large language model companies, I guess, got into the race.
00:03:41.759 --> 00:03:58.240
Microsoft has invested tens of billions of dollars in data centers for generative AI and offers obviously many AI-powered tools and solutions from you know the dozens of you know co-pilot uh offerings across their tech stack.
00:03:58.479 --> 00:04:03.759
Uh, but Microsoft so far, their stock is down a bit this year.
00:04:04.159 --> 00:04:16.079
They're actually down about 20% year to date, the biggest drop among the major tech companies, partly due to concerns that AI could disrupt established software markets.
00:04:16.319 --> 00:04:26.720
So the company did, though, generate about$2 billion in revenue from enterprise and partner services last quarter, up 2.5% from a year earlier.
00:04:26.879 --> 00:04:44.240
So, yeah, this one, not necessarily surprising that Microsoft is starting its own deployment company, but maybe surprising that they committed$2.5 billion and 6,000 employees, especially since Microsoft is in the middle of a lot of cuts, right?
00:04:44.319 --> 00:04:50.959
They've cut thousands of employees uh and kind of pushing the early retirement uh portion as well.
00:04:51.199 --> 00:05:16.160
So it's it's it's one of those things that I think that every single big company uh, you know, that is building any type of AI will have these forward-deployed engineers because what we've seen more than anything else, uh, I think especially it's become very clear over the last year that even companies that are usually somewhat agile and can adopt uh to new digital technologies.
00:05:16.319 --> 00:05:29.199
Well, hardly no company out there has been able to close the capability gap between what AI models can actually do and what the majority of their employees are actually using it for, right?
00:05:29.279 --> 00:05:44.720
Which is why huge companies like OpenAI, Microsoft, Google, Amazon, uh, right, uh everyone, they're now coming up with these roles where their employees go in, work at these other companies and help them essentially use their tools.
00:05:44.800 --> 00:05:48.319
Uh, so it's something we'll probably do a deeper dive later.
00:05:48.399 --> 00:05:51.279
Like, what does an FDE role look like, right?
00:05:51.360 --> 00:05:52.959
What are these companies coming in to do?
00:05:53.120 --> 00:06:05.839
Because I know not every single uh you know uh listener out there is going to be able to at their enterprise bring in uh you know Microsoft FDE or an OpenAI or Anthropic FTE into their organization.
00:06:05.920 --> 00:06:09.839
So we'll probably uh unpack later what that actually looks like.
00:06:10.319 --> 00:06:10.879
All right.
00:06:11.120 --> 00:06:17.839
Speaking of still living in upside-down world, anthropic is gonna make drugs now.
00:06:18.480 --> 00:06:20.720
Yeah, it's apparently a thing.
00:06:20.879 --> 00:06:34.879
So Anthropic has announced Claude Science, an AI workbench designed to streamline scientific research by integrated uh by integrating fragmented tools and data sets into a single environment.
00:06:35.040 --> 00:06:48.480
So yeah, they're making waves with the new Claude Science, uh, making their try to make their splash in the pharma sectors uh as they launch into with Claude Science as they launch into drug discovery.
00:06:48.639 --> 00:06:56.000
So the new Claude Science platform generates figures and visuals aiming to help scientists accelerate the pace of discovery and healthcare innovation.
00:06:56.240 --> 00:07:02.240
Anthropic, though, is already a leading provider of coding tools and AI models, right?
00:07:02.399 --> 00:07:08.160
We all know that, with a strong presence among biotech and pharma customers.
00:07:08.720 --> 00:07:21.839
So at a recent briefing uh event called AI, uh, the briefing AI for science, anthropic revealed plans to help develop drugs faster, focusing on treatments for neglected diseases.
00:07:22.000 --> 00:07:31.040
So this marks one of the most direct attempts by a major AI company to move in drug development rather than just selling AI to other drug makers.
00:07:31.199 --> 00:07:40.879
So the announcement puts Anthropic in a unique position as it will both be a software provider and a potential competitor to its pharma clients.
00:07:40.959 --> 00:08:00.959
And yeah, you're gonna want to stick around uh for this piece later when we talk about mythos and fable five kind of coming back and the data retention policy, because I think it's gonna come into play now that we see companies like Anthropic uh spinning up completely new companies and maybe competing against others that they have access to their data.
00:08:01.120 --> 00:08:15.120
So Anthropic joins a growing race that includes AI first drug companies like Insilico, Google DeepMinds, uh uh Ice of uh Isomophoric Labs, and establish pharma giants building or buying their own AI tools.
