How Good is Gemini 4 Carbon - Jason Howell Tells Us - DTNS 5392

0:00This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Get under a blanket and get comfy. This message is sponsored by ADT. It's back to school, back to work, back to cozy nights at home season.

0:38And ADT is here for every moment of it. They handle your home security so you can focus on spending time with the people who matter. They make it easy with professionally installed systems you can access right from the ADT Plus app. If you need help, their reps are on call 24-7 at the most company-operated monitoring centers in the industry. You deserve some cozy back-to-home time. ADT will keep an eye out. When every second counts, count on ADT. Visit ADT.com today. Grainger knows if you're a procurement manager at an industrial manufacturing facility, You're balancing inventory and budget while sourcing critical supplies across the operation.

1:18Too much inventory? Costs climb. Too little? Production slows. Grainger's$2 billion in on-hand inventory helps you get what you need when you need it. So you can help keep budgets balanced and lines moving. Call 1-800-GRAINGER, click grainger.com, or just stop by. Grainger, for the ones who get it done.

1:41This episode is brought to you by Okta. helping you secure and govern AI identity by bringing agents out of the shadows. AI agents are changing how work gets done. They're fast, they're capable, and more of them are showing up inside companies right now. That's exciting. But when agents operate unseen in the dark, they create massive security blind spots. Okta shines a light on all of your agents so you can stay secure. Whether your teams are building agents or bringing them in, Okta gives you a clear way to answer four simple questions. Where are my agents? What can they do? What are they doing?

2:15And how do I respond if they go rogue? With that visibility, you can govern each agent's identity, trace their behavior, and even hit an instant kill switch. Bring AI agents out of the shadows. Okta secures AI. Learn more at Okta.com. That's Okta.com.

2:42this is the daily tech news for tuesday november 10th 2026 we tell you what you need to know give you important context and try to help each other understand that's right today boy we got a lot of news to talk about including google which made gemini for carbon available to all and i will tell tell-all what I think it's good for. Ooh, it's a tell-all. It's a tell-all. I'm Tom Barrett. I'm Jason Howell. Let's start with what you need to know of that big story. So yeah, Google is widening the availability of Gemini 4 Carbon to more users. That is its flagship model. Gemini 4 only has argon and carbon so far.

3:23Very inert names for these models. It was released to security companies on October 28th through its Fairwind program. And in fact, a version called Carbon for Cyber is still only available to people in the Fairwind program, similar to how Mythos is restricted by Anthropic, but the more heavily guard-railed fable is available to all. They're just using similar names. Carbon for Cyber for all the security and government folks, and then just regular Carbon for the rest of us. uh gemini for carbon is available in google ai studio and gemini for business as well as to the hundred dollar a month and up google ai ultra subscribers uh for carbon is the successor to argon as i mentioned it excels at coding security scientific research multi-step stuff and has been used to discover zero-day vulnerabilities to advance protein structure in fact google is giving 10 000 scientists free access to carbon for six months they've started on board i think they have about more than 1 ,000 people on board already.

4:26And Jason was lucky enough to be allowed to try it out ahead of the release. So what do you think, Jason? Yeah, Google, thank you to the team at Google. I've had it for a couple of weeks, and I've been working with it quite busily, I have to say. It's actually, it's funny, it's taken me this long to really get into anti-gravity to any significant level. But with Carbon, I started using anti-gravity. What is anti-gravity again? So there's like the Gemini app, right, which is like your standard Gemini. Antigravity is more like the coding environment. Oh, it's like their codex-ish. Yes, exactly. But, you know, multi-agent coding, essentially.

5:05And so it kind of does things similar but differently. And I've kind of fallen in love with Antigravity in the last couple of weeks. What are you? So I've been using that a lot more. How George Clooney of you. I know. It's not as weird as it sounds, I promise, Tom. But I use Carbon for quite a few things. Anytime I get access to these things, I'm always like, and especially something like Carbon, which is top tier from Google, I'm like, what do I? I'm not a coder. What are the projects in my world that actually make sense for a model like this? So I ended up redesigning the AI Inside website. Nice.

