Vibe Code Ads Right Off of Your TV - DTNS 5377

0:00With the Spark Cash Plus card from Capital One, you earn unlimited 2 % cash back on every purchase. And you get big purchasing power, so your business can spend more and earn more. Stephen, Brandon, and Bruno, the business owners of SandCloud, reinvested their 2 % cash back to help build the company's retail presence. Capital One, what's in your wallet? Find out more at CapitalOne.com slash Spark Cash Plus. Terms apply. 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.

0:38Its formula with science-driven ingredients include theanine and huperzine A to support focus and memory. And it's caffeine-free. So when you're deep in the work, Alpha Brain helps support enhanced productivity and tune-out distractions. Get focused on what matters. Shop Alpha Brain now at onnit.com. Therapy's biggest impact doesn't show up in some dramatic breakthrough. It shows up in the smallest moments. You bite your tongue instead of escalating. You put your phone down instead of firing off that text. You say what you actually mean instead of what would hurt someone else the most. Nobody applauds those moments, but they quietly change your life.

1:16That's what growing feels like. And that's exactly what Grow Therapy is here for. Grow is therapy that works, covered by insurance. Sessions average$21 with insurance and some pay as little as$0 depending on their plan. There are no subscriptions, no long-term commitments. You just pay per session. Grow connects you with thousands of high-quality licensed therapists across the U.S. offering both virtual and in-person sessions nights and weekends. You can search by what matters, like insurance, specialty, identity, or availability. And you can book a therapy session in minutes and begin in as little as two days, not the national 48-day average.

1:56See if you could pay as little as$0 for therapy in seconds at growtherapy.com slash ACAST. That's growtherapy.com slash ACAST. growtherapy.com slash ACAST. Availability and coverage vary by state and insurance plan. 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. Too 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.

2:36Call 1-800-GRAINGER, click grainger.com, or just stop by. Grainger, for the ones who get it done.

2:56this old thing this is the daily tech news for tuesday october 20th 2026 happy birthday to my brother tim we tell you what you need to know give you important context and help each other understand tim you better be listening today i'm just saying uh today should you vibe code the ads and bloatware off of your smart TV. We're going to talk about that. Maybe, maybe not. I'm Tom Merritt. I'm Jason Howell. Let's start with what you need to know with that big story. So two open tools posted on GitHub recently have gotten a lot of attention for promising to clean up your smart TV and how they were made.

3:33Dumb TV is the name of one of them. It says it'll get rid of ads, tracking, and forced updates from the TV's operating system if you don't want those. It's been tested on the Fire TV and the Android and Google TV platforms. Developer Muzamal Yasin Sabawala suggests telling an agent to implement the app. He's just like, just tell your agent, put dumb TV on your TV. He used a combination of DeepSeek version 4.1 Flash, Claude Opus 5.5, and Claude Sonnet 5.5 to make it. And he hopes to add support for more models. So it's an example of somebody saying, hey i've got a problem let me let me vibe code up a solution to it and then share it with folks clean your tv promises to remove your bloatware from both android tv and google tv devices over wi-fi and then install a launcher that is modeled along the apple tv design its developer gavin t posted on x vibe coded my parents tv this afternoon it was showing a sponsored ad before you could even get to netflix gave claude code the model number it found the tv on wi-fi removed 25 xiaomi packages and wrote a new launcher looks like an apple tv now lol so he did even less planning it sounded like he was just like hey can you fix this and claude code was like yeah sure here you go i found the tv uh he posted the prompt on github so that others could tweak it to suit their own devices his solution does not need root access does not prevent restoring to factory defaults something that might be helpful if you have a warranty repair and then ryan doyle posted on x that he just asked claude code can you hack this and posted a listing of his tv model and it walked him through removing ads and bloatware while also speeding up the user interface he did not share his particular uh code with anyone else but he was mostly saying like you can just ask the tool It'll do it.