00:08:15.199 --> 00:08:24.000
So Anthropic has not provided detailed information yet about its drug development goals or what it will do if it discovers promising candidates.
00:08:24.160 --> 00:08:32.720
But they did say, along with this uh new announcement with Claude Science, that they are going to be working in drug developments.
00:08:33.279 --> 00:08:38.559
All right, more and more, more and more big tech companies doing new things.
00:08:38.799 --> 00:08:47.679
So, yes, first Microsoft with their uh new uh Microsoft frontier company, Anthropic, gonna be trying to discover drugs.
00:08:47.840 --> 00:08:52.399
And now Meta is a neo cloud, they're selling compute, all right.
00:08:52.480 --> 00:09:06.159
So, Meta Platforms is launching their own subsidiary, uh subsidiary called Meta Compute, a new cloud infrastructure business that will sell the company's excess AI computing capacity to outside customers.
00:09:06.320 --> 00:09:23.200
So the move positions Meta as a direct competitor to Amazon with their AWS, Amazon Web Services, Microsoft's Azure, and Google Cloud, marking a major shift beyond Meta's traditional focus on consumer apps and digital advertising.
00:09:23.600 --> 00:09:32.639
So uh Meta's stock did close up about six percent late last week, but still down a little bit month over month and year over year.
00:09:32.799 --> 00:09:45.440
So the company aims to turn its significant investment in AI data centers into a new, in what they hope, a potentially high margin revenue stream by targeting enterprise technology budgets.
00:09:45.600 --> 00:09:57.039
Uh so the success of Metacompute will depend on how quickly the company can attract paying cloud customers and whether this new business offsets the heavy capital spending on AI infrastructure.
00:09:57.200 --> 00:10:10.799
So Metacompute represents the company's clearest attempt yet to monetize its multi-gigawatt AI infrastructure, which has so far mainly powered internal products like ads and recommendations.
00:10:10.879 --> 00:10:18.720
And for those customers, maybe that are using Meta's new Muse Spark model, haven't really met them yet.
00:10:18.799 --> 00:10:32.639
Uh, but entering the cloud market puts Meta in direct competition with established players and also raises questions about how existing cloud providers and AI chip suppliers will respond to a new large-scale entrant.
00:10:33.360 --> 00:10:39.840
So this one, when you look at it now, maybe not super surprising.
00:10:39.919 --> 00:10:46.480
Uh, right, meta has been one of the leaders in building out infrastructure for AI, right?
00:10:46.559 --> 00:10:52.720
They've uh started along with probably uh, you know, X, formerly known as Twitter, right?
00:10:52.799 --> 00:10:56.960
So all of well, now I think they're technically known as SpaceX, right?
00:10:57.039 --> 00:11:00.720
It's all of Elon Musk's uh companies combining into one.
00:11:00.799 --> 00:11:08.159
Uh, I think they've been the leader uh in getting new uh, you know, these gigawatt compute factories up and running.
00:11:08.240 --> 00:11:11.039
But meta is definitely one of the big names, right?
00:11:11.120 --> 00:11:14.799
And being able to essentially churn out these large data centers.
00:11:14.879 --> 00:11:20.720
They obviously uh were using them before to power uh, you know, the meta networks.
00:11:20.879 --> 00:11:35.120
So now when they're trying to use this for their own models, well, uh kind of the chatter and some of the anonymous reporting is like, well, yeah, this makes sense for Meta because they built out all of this infrastructure, right?
00:11:35.200 --> 00:11:45.279
All of this uh tremendous cap X spend, uh, right, billions of dollars buying all these, you know, buying and building all these factories and bringing all this compute online.
00:11:45.440 --> 00:11:49.759
And well, people are saying, well, yeah, Meta's selling it now because they're not using it.
00:11:49.919 --> 00:11:50.480
Uh, right.
00:11:50.559 --> 00:11:58.720
That's the uh the allegations, anyways, is that Meta is getting into this business is well, because they built up all of this.
00:11:58.799 --> 00:12:01.519
Uh, you know, it's kind of like the supply and the demand.
00:12:01.759 --> 00:12:10.720
They built the supply, assuming that their models, right, whether uh externally people using the new meta muse model, which I actually think is a pretty good model.
00:12:10.879 --> 00:12:13.120
I know people kind of have poo-pooed on it.
00:12:13.200 --> 00:12:20.240
I think it's you know pretty good, and we'll see what Alexander Wang and uh Zuckerberg and crew bring out with the next meta models.