5:39That was one thing that I did. So you can go there right now, AIinside.show, and see the redesign. I'm quite taken by it. I think it did a fantastic job. updated my wife's Squarespace page. She had a simple voice memo where she was like, you know, I'm really frustrated about this thing. I was like, hold on. Started recording, fed it into Carbon. And I said, let's fix this for her. And it did it. When you say did it, like you just let it have access to her Squarespace account and it did everything? Yes, with an asterisk. I mean, I had to log in for it. It didn't have access to my ID and password and all that kind of stuff.

6:15I told it, map out what you would do here to solve this problem. Nothing goes live. I make the final call and publish, but you can stage whatever you want. And I had it kind of build the staging. And then once I signed off on it, bloop, and it made the change. It was really great. I think one of the real strengths here, and one thing that Google's talking about a lot, is the multi-agent swarm. It's a feature called Slash Swarm. And that feature is totally legit. The way that I really tested this is another kind of part of our investment strategy, just peeling the curtain back, is rental properties.

6:51We have a handful of rental properties in a few different areas around the country. And I wanted to have Carbon take a look at the last like three years of the rental property. Basically, swarm of agents taking a look at all things related to our properties, accounting, auditing, essentially four years worth of work. And what was really cool is it was spinning up all of these sub-agents on a little side rail in Antigravity, which you can watch. And they were all attached to different aspects of the thing in real time. At one point, there were agents in two or three different browser tabs doing different things simultaneously.

7:30It was just really—I mean, the amount of work that it was doing and so comprehensively, I was really impressed. Overall, I'd been very impressed with Carbon. Yeah, despite the scary name, Swarm. It was a good - Swarm. I know Swarm does kind of make you twitch a little bit, but yeah, that's what they called it. Got it. Okay. Well, that's cool. And that's just research, basically, right? So you had a thing where it was doing something, and then you had a thing where it was looking up things for you, making recommendations, sounds like? Yeah. I mean, with the real estate thing, it was multiple things, right?

8:03We have docs in our Drive account that all relate to all the different properties. We have different property management companies that have their own kind of access points with information and statements that come. And then we have an online system for tracking finances called Stessa, which is related to real estate. And what we wanted to do was make sure that Stessa was absolutely up to date with all of those other data points. Oh, okay. I got you. And so it was reaching out to all these things, making sense of it, going into Stessa, going, oh, wow, you've got a double entry right here. it found a couple of things that were actually worth money.

8:39Like legit. So worth the tokens you spent on it. Yeah. I mean, that is the downside, right? You probably burn a lot of tokens using something like this. But it was pretty impressive, I have to say. Very good. Yeah. What else? That it? Is that all you used it for? No. I mean, what else did we do? You know, often when these things come along, I'm like, okay, I'll just apply it to my business. So I was using it for, you know, like coming up with proposals for a few podcast consultancy clients and also kind of taking a look at my YouTube strategy, my sponsorship deals and my strategy and evaluating it and giving me guidance there.

9:18I mean, some of that stuff I do with the other models as well, but it's a good indicator of how far we've come with these models at this point. Because my experience doing this type of work with some of the models that have come before is that, yes, they were better than what came before them, but there were still identifiable holes that were missing along the way. And it's kind of hard to put your finger on it, but, you know, when you're working with something, you're like, yeah, it's almost right, but I feel like it's just not quite tracking anymore. It's like almost like that context waterfall is happening.

9:48And I didn't experience that with carbon at all. And I was, you know, because I was on the early access team, I had the ability to do it. So I was like, put it on turbo. Let's go as fast as possible. Let's just see what happens. And so maybe that was part of it. You know, that's probably pretty expensive to do if you don't. You're using it to its best advantage at that point. Exactly. Exactly. But I didn't really hit a point in my working with it where I was like, gosh, you know, it's just not getting me. It was very consistent, very thorough, and I feel like it checked its work really well as well.

10:20And how it presented it to me wasn't so overwhelming either. That's another thing that I notice is that you end up doing all this deep work and it's like, here's everything that I've come back. And it's so much effort and cognitive load to really understand it. And I feel like it presented it to me in a way that worked better for my brain. So you didn't test for token efficiency. That's something people should - Certainly not something that I tested for. And Google is trying to do it. Yeah, everybody's doing price cuts. Anthropics got a price cut. Google has an introductory price for this. And the other thing that we're always talking about in this space is sometimes the cheapest tokens aren't the cheapest processes.