5:30Now, not everybody has had success with this. Sometimes trying to avoid hacking the TV and requiring complex prompts will show up. Claude will say like, oh, I'm not supposed to do hacky things. And so you might have to make a more complex prompt to get around Claude's guidelines. more than a few people have bricked their device along the way there's all kinds of ways that could have happened uh it could be the code it could be user error we don't really know uh tech crunch points out there are existing apps already for android-based smart tvs if you don't feel comfortable vibe coding you might not realize that there are home screen replacement apps like Like ATV Launcher, Projectivy Launcher, F Launcher.

6:15Jason, I don't know. Sounds like we got another example of people saying, well, I don't need to find an app. I can just make an app to do the thing I want. Yeah, and we should invite a host of Court Killers to join us for this segment to talk about it. It's Mr. Tom Merritt. Welcome, everybody. Tom Merritt. Oh, hi. Thanks for having me, Jason. I just wanted to bring you in the conversation. I'm a co-host of Chord Killers with Brian Brushwood. We talk about this sort of thing all the time. In fact, we'll be talking about it on this week's episode. I'm sure you will. When I think about this, it's good to have you, Tom, Eric, have you here.

6:51When I think of this, a couple of things come to mind. First of all, TV interfaces, smart TV interfaces, really kind of are a double-edged sword right now. Because I feel like for many years before we got sick of how bloated they got, it was at least for a while, it was kind of like, well, can my TV actually make the experience better when I'm splintered among 20, 30, 40 different services and my content's everywhere so that I don't have to go into each one of these? Like, can it bring it into the interface? And so that ended up being kind of like the design and development, you know, worked towards that.

7:25But these teams always have to keep iterating and adding new things, and then it's all ads, and then there's some sort of sponsor interest in directing you to this Netflix show versus that or sports or whatever. And so now they've kind of gotten to a point to where people are like, I want my TV dumb again. I don't want it to be smart anymore. So I'm not surprised. This is basically the end-ishification of TV interfaces is where we are right now. And now we've got these systems, these AI systems, that whether you know what you're doing in code or not, it's very easy to just tell a system like this, hey, this thing's doing this thing.

8:03I don't like how it does it. What are my options? And it might give you a great option. And in this case, for a couple of people, it actually executed them, which is really fascinating. I do note that most of the stories are, I was using a relative's TV and they were having a problem, so I fixed it for them. I still think you need to have a bit of tech savvy to pull all of this kind of thing off. No, you don't have to be an expert in coding. Maybe you don't even have to know a lot of coding at all, but you should probably be a little familiar with it. Maybe have some actual coding friends you can call on if you get yourself into a corner and brick something.

8:42uh, it, it is not without its dangers, but that's not me discouraging you. That's me saying like, if you know what you're doing, yeah, you could, you could probably do this. The other thing I, that it makes me, uh, think about is I saw some rhetoric around the idea of like, I finally want to get the TV I paid for. The sad fact is you didn't pay for that TV. That's TV was subsidized by all the ads. It hoped it would sell you go look at the cost of an equivalent monitor that has the same specs and none of the software, it's going to be a lot more to buy that technology without the subsidy. And again, that's not me discouraging you from doing it.

9:19It's just like, I don't know that I'm going to feel super righteous about doing it because I got a subsidized price and I'm getting around it, but that's no different than using an Apple TV, right? Like with my LG, I rarely ever see the LG interface in its ads because I'm always using an Apple TV or a Roku or a Shield or something like that. So I think it's fine to do this. But do be aware that it's not entirely risk-free, but it would pay off if you take the risk. Yeah, it's certainly not bulletproof. A couple of the ways that could potentially get you into trouble, long-term maintenance. If you're making a change like this, that works right now, is it going to work in two months, are they going to flash a firmware update to the TV that wipes it out?