00:12:20.399 --> 00:12:26.159
But regardless, there's not many companies that are sitting there saying, hey, we have a ton of extra compute, let's build it.
00:12:26.320 --> 00:12:27.919
Surprisingly, or let's sell it.
00:12:28.080 --> 00:12:29.519
Surprisingly enough, right?
00:12:29.679 --> 00:12:37.600
The other company that did it successfully, albeit is you know, SpaceX slash X slash Rock slash Elon Musk's companies, right?
00:12:37.679 --> 00:12:41.840
And you can make the same argument for them that they didn't really have the demand.
00:12:42.000 --> 00:12:44.320
They thought everyone's gonna be using groc.
00:12:44.480 --> 00:12:49.279
Turns out not a lot of people are, but they've successfully, very successfully, by the way.
00:12:49.360 --> 00:12:56.799
SpaceX has uh been putting out all of these uh you know longer term uh partnerships where they're selling compute.
00:12:56.879 --> 00:13:13.279
So now Meta looks like is hoping to take a page out of SpaceX's or X's book when maybe they were preparing for more demand for their AI models, it wasn't there, and then they quickly sold it uh or rented it out to companies, well, like Anthropic.
00:13:13.919 --> 00:13:31.919
All right, so here's an interesting one that I didn't think was gonna make a splash on this week's AI news, but looking at its potential long-term ramifications, I think this new move from Cloudflare definitely uh, you know, requires us to spend a little bit of time on it.
00:13:32.080 --> 00:13:46.159
So here's what's new Cloudflare has announced that they are blocking some AI crawlers on ad-supported pages, which could reshape how well humans and AI bots access the web.
00:13:46.399 --> 00:13:57.600
So starting in September, Cloudflare will default to blocking AI training and agent crawlers on ad-supported pages while still allowing search crawlers.
00:13:57.759 --> 00:14:06.480
And that's a shift that could affect a large share of the web because well, cloudflare essentially runs like half of the internet.
00:14:06.879 --> 00:14:22.960
So the key change is that Cloudflare will judge crawlers by what they do, not just who they are, meaning that mixed use crawlers that handle both search and AI training could be blocked from training access if publishers choose that setting.
00:14:23.200 --> 00:14:24.720
Here's why that's important.
00:14:24.799 --> 00:14:32.080
Uh, well, Cloudflare says that these mixed-use crawlers account for about 36% of crawler activities.
00:14:32.320 --> 00:14:34.480
So you're like, okay, what's these crawlers?
00:14:34.639 --> 00:14:35.759
Why do they impact me?
00:14:35.919 --> 00:14:44.320
Well, the big one is Googlebot, you know how your company probably has gotten over the years a lot of business, a lot of visibility from Google search.
00:14:44.559 --> 00:14:44.960
Yeah.
00:14:45.120 --> 00:14:50.879
So now a lot of publishers may just go click that block button, and well, maybe it's a good thing.
00:14:50.960 --> 00:14:52.080
It might be a bad thing.
00:14:52.159 --> 00:14:56.480
That's not for us to determine here uh on this news recap show.
00:14:56.639 --> 00:15:07.120
But um, but those bots tied to companies like Google, Apple, and Microsoft uh may now face new limits when their search and training functions overlap.
00:15:07.440 --> 00:15:22.080
So for publishers, the move creates new leverage because public content is no longer treated as automatically free training material for these big AI companies, especially on pages designed to earn revenue from human readers through ads.
00:15:22.240 --> 00:15:39.039
So, according to reports, Cloudflare is moving from a pay-per-crawl marketplace toward a pay-per-use model using the new X402 standard, an open payment protocol reportedly backed by Cloudflare, Stripe, Amazon, Google, and Microsoft.
00:15:39.200 --> 00:15:55.919
So the risk is that AI answers could start favoring large publishers with these licensing deals, making smaller blogs, local news sites, and niche experts harder to find, which matters for anyone building a career or company around online visibility.
00:15:56.559 --> 00:16:00.559
So this one, another one to keep an eye on, right?
00:16:00.720 --> 00:16:04.720
A lot of times we're not paying much attention to the plumbing, right?
00:16:04.879 --> 00:16:09.759
To the AI plumbing, you know, the data centers, but also things like this, right?
00:16:09.919 --> 00:16:17.679
I think we become overly, maybe overly reliant on search inside of AI, right?
00:16:17.759 --> 00:16:24.879
So when you go and you use uh Chad GBT or Codecs or Claude or Claude Code or Google Gemini, etc., right?