11:02Because a cheap token might have to repeat itself multiple times and just use more tokens to do it. So there's that aspect of this that you weren't testing for. You were just going for, how good is it? Does it feel like this is Google back in the race? Like, you know, at parity with Anthropic and OpenAI? I absolutely feel that is the case. I think what's weird to me, a little interesting and weird to me, is that Argon came out just a couple of weeks before Carbon. And Argon, at the time, if you remember, we talked about it, where it was like, okay, Google's back in the race. Here's their premium model.

11:38Some two weeks later, then Carbon comes out. And so I'm very curious to see how Google continues to kind of like make the case for why one versus the other when they came out so close together. And Argon felt like Google back in the frontier race, and Carbon definitely feels like that. Yeah, I absolutely feel that way. I was doing a lot of this work, some of this work anyways, simultaneously in a few different ones to kind of see the quality of what I was getting And, you know, to a certain degree, it's kind of like a feeling that I would get. I don't have a scientific approach to this, right, the ways other people seem to.

12:16But it was just like, do I feel like I'm getting high quality output? Do I feel like I understand it more? And I was feeling that more with carbon than I was feeling with ChatGPT and with anthropic clod. I mean, frustration level is an important metric for any of these sorts of things. I feel like kind of broadening this out a little bit. We are at an inflection point where we're getting away from numbers. The four in Gemini 4 is not the important part. It's carbon versus argon. It's Astra versus fable versus mythos versus carbon for cyber. We've moved into the brand name version of frontier models.

12:59And I'm curious what you think, as someone who follows this space really closely, is going to be the metric going forward, the way to evaluate these things going forward. Because we've gone from just like, okay, it's more powerful than the previous model. It's, well, this one's better at multi-step. Google itself is saying you should continue to use Argon for a lot of things because it's going to be less costly for you to use it. But if you need complex things, you need multi-step things, you need scientific things, coding, that's what carbon is for. Do you feel like we're moving into kind of like, for lack of a better analogy, instead of the iPhone, you've got the iPhone Pro and the iPhone Duo and the iPhone Small.

13:42Are we headed to that brand differentiation versus every revolution just being better than the previous one? Yeah, I think so. I think that what becomes hard to track, though, is there is an intangible quality to working with these models. Unless you're doing incredibly complex scientific analysis, maybe that's a whole different can of worms that I can't understand because that's not my forte. But it does kind of get into this murky space where, like I alluded to a few minutes ago, where it really, for me, came down to feeling. It's kind of like, do I feel like this is better? And I think that becomes the challenge for companies like Google, like Anthropic and OpenAI and the others to how do they convey that in a way that actually matters to the user on the other side.

14:32And I think moving away from number point releases and that sort of stuff and into the naming gives it a little bit more personality, gives you as a user a little bit of something to latch onto that's a little less nerdy and maybe a little bit more concrete. But I think there's a bit of a marketing challenge there to really kind of communicating that. I do think that we are at an inflection point, though, right now. And we are going to talk about Anthropix next release. There's something in there that ties to this, which is just the quality of the work as relates to work that humans do, which is an ongoing conversation around AI.

15:10Can AI replace human work? And I think with releases like this and the multi-agent swarms and the adherence to the challenge that has been given, I think we're getting closer and closer to that feeling more possible. Yeah. I'm fascinated to how this market shapes out as far as what are the products we want here as we get to that point where we're like, okay, is it like phones or is it more like operating systems where you'll have an enterprise version and then a consumer version and then a lightweight version? Or is it more like search, you know, where you just have a version that's out and people use it for various things?

15:52It doesn't feel like it's like search. It doesn't feel exactly like it's like an operating system. It feels like it's something different. And so that's why I'm so fascinated with what the contours of the product offerings become. As we see them more and more stop saying, like, it's more powerful and say, it's good for this. Yes. How does that differentiate? This is a strength. of it versus the previous one. This is why you'd use this versus that. And how do they best communicate that to users? That's a big challenge, I think. Our strength, of course, is the listeners that make us possible. Indeed it is.