10:05Is there a possibility of bricking, which if you can go back to your factory defaults, then maybe that's less of a case. But for just random person going, hey, I've got a$20 subscription to ChatGPT, I'm going to do this too. Something to consider. And then also one of these was done in a way, which I haven't really seen this very, very often. Like we're used to getting like these, these installer files or an APK or something like that to be like, put this on there and it'll do the thing. Instead, in one of these cases, it was, here's my prompt, run this prompt and it will make the changes. And one thing that I know through working with these systems a lot is that prompts don't always lead to exactly the same place from person to person.

10:48Little details and little things, and you know, it might decide to go left instead of right at this particular point. And you aren't guaranteed to end up with exactly the same thing on the other side that this person was. His prompt worked for him in that regard. It might not work the same for you. Yeah, that's a really good thing to remember. Prompts are not determinative. No, they are not. I think it is yet another example of the beginning of this new, it's not even a new world because we've been living in the world, but the new trend of realizing oh, I could make my own app. I could do this myself.

11:23Or I don't like how this works, so I'm going to change this one thing. And actually, Rich Straffolino and I talked last week, and I think it's coming out this weekend on the DTNS feed. Yes. As far as anyone listening is concerned, Jason will be talking to Rich on Saturday. Oh, yes. I will be. Yes, exactly. Jason's a time traveler, though, so in his mind, he's already done. Yeah, sorry, I've said too much. Basically about using AI to create and flash new firmware onto audio hardware that make it run different, bring it new sounds, you know, completely changes the firmware essentially, and how you can vibe code using these AI tools to create new firmware for your devices.

12:07And I think this story that we're talking about today is firmly in that realm. I think now more and more, like it even has me like looking around the house and being like, is there anything here that I want to experiment with? Like that I'd be okay if I totally bricked it. But like what could I make my toaster do that it couldn't do before? That's probably a horrible thing to do. Don't do that. Don't want to set your house on fire. But I think we're in this weird moment where people can kind of use their imagination a little bit. And who knows, you might actually be able to use your LLM to create alternative firmware for the thing that you use.

12:39Yeah. And again, if you're comfortable doing that, it is not a dead simple thing. But if you're comfortable doing it, yeah, you might be – right now I'm thinking about my laundry machine. I'm thinking about my washer and dryer and how the firmware just never worked with my networking and I gave up on it. so I don't get notifications from my washer and dryer, and I'm fine living without notifications from my washer and dryer. But I'm sitting here thinking like, ooh, do I want to risk messing with my washer and dryer's firmware in order to make that work the way I want it to work instead of the way LG wanted it to work?

13:15But I don't want to screw up my washer and dryer, so I don't know if I'm going to do that. Totally. Yeah. Well, if you do, bring it on the show. Oh, I will. You know I will. All right. DTNS is made possible by you, the listener. Big thanks to Vinnie Yack Menon, ThatCharlieDude, Justin Zellers, Chris Beneteau. And a big welcome to Jason A. I keep saying, if you become a new patron, you're going to hear us thank you and clap and stuff. So become a patron like Jason did at patreon.com slash DTNS. Episode prep gets messy when research, slack threads, and script trackers are all in different places.

13:55That's where ChatGPT work comes in. Give it the goal and connect to the tools where work happens, like apps and files, and it can help pull the pieces into a draft you can review without you having to re-explain the project at every step. For example, ChatGPT work can gather team feedback from Slack and use it to suggest updates to a project brief in Google Docs. You stay in control reviewing the draft and deciding what moves forward. Put ChatGPT to work on your most ambitious ideas and projects. Get started at chatgpt.com by selecting work mode. Available on Plus and Pro plans.

14:54more at CapitalOne.com slash SparkCash Plus. Terms apply. Some days, getting the work done isn't the problem. Staying locked in is. AlphaBrain 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 huperzine 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. Get focused on what matters. Shop Alphabrain now at onnit.com. Do you snore? Did somebody hand you a CPAP and did you quietly stop using it?