00:16:25.039 --> 00:16:30.480
Almost every single query, whether you see it or not, is pulling from the web.
00:16:30.799 --> 00:16:45.759
So there's obviously been a lot of uproar in the publishing community over the past two to three years as more and more of that direct traffic is now just kind of ingested inside of these answers that come inside of these chatbots.
00:16:45.919 --> 00:16:48.000
And well, what does that actually mean?
00:16:48.240 --> 00:16:54.639
Well, it means a lot of things because chances are you might want someone to land on your page and become your customer.
00:16:54.879 --> 00:17:08.720
Or in the case in this example of what Cloudflare is kind of uh uh talking about here, well, maybe you get uh a little penny from Google AdSense or something like that, you know, for every certain number of users that end up on your page.
00:17:08.880 --> 00:17:18.000
So for those people who are more, you know, trying to draw, you know, readers in the masses to come to their website and they just get paid by putting ads on their website, right?
00:17:18.160 --> 00:17:35.680
They've obviously been hit very hard as well by AI, uh by big AI companies, essentially consuming all of this information and then just more or less kind of displaying a um, I guess, collection of the answers, uh, right, inside of the AI chatbot.
00:17:35.839 --> 00:17:39.440
So uh this one from Cloudflare, definitely one to keep an eye on.
00:17:39.680 --> 00:17:44.319
All right, let's take a quick break for a word from our partners.
00:17:44.960 --> 00:17:50.000
Here's what most AI tools still can't do work outside their old little box.
00:17:50.160 --> 00:17:52.480
The all new Slack bot, just change that.
00:17:52.559 --> 00:17:59.440
It's your AI teammate inside Slack, and now it can read, write, and act across the other apps your team already uses.
00:17:59.599 --> 00:18:04.960
No more tool switching, no more re-explaining yourself every time you open a new tab.
00:18:05.119 --> 00:18:11.599
One ops team at engine says the summary feature alone saves them 15 to 20 minutes of use.
00:18:11.759 --> 00:18:15.039
See what Slackbot can do at Slack.com.
00:18:16.640 --> 00:18:23.519
All right, now we've got a lot of news when it comes to AI models and the US government.
00:18:23.680 --> 00:18:31.680
Yeah, we're gonna get to Fable and Mythos and even something with OpenAI, maybe partnering with the federal government.
00:18:31.920 --> 00:18:43.519
But first, we have some new news and developments when it comes to this new uh voluntary model standards as part of President Trump's executive orders on AI models.
00:18:43.680 --> 00:18:55.680
So the US government is moving closer to announcing new voluntary standards for AI models, a step aimed at tightening oversight and addressing national uh security concerns.
00:18:56.000 --> 00:19:09.839
So, uh, according to reports, the US is in advanced talks with leading AI companies to create voluntary standards for releasing new AI models, with an announcement expected as soon as this week.
00:19:10.079 --> 00:19:22.240
So the new standards would set clear benchmarks for advanced AI models, establish release timelines, and clarify who can access these technologies both in the US and internationally.
00:19:22.480 --> 00:19:36.160
Uh, national security is a driving concern, with Washington seeking to prevent misuse of powerful American AI by military intelligence services in China, Russia, and other countries of concern.
00:19:36.400 --> 00:19:50.880
So in June, President Donald Trump issued an executive order requiring federal agencies to work with top AI developers and AI labs to test advanced models before public release and to draft new standards for their deployment.
00:19:51.200 --> 00:20:01.119
Google has been actively involved in discussions with the government, according to reports, particularly regarding the release of advanced coding models with more sophisticated capabilities.
00:20:01.279 --> 00:20:13.839
So the US Commerce Department recently lifted export controls on anthropics, uh Fable and Mythos models just weeks after suspending them due to national security concerns.
00:20:14.079 --> 00:20:24.960
Then we heard from OpenAI when they announced their uh GPT 5.6 uh about a week and a half ago, said that they were doing this tiered release.
00:20:25.039 --> 00:20:36.559
They weren't releasing it to the public, and they were only releasing it to trusted uh testers that had been essentially vetted by the government or approved between OpenAI, the testers, and the government.
00:20:36.720 --> 00:20:50.160
So both OpenAI and Anthropic are preparing for IPOs, which makes this news with a government extremely interesting and one to watch, making government oversight and standards especially relevant for their business strategies.
00:20:50.400 --> 00:21:02.000
And then these voluntary standards could impact how quickly new AI tools reach businesses, developers, and the public, potentially slowing releases but increasing safety and transparency.