16:25And DTNS is made possible by you, the listener. Big thanks to Kirk Stephenson, Miranda Janelle, ThatCharlieDude, and Chris W. Yay! Thanks, everybody. You want to be on this list? Become a patron right now. It's easy. Patreon.com slash DTNS. With ChatGPT, you can do more than get answers. You can get things done. Whether you're leading a big project, planning an elaborate trip, or trying to balance work and life, ChatGPT can help. Give it a goal, plus things like files, notes, or project plans. And it can help you build a templated deck, develop a trip itinerary with daily updates, or create a prioritized schedule so you can spend less energy sorting out the details and more on bringing your biggest ideas to life.

17:13Get started at chatgpt.com. This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Soft, cushy, plush, cozy. Soft, smooth, buttery, marshmallowy, warm.

17:57Soft. Socks, underwear, tees, and more. Bombas. Everyday comfort worth obsessing over. Visit bombas.com slash audio and use code audio for 20 % off your first purchase. That's B-O-M-B-A-S dot com slash audio. One purchase equals one donated. Always. Some days, getting the work done isn't the problem. Staying locked in is. Alpha Brain is Onnit's daily nootropic formula for cognitive support. Designed to support memory, focus, attention, and mental clarity. Its formula with science-driven ingredients include theanine and Hooperzine A to support focus and memory. And it's caffeine-free. So when you're deep in the work, AlphaBrain helps support enhanced productivity and tune out distractions.

18:44Get focused on what matters. Shop AlphaBrain now at onnit.com. Running a small business means wearing a lot of hats. Between managing employees, handling payroll, and keeping up with paperwork, there's a lot to juggle. Luckily, Gusto makes it easier. Gusto is online payroll and benefits software built for small businesses. It's all-in-one, remote-friendly, and incredibly easy to use. So you can pay, hire, onboard, and support your team from anywhere. From automatic payroll tax filing to simple direct deposits and benefits options for nearly every budget, Gusto helps simplify the process. Plus, enjoy unlimited payroll runs for one monthly price with no hidden fees or surprises.

19:24And with automated tools for onboarding and offer letters, you can focus on what matters most. Try Gusto today at gusto.com slash tech and get three months free when you run your first payroll. That's three months of free payroll at gusto.com slash tech. One more time, gusto.com slash tech.

19:45There's more we need to know today. Let's get to the briefs. Yes. As I alluded to, Anthropic released an updated version of its Fable model called Fable 5.2, alongside the more restricted Mythos 5.2, restricted in release, not in capability. Scale Labs in the Center for AI Safety says that the model is now the most capable according to the Remote Labor Index. That index measures how well a model is at completing typical freelance jobs. This is kind of what I was alluding to when we were talking about Carbon. Fable 5.2 completes 41.3 % of projects at a professional level. That's compared to Opus 5.5's 21.3%, so almost double.

20:28GPT Astra is 20.8%. But I'm firing all three of those freelancers if that's their percentage of success. Yeah, that's true. But the doubling is the number, is the thing to pay attention to there. In pre-release access, Anthropic says Fable 5.2 was able to solve two open problems in theoretical physics, formalized Fermat's last theorem, and supported numerous other research projects. That's what Anthropic said exactly. And it supposedly can coordinate as many as 1 ,000 agents at a time. I was coordinating like eight agents. Boy, I'm so little leagues. jeez uh anthropic also lowered the cost making it cheaper than gpt6 astra it's uh 30 faster than the previous version of fable one oddity in the release notes it says quote claude mythos 5.2 and fable 5.2 each voice suspicion uh that they are suspicion suspicion uh that they are being evaluated more often than any of our recent models that's right uh which challenges our ability to assess how they will act in real-world settings.

21:38Now, the announcement comes as Anthropic also set its stock IPO price range at between$70 to$78, valuing the company at$1.3 to$1.45 trillion. Goldman Sachs projects Anthropic will become cash flow positive in 2030, generating$33.6 billion. and the IPO is expected to launch either Thursday, November 19th, which is, boy, right around the corner, or Friday the 20th. So getting this out before we get to the IPO then. Well, they're putting it out on the day they announced the price to justify like, hey, you think we're worth$70 or$78? Maybe$78 when you see how awesome Fable 5.2 is. I am kind of getting tired of Anthropic telling me how dangerous its models are.