15:36The mask that never quite seals? The hose tugging at you every time you roll over? A machine sitting abandoned in a closet? Excite OSA works a different way. It treats a root cause instead of managing the noise, and you use it while you're awake. 20 minutes a day, while the coffee brews, while you set your fantasy lineup. An FDA-cleared prescription device for primary snoring and mild obstructive sleep apnea. Then you go to bed like a person, nothing strapped on, nothing plugged in. In clinical studies, 90 % of patients saw a reduction in snoring. No sleep study required. Prescribed online, shipped to your door.

16:20Excite OSA at springsleep.com slash sleep20. Code sleep20 for 20 % off. Excite OSA is an FDA-cleared, prescription-only device for primary snoring and mild obstructive sleep apnea. Not for everyone. Individual results vary. Risks and benefits at springsleep.com. Harvard Business School Executive Education delivers a world-class learning experience that energizes aspiring and established changemakers. Prepare for the next elevation for your organization and for yourself. Learn more at hbs.me slash breakthrough. That's hbs.me slash breakthrough.

17:07Oh, there's so much more today. Let's get to the briefs, briefuses. The robotic Martian moons exploration mission. This is known as MMX, lifted off from Japan's Tenegashima Space Center on Monday, beginning a 10-month journey to the Martian moon Phobos. After reconnaissance in late 2028 or early 2029, a rover called IDAFIX will land on Phobos and spend 100 days exploring. Then in late 2029, the main spacecraft will attempt up to two landings itself and spend several hours collecting samples. A coring drill will collect a cigar-shaped sample, and a pneumatic sampler will use nitrogen gas to sweep bits into a collection canister.

17:55It'll then lift off in 2030 and head back to Earth with an expected landing in South Australia in 2031. Japan's Hayabusa and Hayabusa 2 missions were the first to bring back samples from asteroids. Fascinating. Japan's become the specialty digging on other astral bodies space agency, which is pretty cool. The Hayabusa missions were fascinating. I forgot there was a second one. I paid a lot of attention to the first one. That's kind of the way it always goes, right? Yeah, right. but Phobos is thought to be possibly a captured asteroid, but we don't know because we're so far away. We can't tell for sure.

18:37So that's one of the things that MMX is going to do is land on Phobos, take a sample so that we can try to figure out, hey, is this a captured asteroid or is it ejecta from Mars, from some kind of collision with Mars, or is it a combination? Did an asteroid bump into Mars and this is what's left of it plus pieces? Anyway, we're going to find out if this mission is a success. Russia tried one in 2011 that failed. The Soviet Union before that tried one in the, I think, late 80s that also failed. So third time's a charm is what I'm saying. Yeah, I just, in reading through this, have a lot of admiration for the patience that you have to have when you work on projects like this.

19:19Like you're talking about three, four years down the road. and you're like, if our calculations are correct, we're going to be back in 2030 and have everything that we hope to have. So I have a lot of admiration for that kind of patience, because I sure don't have that kind of patience myself. Well, you're going to have to have a lot of patience if you want to read the original post regarding this next story. One of the challenges with large language models is you cannot easily see how they come up with their actions. There are workarounds, like having the LLM report its thoughts. But the actual underpinnings are often called a black box because when you look inside, quote unquote, the software, most researchers agree it is a tangle of a neural network, right?

20:05They talk about it as a neural networks with neurons, but there's a tangle of how they work. It's not obvious how the network actually works to make its answers. So when agents go rogue, it becomes detective work that involves looking at logs and messages to try to deduce why, rather than being able to look at the actual operations of the software. However, that might, and I emphasize might, be changing. Nick Camarata, who is a researcher at the AI startup Goodfire, posted an article called Neural Networks Are Already Clean. And I'm going to come clean and say I have not spent the entire 65 minutes reading all 26 ,000 words, but I have read through his summaries and scanned through it.