00:21:02.960 --> 00:21:08.240
So this is what we all knew was going to be coming, right?
00:21:08.400 --> 00:21:14.799
Uh, the debacle between Anthropic and the US government obviously was pre-right.
00:21:14.960 --> 00:21:16.799
We knew something like that was gonna happen.
00:21:16.880 --> 00:21:23.519
Uh Anthropic and the US government had been going uh tit for tat since you like, I don't know, February.
00:21:23.680 --> 00:21:38.480
So we knew when uh you know Anthropic spent a couple of months hyping up how dangerous uh you know their mythos model was and it was a cyber weapon, and they were begging for uh you know uh government involvement, essentially, right?
00:21:38.559 --> 00:21:40.000
We knew that that was gonna happen.
00:21:40.160 --> 00:22:03.440
But this is a new development that we see that the uh the companies and the US government might be close to announcing something this week on a set of actual standards because one thing that Anthropic responded to, uh, you know, when they announced uh and released Mythos and Fable on a Friday and then By the uh weekend, it was essentially pulled, right?
00:22:03.519 --> 00:22:08.000
So this model was quote unquote out in the public for like two or three days.
00:22:08.160 --> 00:22:21.119
It was pulled, and you know, it seemed like there was not going to be a good uh way to have these set of rules um and safeguards that every single company would have to play by, right?
00:22:21.359 --> 00:22:37.039
According to reports, essentially, we heard it was Amazon CEO uh that ultimately flagged this to the US government and said, Hey, you know, our team here at Amazon found this this jailbreak or this vulnerability in the mythos model.
00:22:37.119 --> 00:22:42.000
They told that to the US government, and then the US government essentially said, All right, well, you got to pull it immediately.
00:22:42.160 --> 00:22:43.519
Uh, Anthropic, right?
00:22:43.599 --> 00:22:48.640
Uh, and Anthropic didn't really have a way to do this, they didn't have a process to go through to appeal it.
00:22:48.720 --> 00:22:54.960
Uh, so this is good news, I think, for all AI companies, especially here in the US.
00:22:55.119 --> 00:23:03.119
Uh, yes, we were supposed to have you know new models from uh open AI and Google that were supposed to land in June.
00:23:03.279 --> 00:23:19.519
Things have been delayed, but I think ultimately uh a couple of weeks of delays uh for ultimately a more secure, uh, more um, you know, strictly vetted AI is probably for the benefit of all of us.
00:23:19.599 --> 00:23:20.720
So, this is one of those things.
00:23:20.880 --> 00:23:34.480
Hey, a couple of weeks of patience and getting something set up with the government, uh, you know, even though you don't want to go into the state of sovereign AI, it's probably for the best versus just having all of these one-off reactions.
00:23:34.559 --> 00:23:45.200
Because if your business, as an example, let's just say your business shut down for a day and said, All right, we're reworking our entire organization around mythos, and then mythos drops, right?
00:23:45.359 --> 00:23:48.000
And it was gone for 19 days, right?
00:23:48.160 --> 00:23:49.119
19 days.
00:23:49.359 --> 00:23:53.359
If your company did that, you may have been kind of screwed, right?
00:23:53.519 --> 00:24:06.079
Uh, I'm sure most companies didn't do that, but that's just an example on why I think we need this set of uh streamlined safeguards and accountability between the US government and the AI companies.
00:24:06.559 --> 00:24:07.039
All right.
00:24:07.200 --> 00:24:12.799
Uh, speaking of that, yeah, the other news is well, Fable V is now back.
00:24:13.039 --> 00:24:14.160
It's actually back.
00:24:14.319 --> 00:24:22.640
So Anthropic has reinstated its advanced Fable V model after a 19-day global suspension.
00:24:22.880 --> 00:24:33.839
And they did add some kind of aggressive new cybersecurity classifiers to address the US government concerns about that jailbreak exploit that I just talked about.
00:24:34.000 --> 00:24:53.440
So Anthropic says the new classifier blocks over 99% of the previously reported exploit attempts, but it is also so strict that it frequently misidentifies harmless software development, coding, and even simple personal biology-related prompts as security risks.
00:24:53.680 --> 00:25:07.759
So when a prompt triggers a new safety filter, uh Anthropic says the system automatically reroutes the task to Claude Opus 4.8 instead of Fable 5 or Mythos 5, whatever you may be using, and then notifies the user.