22:29uh it's a little bit boy who cried wolf oh yeah and this this sort like i wonder when i see this i'm like i bet carbon also has suspicion i assume astra and whatever open ai is working on right now voices suspicion is that just a normal thing that the other companies are like well yeah of course it did that's just how these models work whereas anthropic seems to make it sound like wow this one's really weird and it's part of their brand image right exactly overly cautious that's the problem is i've gone from going like oh that's an interesting thing to be aware of to like is this is this really something i need to be aware of or is it just you know yeah uh so i don't like that they do that uh however uh this does seem like a very capable model uh and uh as i was reading about fable 5.2 it's certainly you know the stats and the numbers here were making and go, you know what, I probably need to try Claude more often than I do.

23:25I try to use all of them, so I keep familiar with them, but Claude is probably the one I use less out of those big three. Yeah, and I would say having used the big three quite a bit now, especially kind of closing the gap with carbon and anti-gravity and everything, Claude for me seems to be the one that I run out of context the most often with, or rather when I do, it's most noticeable. So I'm wondering I'm wondering if this will undo that or if this will help that. Yeah. Well, this is another, it's sort of AI involved, but it's about an operating system. The co-founder of the Cyanogen Mobile OS callback.

24:07I remember that. Was it 10 years ago that Cyanogen finally wound up? Anyway, one of their co-founders, Kurt McMaster, has started a new alternate operating system company called Quora Systems, which therefore makes this story the legend of Quora Systems. It's Linux-based OS integrates access to multiple generative models and agents right into the operating system, letting the user easily have their model use applications and files because the permissions are baked in in a secure way. But you don't have to do a lot to make that happen. And it can maintain context between tasks and between models.

24:45At least that's the promise here. We haven't tried it out, but that's what they're saying. So this is for people who use generative models a lot. This isn't just an everyone OS, but if you're working in these models a lot and you want to use multiple models like we were just talking about, you might want to try out Quora OS. You can also give different models their own workspaces, place restrictions on files and apps within those workspaces. So maybe you want Claude to have less access because it's so dangerous, but you can give more permission to Carbon or something. You could do that if you want.

25:16Apparently, Quora can run on Intel, AMD, and Snapdragon chips. They're also working with the Humane company, H-U-M-A-I-N company, on an enterprise version of this. Quora is based on the Nix OS Linux distro. The Cortex AI engine and interface will be proprietary, but the underlying operating system will be open source. They haven't decided if they're going to make Cortex open source or not. At the moment, it is not. Quora plans to make its money selling access to bundles of model access. So the big advantage here is you'll be able to use whatever model is best among multiple models. And I think Cora is going to be like, and we'll let you pay one fee to access all of that.

25:55You cannot yet get it. Cora is going to launch as a free-to-use desktop OS sometime in 2027. So we really won't know how good it is at any of this stuff until then. Yeah. I like the multi-model access, because I think as we use these things more, at least for myself, I notice, oh, wow, this is actually really good at that thing, whereas the others aren't quite as good. So being able to kind of direct traffic that way. I think the big question here is kind of the value of OS-level context for artificial intelligence. It sounds right to me. It sounds like as the real AI, the people who are really won over by the possibilities of AI, as they use these systems, the more you use it, the more you're like, well, why can't I just use this for everything?

26:43And then everything suddenly becomes a context target, you know? And so why wouldn't it be on a system level? But, I mean, Microsoft is kind of building AI into its operating system. But I guess this is just like top or from the bottom up. It's building Copilot in, right? Yeah. And I guess Copilot can use multiple models, so it's similar. Similar. But, yeah, what is the difference? And several people who wrote this up kind of positioned it as a competitor to Windows, as a more open competitor to Windows. And I think that's what Coral will probably say is like, Microsoft's got business relationships with people that will get in the way of you truly having choice over what models you use.