20:48and MIT Technology Review has a really good interview with him that kind of sheds some light on it. He thinks he has discovered a way to understand LLMs, which would make them easier to monitor and easier to control. Now, reception of his post has been mixed. Nobody hates it, but some researchers are very enthusiastic, thinking he nailed it and it might become one of the most important points in the development of LLMs. Plenty of others, though, are looking at it and saying, yeah, no, these observations are good, but I don't know that they make the problem significantly easier to solve. Kamarada himself told MIT Technology Review, he was kind of surprised by the enthusiastic responses.

21:28He didn't think it would cause people to update their beliefs as much as it did in some cases. But Kamarada has been working on this for a long time. He worked at OpenAI, developed a method called mechanistic interpretability. he helped develop that that method mechanistic interpretability examines those tangled paths inside a model to look at the individual neurons to see what they represent and try to therefore understand how the model works a little better well if this neuron is always firing when it sees a cat we know that's the cat neuron and if we can connect it to other neurons maybe we can understand what's going on current models are too large and complex to make sense of more than a small piece of what's going on.

22:09So this tangle has gotten in the way of understanding. Camarada now says, we were looking at the patterns wrong. He says, if you don't assume the model is a tangle, you can actually start to map out a kind of circuitry. Don't go into it thinking, okay, this is a tangle I've got to untangle. Just look at what happens. For example, in the 2014 Inception V1 model, they found a single neuron that was used to recognize cats and cars. So the researchers were like, why? What do cats have to do with cars? The grouping of more than one use like that was called polysemanticity. And they assumed this was an example of a superposition, that the model was just reusing the neuron somehow that they didn't understand, that it could be both for cats and for cars somehow.

22:58And that was part of the Tangle way of looking at things. Like, we don't really understand why that works that way. The model using one neuron for more than one thing. With his new perspective, Camerota figured out it's because the face of a cat and the front of a car seen from the side both have a V shape. He then found that that same neuron responded to other V shapes in images, for example, the corner of a pool table, and turned off when it saw rounded shapes. So he calls this recursive generalization. His argument has four steps. networks use one learning principle not two if not everywhere then nearly enough that it doesn't matter for the study so there's no blocker to beginning ordinary science on them like people thought if they're like well they're using this neuron for two things ordinary ways of looking this aren't going to work he says no they're only using it for one thing it's the v-shape in that case with modern models and tools that science goes much better than most people expect understanding compounds.

24:01So as the map fills in, he thinks it gets easier to understand what's going on. And he says at minimum, a network's own mechanisms can be part of the toolkit for understanding. Now he calls his last step highly speculative, but he says this might be most of what we need and interpretability might be far more tractable, possibly in the near term than people assumed. Kamerata says, I now feel like if I had 100 ,000 years to go look through a frontier language model one circuit at a time, there's like a pretty good shot I would come up with a good understanding. And he points out that he doesn't have 100 ,000 years, but agents can do things that would take a human 100 ,000 years much faster.

24:43So maybe putting an agent on the case might be the solution. Kamerata and his colleagues are now using this approach with current models. They plan to publish technical results soon. There's a lot of let's see what they come out with first. And they want to be able to publish so people can try to replicate their work and all of that. So how much this is actually going to advance understanding, we don't know. Camerata told MIT Technology Review, it's not infinitely good news. It's not everything solved. There's a bunch of open questions around like how much the combination of neurons matter versus just one at a time.

25:17but it has got a lot of people excited because it's a different way of looking at things and it makes them hopeful that it will work yeah well and and there is i think more and more kind of a need to understand exactly what's going on like yeah there is at the end of the day an explanation for why these systems do what they do the way that they do and the black box idea has been you know where we've been because simply from a point of like well we're making guesses We're trying to understand. They're leaving their kind of their trail of decision making and everything. That's a very reactive way to analyze how these systems work.

25:56That tells you, yes, the twists and turns and decisions that they made along the way, it's some of the picture of what those decisions were in the process, but doesn't really necessarily tell you why those decisions were the way that they were. And this seems to be more of a step closer to understanding the why behind how the system works and how those decisions are made. And I think that's really important, especially when we're talking about things like alignment and safety and everything. Being reactive means that there's a thing that happened that we respond to, that we react to. And if we don't truly understand the systems and how they are designed and how they operate and everything, then we're always kind of reacting and a little bit behind.