00:25:07.920 --> 00:25:14.799
That's good, it actually notifies them, leading to a noticeable increase in false positives for regular users.
00:25:15.039 --> 00:25:28.880
So that is, I guess, the good news is that uh after a 19-day delay, uh Mythos 5 and Fable 5, which is essentially mythos five with guardrails, are back.
00:25:29.119 --> 00:25:33.839
But here's the downside it's not truly back for most people.
00:25:34.240 --> 00:25:40.319
Uh so yeah, look at your watch, it's gonna be gone for most users tomorrow again.
00:25:40.480 --> 00:25:54.400
That's because anthropic is not going to include uh Fable Five and Mythos 5 wasn't available anyways, but Fable 5 had been available if you were a paid subscriber, so that's no longer the case.
00:25:54.640 --> 00:26:06.720
So uh after Tuesday, tomorrow, uh, Anthropic says that Fable Five access will only be if you are paying those very expensive API prices.
00:26:06.880 --> 00:26:16.400
They did hint and say, uh, maybe we'll bring it back in the future if we have enough compute, but the only promise is that it's gonna be gone in a matter of hours.
00:26:16.640 --> 00:26:29.599
So Anthropic says the restriction is due to the model's high computational demands and the current infrastructure limits, and they do hope that they might be able to restore broader access once capacity improves.
00:26:30.160 --> 00:26:46.160
So before you go and rush out and you know, have your company use all the limits you can in the next day or two, it is important to remind people that the updated Fable 5 and Mythos 5 explicitly overrides zero data retention.
00:26:46.319 --> 00:26:52.960
So, yeah, you know that whole ZDR thing that makes uh anything in enterprise possible, yeah, that's out the window.
00:26:53.119 --> 00:27:05.440
Because all out all inputs and outputs in Fable 5 and Mythos 5 are stored for at least 30 days, even for those enterprise clients who previously had strict privacy guarantees.
00:27:05.599 --> 00:27:25.599
So if a user's prompt is flagged as a potential policy violation, which we've seen can be literally anything, the data can be held for up to two years for human review, and then automated safety scoring data may be stored for up to seven years to monitor abuse patterns, according to well, anthropic's privacy policy.
00:27:25.839 --> 00:27:30.720
So that's the reason that big enterprises like Microsoft banned internal use.
00:27:30.799 --> 00:27:46.880
Yes, Microsoft, the company that is one of Anthropics' largest investors and makes a ton of money selling anthropics AI through their systems, banned internal use of mythos and fable because you can't get away with them taking your data.
00:27:46.960 --> 00:27:48.160
There's no way around it.
00:27:48.319 --> 00:27:55.279
All right, and yes, that is a first for enterprise AI, has never happened before.
00:27:55.440 --> 00:28:03.440
So they became the first frontier AI provider to essentially say, Hey, yeah, even if you're an enterprise company paying us millions of dollars, yeah, we're gonna keep your data.
00:28:03.599 --> 00:28:10.480
So just keep that in mind before you say, Oh my gosh, you know, we have this fable thing for another 30 some hours.
00:28:10.640 --> 00:28:14.319
Let's go out and dump our company's entire code base.
00:28:14.400 --> 00:28:20.640
Or remember how anthropic is well trying to uh create, go down the drug discovery route.
00:28:20.799 --> 00:28:28.640
Maybe you're a pharma company and saying, Hey, we pay all this money for you know enterprise subscriptions, subscriptions to anthropic.
00:28:28.720 --> 00:28:33.200
We should all go, you know, use our our Fable Five and see what we can cook up.
00:28:33.359 --> 00:28:41.200
Yeah, you might want to pause on that, at least talk to your legal and your IT, and I don't know, maybe take a page out of Microsoft's book and don't do it.
00:28:41.440 --> 00:28:42.000
All right.
00:28:42.319 --> 00:28:47.759
Last but not least, maybe the ultimate curveball in our upside down AI week.
00:28:47.920 --> 00:28:54.640
OpenAI is reportedly considering giving the US government a 5% stake in the company.
00:28:54.960 --> 00:29:04.400
So, according to reports, open AI is in early discussions to offer the US government a 5% stake in the company.
00:29:04.559 --> 00:29:16.000
So the talks are part of efforts by major AI firms to improve relations with the Trump administration as regulatory pressure on the industry increases, according to reports.
00:29:16.160 --> 00:29:27.200
So open AI CEO Sam Altman is advocating for the US public to have a financial stake in the company, aiming to share the economic benefits of AI development more broadly.