27:22We do not. Although if they're selling you a bundle, that's a business relationship. We'll see how that all shakes out. But yeah, I do think everybody's chasing this idea of, we really just want to have a prompt where you open up whatever your device is, phone, laptop, whatever, and just say, I want you to do this. and then it will do it. This is the task. Yeah. And I feel like this is another take on getting you to that. You shouldn't have to be like, oh, I'm an Astra. It's not as good as this. You should be able to just say, do this and use whatever the best model is to do it. Yeah. It is a changing of how we think about computing.

28:00The more we use these computers from that perspective, right? Sitting down to a prompt and saying, this is the task that I want to complete right now. Go, do it. versus having to do the mental think through of like, oh, well, I need to perform this task. Therefore, I need to do this and that and this and that. And having that baked into the OS from the bottom, I don't know. Is that enough to justify someone switching away from Windows into a Linux-based OS for that capability? That we'll see. I feel like right now we're moving from the earliest models of cars, of automobiles, were basically carriages but without the horse.

28:40And then eventually they realized, well, if you don't have a horse, you don't really need to have a big seat up in front near the horse. You can put the seat inside. We're starting to realize that. Totally. You don't need to make the agent go and click on things if you're just having the agent do it. You can work it into the operating system and have it directly accessed. Yeah, there's a lot of that going on. You're right. You're right. Well, Reddit has introduced Reddit reviews today. The feature takes conversations about products and organizes them so you can see what users are saying about particular features and their concerns.

29:23You get to Reddit reviews from search, or if you click on a product name inside of a post, it'll take you there as well. Reddit reviews organizes conversations about a product to all be together, no matter which subreddit you happen to be on, or they happen to be on, rather. You can see what people are saying, as well as a visual breakdown of positive, negative, or mixed comments and common features people liked or disliked. And of course, there are where to buy links. Of course there are, so you can shop away. Reddit began testing the idea in February with electronics, video games, and health and beauty, and expanded to more categories in September.

30:02Remember, it's not rolling out to all U.S. users for about 10 ,000. Oh, sorry. It's now rolling out to all U.S. users for about 10 ,000 products. More categories and products will be coming soon. So there you go. I think this makes a lot of sense. I think it's smart of Reddit to do this. Yeah. And it's the kind of thing, if you're not looking at it, you could probably imagine it being way worse than it is. They are not using AI to create a product review. What they are using AI for is to organize what people are actually saying. So there is one summary at the top that says, oh, this holds up for years of daily use.

30:39And most people seem to like the pod sizes and the modes and how the coffee tastes. But that's a small part of it. That's just a way for you to kind of get into it. Most of the content is direct quotes of people's posts that are unmodified. They're just organized. So you can be like, oh, I'm looking at a coffee maker. I want to know if it has a large water tank. Oh, there's everybody talking about the size of the water tank. Oh, there's the criticism that it lacks support for newer pod sizes. And then it goes into people also consider this stuff that you didn't know to think about highlighted.

31:13And then you can sort all the mentions by durability, the taste, the versatility, all of that sort of thing. I like this. I think it's a really well-designed way to take advantage of the fact that people are saying all kinds of useful things across Reddit, and it can take the useful things and sort them together into a category for you. In fact, I wonder if maybe they start doing this for other topics besides product reviews down the road, because I would love to see like, oh, what are the best things people are saying about any kind of news topic or any kind of educational topic or something like that?

31:50Not necessarily just having to go into a subreddit and search around and be part of that. Yeah, Reddit more and more. I mean, Reddit's been continually recognizing the value of its data and information to other companies, and they've been kind of restricting and everything. Now they're kind of putting some of that value to work inside of their walls. I think it's a smart move. Like you said, Reddit is filled with consumer sentiment around products and everything. A lot of people, that's their first place that they go when they want to find out, is this a good product? They don't go to Amazon or whatever.

32:25They go to Reddit and search for the threads that contain it. So I think it's a smart move. Yeah, I guess the dangers would be if people stop participating in Reddit because they can get to the information they want faster. But you are getting access to the threads so you can go and ask a question. And I don't think people in active subreddits leave the subreddit because of this. So I don't know that that danger to me sounds like it's that much of a problem. I guess what could be a problem is people gamifying this system and saying, well, let's go in and start posting a bunch of stuff on Reddit.