26:37And I think everybody would probably agree. We kind of want to be ahead of that, to understand it so that we can be ahead of that curve. I think this is one of the keys to understanding why things are happening the way they're happening. Because a lot of our responses to this are, well, if this is a computer, we know how programs work. Just find out where the program went wrong. And I think what this shines a light on is you couldn't do that the same way with an LLM. and it pushes you towards wanting to anthropomorphize it, right? And say like, well, what we had to do is look at what it did and what it said, like you would with a human.

27:16Like, why did they do this? Let's ask them why they do it. Let's look at what they did. But I can't actually look in the brain and find out how they did it. And what Camerata is saying, like, maybe we can. Maybe we can actually understand how these neurons work, In which case, we can start developing a theory of why things happen the way they happen in a way that you cannot with a human. At least not yet. Yeah, exactly that. And it turns it more into like a scientific style investigation. It makes it a little more deterministic. It makes these tools less scary. Yes, totally. If you know how they work, right?

27:52Absolutely, 100%. 100%. Meta's Muse agent, like many agents, lets you share a conversation with someone else by creating a link. Researchers at security firm Cloric discovered a central design decision in the way Muse creates these links that makes them easy to guess. All Muse links use the same domain name. That is cf.metaaiusercontent.com with a conversation name and string of text as the first subdomain. Like the part that is often www.exceptitshome-sale-budget-93f.cf.metaaiusercontent.com. Because it's in the subdomain, it's stored in a passive DNS resolution database, DOA. DOA search of that database and you get the whole...

28:47Do a search of that database. Sorry. Do a search of that database and you get a whole URL at the end of it. Now, among the things they were able to read from this kind of public presentation, a small business owner's loan application with the amount requested and a workplace's full year staff vacation calendar. It listed dozens of employees against the week. Each one would be away. Meta says it has, quote, paused this functionality while we review this report and consider changes to how these links can be discovered. Tom, are you surprised that when you share something publicly through a system like this, there's a possibility that it might somehow be discoverable by others?

29:31There's always a possibility that any link could be discovered. You make it sufficiently complex so that people can't guess it. Now, your example was intentionally simple, home sale budget 93F. In reality, it'd be something like home sale budget, 93F76, pound symbol, blah, blah, blah, blah, blah, right? It should be hard to guess that. That's not what they did wrong. What they did not wrong was not knowing how DNS works and putting it in the subdomain. Best practice is to put those things after a slash so that it's really, really difficult to discover. And it isn't logged in the DNS database that's searchable by anyone.

30:12This is a really simple one to fix. I feel like you just change your URLs to have a path and not a subdomain. I'm hoping that's all there is to it. I'm not sure why Meta did it this way. Maybe there was some kind of functionality efficiency that they gained. And maybe I'm missing something. You know, when fixes seem this simple, it's usually not because the company's dumb. It's because the company had a good reason for doing it that just didn't outweigh the consequences. I feel like they just didn't know that the DNS user database was searchable like this or didn't think that anybody go looking.

30:49But, hey, we got security researchers, thankfully. So they did. Yeah. So at least there's that. But, you know, they did claim when they were kind of announcing and revealing, you know, Muse and everything along the way. Zuckerberg claimed that it was built from the ground up. Yes. Privacy and security. This is an example of it being built from the ground up. they did a subdomain instead of using a path. So like, yeah, okay. We can all laugh and point and go, you said it was secure and feel schadenfreude about it. This is a perfectly understandable mistake to me. This doesn't immediately cause me to question their ability to secure things, but I'm not sure what's taking them so long to fix it.