00:29:27.440 --> 00:29:39.519
So the proposal could see other leading US companies, including Anthropic, Google, and Meta, asked to give similar stakes to the government, though the participation is not confirmed.
00:29:39.599 --> 00:29:49.680
And yeah, reports said that at least Anthropic and the US government had not yet had these similar uh talks as OpenAI and the US government have.
00:29:49.839 --> 00:30:00.400
So the plan is actually inspired by the Alaska Permanent Fund, which invests oil revenue for public benefits and could require an act of Congress to implement.
00:30:00.640 --> 00:30:10.559
So the discussions are right now being described as consent uh conceptual and at an early stage with no formal agreement yet in place.
00:30:10.720 --> 00:30:23.200
So the move comes as Washington intensifies oversight of AI companies, highlighted by the obvious uh anthropic recent suspension of its latest model due to those national security concerns.
00:30:23.359 --> 00:30:31.920
So both OpenAI and Anthropic have previously floated the idea of a public or sovereign wealth fund to distribute AI-generated wealth to citizens.
00:30:32.079 --> 00:30:45.200
OpenAI and Anthropic are also both obviously preparing to go public, and it is estimated that both companies are gonna come in valued at more than a trillion dollars at their market cap.
00:30:45.680 --> 00:31:05.279
So Altman has reportedly discussed the idea with President Trump, Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessence, and Senator Bernie Sanders, who uh Bernie Sanders, who had proposed a sovereign wealth fund financed by a one-time 50% tax on leading AI companies.
00:31:05.519 --> 00:31:23.200
So, yeah, definitely all the big companies this week, either starting new companies going in completely different directions, or apparently talking with the federal government to give the American people a small slice of the AI pie.
00:31:23.599 --> 00:31:25.359
I don't know, we'll have to find out and see.
00:31:25.599 --> 00:31:27.359
But that was a lot.
00:31:27.680 --> 00:31:33.759
But we always have our little roundup at the end where we go over what's new and what's next.
00:31:33.920 --> 00:31:40.640
So some of these are maybe big stories that any other week might have made our top AI uh news list.
00:31:40.799 --> 00:31:44.319
Others are some rumors, uh, some things around the corner.
00:31:44.480 --> 00:31:45.440
But let's get to it.
00:31:45.519 --> 00:31:48.880
Here we go, our bullet point, what's new and what's next.
00:31:49.039 --> 00:31:56.720
So anthropic released sonnet five as the default clawed model, offering near opus performance for all users.
00:31:56.880 --> 00:32:01.759
Yeah, this didn't even make our top story because Sonnet 5, I personally like it.
00:32:02.000 --> 00:32:07.119
Um, but it's not really even a top three anthropic model right now.
00:32:07.200 --> 00:32:11.519
So I don't even think it's actually that big of a story, although I think it's actually a really good model.
00:32:11.599 --> 00:32:15.839
Uh, next, open claw has landed on iPhone and Android.
00:32:16.000 --> 00:32:20.240
Uh, a new report says OpenAI has found a way to cut inference costs in half.
00:32:20.400 --> 00:32:23.839
And if that's true, that is a monumental deal.
00:32:24.160 --> 00:32:31.920
Apple has added MCP server to Safari, enable a direct AI agent web debugging and preview.
00:32:32.079 --> 00:32:32.880
That's big.
00:32:32.960 --> 00:32:46.000
Uh Meta is reportedly cooking up a new model, which they've codenamed Watermelon, and it allegedly matches GPT 5.5 uh results, and it uses 10 times of Muse's compute.
00:32:46.079 --> 00:32:50.720
So, yeah, their last model was Meta Muse, previously codenamed Avocado.
00:32:50.880 --> 00:32:57.680
So we'll see if this is their medium size or maybe their large size muse model in watermelon.
00:32:57.839 --> 00:33:06.319
Uh, next, a report says that Alibaba banned the use of clawed code and ordered employees to remove clawed models over security concerns.
00:33:06.400 --> 00:33:08.960
We're not sure if that is data retention or something else.
00:33:09.119 --> 00:33:19.119
Uh, a new report says Anthropic is working with Samsung to make its own custom AI chips after OpenAI and Broadcom launched the Jalapeno chip.
00:33:19.200 --> 00:33:24.400
Uh, the White House is set to announce a new AI model and standards, which we just talked about.
00:33:24.480 --> 00:33:31.839
Uh, OpenAI teased a codex-related keyboard device with popular hardware uh maker work louder.