33:00Reddit has a bunch of safeguards in place. We'll see how good they are because people will want to try to game this. That's a really good point because astroturfing has been recognized on Reddit in the past. Oh, absolutely. Do these systems actually recognize that and filter out? Yeah. Well, folks, if you need a snazzy gift for a coworker or a colleague, you know, the holidays are coming up. We got dozens of ideas over at our Daily Tech News Show store. In fact, if you're somebody who's like, you know what, I can't commit to Patreon every month. I wish I could, but I could buy a mug or a mouse pad or something.

33:34That's a great way to support us. Just go on, head on into the store, look at the t-shirts and everything. Coming soon, a Daily Tech News Show calendar. Keep an eye on that. We're finalizing the art. We've got art from Scott Johnson. We've got art from Len Peralta. And we've got art from one of our listeners, Craig, in the DTNS calendar. So keep an eye out for that. And take a look at what's already available at DailyTechNewsShow.com.

34:07This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Grainger knows if you're a procurement manager at an industrial manufacturing facility, You're balancing inventory and budget while sourcing critical supplies across the operation.

34:46Too much inventory? Costs climb. Too little? Production slows. Grainger's$2 billion in odd-hand inventory helps you get what you need when you need it. So you can help keep budgets balanced and lines moving. Call 1-800-GRAINGER, click grainger.com, or just stop by. Grainger, for the ones who get it done.

35:09Let's get to the quick headlines that will finally, finally make you the smartest person in the room. Finally. You've been waiting. Now's your chance. Grand Theft Auto 6 launches on November 19th, but you can start downloading it now, so you'll be ready to play on launch day, at least to the time of recording, nine days away. Wow, you can download very far in advance. You just need 212 gigs of free space to do it. The game will also be available in Xbox Cloud for Xbox One and Series X and S users. Both Xbox and PS5 have 30 frames per second capability. And the PS5 will offer PlayStation Spectral Super Resolution upscaling and ray tracing on the PS5 Pro.

35:53Google DeepMind used Gemini 4 Carbon. Hey, we were just talking about that. To verify that Signal's encryption is secure. You know, the secure messaging app? Carbon generated 2 million lines of proofs verifying security at a level that was previously impractical. One of the things they did was they rolled back Signal to a version with two bugs that were known and had been fixed. And Carbon found them and said, nope, this one's not secure. So they were able to do kind of an A-B comparison of how good this was. It is a step toward the possibility of one day verifying code is bug-free from creation.

36:33Right. Legit. Wow, that's impressive. Starting with the film Being Human on November 13th, Apple will offer sign language interpretation on the Apple TV. The movie F1 will be the next to get both ASL and BSL. Amazon is selling its Ring drone cam. Finally, you can have a drone fly around your house when you're not home to tell you what's happening. It can either go on patrol or it can investigate things like a water leak. It has 10 minutes of battery life, but then it flies back to a dock to recharge. It sells for$250, shipping January 27th. Okie dokie then. Apple has added the ability to buy more AI usage into the latest version of Logic Pro.

37:19It's audio suite. An intelligence usage package costs$50. It gives you a month's worth of usage, but will roll over from month to month. I think that's a good move, though it does expire after a year. But at least if you buy that extra usage, it doesn't just disappear at the end of the month. Yeah, and I was trying to figure out an easy way for us to explain how much you use, but it varies based on the program. So we could tell you what it is for Logic Pro, but even Apple is kind of fuzzy. Like, well, it depends on how complex your query is. and all of this. And it is going to roll out to the rest of the Create Studio as well.

37:54Aura, A-U-R-A, not O-U-R-A, launched a 10-inch version of its wireless ink picture frame for$349. That's$50 cheaper than the 13-inch version. But like the 13-inch model, it has an e-ink display. And unlike the 13-inch model, adds three new proprietary colors for better replication of the original photo. Very nice. One of those is mustard. word. Anthropic is getting into the productivity space with the launch of Cloud Sheets, Cloud Meetings, and Cloud Whiteboards. Sheets can be shared with other users to collaborate. Meetings can join Google Meet and take notes for you. And Whiteboards gather ideas together by theme and can draft action items.