00:33:32.000 --> 00:33:39.839
Uh, the US House passed the Kids Act, adding a new online safety rules for social media and AI.
00:33:39.920 --> 00:33:43.759
It's not a law yet because it still needs the Senate in presidential approval.
00:33:44.000 --> 00:33:51.119
California signed the largest US Gov AI deployment with Anthropic giving Claude at 50% off.
00:33:51.279 --> 00:33:55.599
Google launched Nano Banana 2 Lite and Gemini OmniFlash.
00:33:55.759 --> 00:33:59.279
Anthropic launched Claude Science for drug discovery and biology.
00:33:59.519 --> 00:34:03.359
Anthropic released enhanced admin controls for Claude Enterprise.
00:34:03.519 --> 00:34:11.199
A new report says Zuckerberg uh talked about Meta's AI agents and said that they're progressing slower than expected.
00:34:11.440 --> 00:34:18.320
A new report said Ford brought human engineers back after AI missed quality check issues.
00:34:18.480 --> 00:34:24.719
Uh Grock 4.5 has reportedly entered private beta at SpaceX and Tesla.
00:34:24.800 --> 00:34:29.039
So we'll see what is cooking with SpaceX's next model.
00:34:29.199 --> 00:34:35.519
Uh OpenAI confirmed that GPT 5.6 Soul Ultra will be included in Codex.
00:34:35.679 --> 00:34:48.480
I know that's a small detail, but right now, unfortunately, you don't have GPT 5.5 Pro in Codex, but OpenAI did confirm that GPT 5.6 Soul Ultra, the big bad one, will be included in Codex.
00:34:48.719 --> 00:34:49.360
That's good.
00:34:49.679 --> 00:34:58.320
And then we have a couple things we went over in our Friday feature show, which was actually on Thursday to the uh uh due to the observed federal holiday.
00:34:58.480 --> 00:35:09.360
So uh we talked about Notebook LM uh letting paid users generate 60 second vertical cinematic videos, so essentially 60 second vertical videos now available inside Notebook LM.
00:35:09.440 --> 00:35:14.239
Uh ChatGPD Plus users in the US now have the new finance app.
00:35:14.400 --> 00:35:17.840
Uh X launched a hosted MCP server.
00:35:17.920 --> 00:35:24.639
So now you can get all your Twitter data without having to go through the API, a little easier to send your agents to an MCP server.
00:35:24.800 --> 00:35:33.199
Uh, like I talked about, OpenClaw now has their free iOS and Android app and cursor released in iOS app as well for paid users.
00:35:33.360 --> 00:35:37.280
Uh and it looks pretty powerful by all accounts so far.
00:35:37.519 --> 00:35:40.239
All right, so that is a wrap.
00:35:40.480 --> 00:35:43.679
A ton of big AI news and development.
00:35:43.840 --> 00:35:47.519
So I understand because I do this every single day.
00:35:47.840 --> 00:35:56.079
But if you constantly find yourself trying to keep up with what's happening in the world of AI, that's literally what everyday AI is here for.
00:35:56.239 --> 00:35:58.000
So on Mondays, we do what we just did.
00:35:58.079 --> 00:36:01.920
It's the AI news that matters, mostly facts, a little bit of my commentary.
00:36:02.000 --> 00:36:06.639
Uh, sometimes I accidentally throw in some hot takes, uh, but we do that on Mondays.
00:36:06.800 --> 00:36:14.719
On Wednesdays, we normally do a uh AI working Wednesdays, we go hands-on demo style uh with a popular new release.
00:36:14.800 --> 00:36:18.400
We'll either have a poll in today's newsletter or tomorrow's newsletter.
00:36:18.480 --> 00:36:20.559
So I just do whatever you tell me to do for that.
00:36:20.719 --> 00:36:27.599
And then on Fridays, we do our uh Friday features, which is going over new features that are actually available today.
00:36:27.760 --> 00:36:30.880
So uh, and then Tuesdays, Thursdays, we switch up the shows.
00:36:31.039 --> 00:36:32.559
So I hope this is helpful.
00:36:32.719 --> 00:36:43.519
If so, I'd appreciate it if you could take 30 seconds out of your day uh to go subscribe to the podcast on Spotify and Apple Podcasts, then go to our website at your everydayai.com.
00:36:43.599 --> 00:36:44.559
So thank you for tuning in.
00:36:44.719 --> 00:36:47.840
Hope to see you back tomorrow and every day for more everyday AI.
00:36:47.920 --> 00:36:48.960
Thanks, y'all.