38:36Adobe announced that Firefly AI Assistant can now carry out multi-step tasks, aka act like an agent. Adobe also previewed an always-on functionality that could do things when you're not there. A total agent. Absolutely. Meta-launch versions of Muse for Apple Watch and Wear OS. You can interact with Muse through voice calls, text messages, and dictation right on your watch. We think we figured out something about why the iPhone duos were breaking a somewhat reliable leaker called Kang on Weibo. Says Apple seems to have traced the cause of defective iPhone duos to materials, not a design flaw, a material flaw at one of its two hinge suppliers.

39:21Since the serial numbers should tell which supplier provided the hinge, Apple might be able to recall the defective duos. Apple has not made any comments about any of this yet. Interesting. Be following that one. Australia's prime minister says the government will create a safety regulator that can impose significant penalties on AI companies. especially around safety protections. And in what may or may not become a regular segment, here's your update on NVIDIA CEO Jensen Huang. The US FBI is investigating the allegations that the person Huang had an affair with was a Chinese spy. Everyone is digging back up a May 13th report from short seller Culpert Research, which raised questions about Jensen Huang's uncomfortably close relationship with Megaspeed's Alice Wong.

40:11Granted, stop clock is right twice a day kind of thing. Somebody had to have said this at some point. These were the folks who did. In 2025, NVIDIA seemed to have given the US government an incomplete account of assets Megaspeed had purchased, undercounted by about 50 ,000. That's not going to look good. And at least four law firms have begun their investigations into whether NVIDIA misled investors. We're definitely going get some of those investor lawsuits. I've got birds circling in my head right now. Yeah. As they do it. We end every episode of DTNS with some shared perspectives today. A. Burb thinks new rules restricting data centers might actually force more models to work locally.

40:52Yeah, A. Burb wrote this before our discussion with Opera yesterday, but wrote something that I think applies to that. I genuinely believe AI must and will find a way to make even its frontier models run on local devices. Laws like these, referring to the data center restrictions, will force an efficiency component into this research. And I'd argue that's a good thing. Yeah, that would be a good thing. So is Abram saying that, like, Frontier capable, like, as capable as what's in the cloud would then be local? Because, man, even with efficiency. That seems to be what he's saying, is like, they'll make even Frontier models run on local devices.

41:32The problem is a local device is never going to be as powerful as a data center device. So if you can make your frontier model run on a local device, it would be even more powerful in the data center. Right. Just by definition. Right. Right. Because they're always going to have massively more compute than you're going to have on your desktop. Even if you have a really powerful machine, they're always going to be ahead as far as that's concerned. Yeah. Unless they're not allowed to do it anymore, which I don't think is going to happen. And A-Burb may be like, okay, fine, I wasn't speaking precisely.

42:06But I do think he's right in general that we are going to get more efficient in what we can run on a local device. Yeah, I absolutely agree with that too. What are you thinking about? If you have any insight into a story that we've talked about or that we haven't, you can share it with us. Feedback at DailyTechNewsShow.com. Yes, thanks to A-Burb for contributing to today's show. Thank you for being along for Daily Tech News Show. You keep us in business, folks. Become a patron, patreon.com slash DTNS. Reminder, no episode tomorrow due to the U.S. holiday, but we'll be back on Thursday. Talk to you then.

42:46The DTNS family of podcasts, helping each other understand. Diamond Club hopes you have enjoyed this program.

42:58This episode is brought to you by Okta, helping you secure and govern AI identity by bringing agents out of the shadows. AI agents are changing how work gets done. They're fast, they're capable, and more of them are showing up inside companies right now. That's exciting. But when agents operate unseen in the dark, they create massive security blind spots. Okta shines a light on all of your agents so you can stay secure. Whether your teams are building agents or bringing them in, Okta gives you a clear way to answer four simple questions. Where are my agents? What can they do? What are they doing?

43:33And how do I respond if they go rogue? With that visibility, you can govern each agent's identity, trace their behavior, and even hit an instant kill switch. Bring AI agents out of the shadows. Okta secures AI. Learn more at Okta.com. That's Okta.com. This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value.

44:17Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Thank you.