31:34Unless I'm missing something. Yeah, I don't know. So I guess I'm a little, I guess I'm a little, when I see a share button inside of an AI tool, my automatic reaction is I assume that by sharing it's publicly available in some way, shape, or form. And that's smart because links can be shared, right? Like if you share a link with me, there's nothing stopping me from sharing it with Hammond and Sarah and Jed, right? Except, you know, the fact that I like you and respect you and don't want to get you in trouble. But yeah, a link is never secure, but this is even extra non-secure. Interesting. Well, folks, hopefully you have checked out our dailytechnewsshow.com store.

32:18We have a lot of cool stuff in there. One of the things I'm proudest of is our DT &S pint glass. R.W. Nash has made many videos on his YouTube channel where he shows off the DT &S pint glass filled with beverages of his choice just to prove that it's an excellent pint glass. If you would like to have a Daily Tech News Show pint glass with the Len Peralta rainbow logo on it, you can just head to our store. They're right there available for shipping around the world. DailyTechNewsShow.com slash store.

33:00Get big purchasing power so your business can spend more and earn more. Stephen, Brandon, and Bruno, the business owners of SandCloud, reinvested their 2 % cash back to help build the company's retail presence. Capital One, what's in your wallet? Find out more at CapitalOne.com slash SparkCash Plus. Terms apply. Losing weight can feel like a roller coaster, full of ups, downs, and being thrown for loops, like when someone brings donuts into the office, or your kid's dino nuggets start giving you that look again. That's why there's Noom. They combine GLP-1s to quiet the food noise with healthy habits that help you keep it quiet, even if you get off the meds.

33:39Their program is designed to make weight loss a smooth ride, helping you understand what drives you, so you can build better habits to not just lose weight, but keep it off. Noom. Lose for keeps. Get started at Noom.com. Not all customers will clinically qualify for medications. Individual results may vary.

34:00Let's get to the quick headlines, because knowing this stuff makes you a little smarter. Yeah, right up here. And Gravity announced a new Follow Me bundle for its A1 drone with propeller guards, a watch remote, and a ring remote. The bundle adds accurate tracking, and you can use the two remotes together for gesture control. You can get the accessories for$340 or with the A1 drone for$799. Yeah, this is a nice little upgrade. Apple has begun limiting the number of iPhone duos you can buy per customer to two. That's a duo of duos. It's not certain if this limit was always in place or if it was added later.

34:40Some people have claimed they ordered more than two. If it was added later, we don't know if those earlier orders for more than two duos have been reduced. There have been some people reporting that scalpers who ordered a bunch of these have now found that their orders have been limited. But one thing we do know is if you try to order a duo or two right now, they will not ship until January 22nd. If you try to order a duo of duos, then that will happen. Then they will arrive on a day with a duo of twos. Of dues. No. The UK's competition in markets. The do's not a number. Do's not a number. However, the UK's Competition and Markets Authority says it wants Apple to lower the fees it charges companies to access the NFC chip on iPhones and open up its wallet to the main contactless standards used internationally.

35:36The EU forced Apple to make similar concessions in that market in 2024. AWS released a tool called Torch Neuron on Monday that lets startups move their apps to the Amazon Tranium chip easily. Torch Neuron works with PyTorch. So if you know, you know, but that lets app makers move from an NVIDIA chip to an Amazon chip if they want. I don't know why they want, but if they want, they can. And Amazon just made it easier and NVIDIA can't be happy. There you go. There you go. DeepSeq has opened up a tool as well. They're like, hey, we got a tool too. It's called DeepInfer that makes its DeepSeq model run faster and use resources more efficiently.

36:14Mm-hmm. However, the U.S. government ordered federal civilian agencies to stop purchasing services and products developed by DeepSeq and remove any existing ones from government IT systems. That includes third-party systems built using the DeepSeq model. Hmm. So one step forward, one step back. OpenAI announced the GPT-6 Astra is capable of migrating software in older languages like COBOL, FORTRAN, and Visual Basic. OpenAI developed a benchmark called MultilangBench to evaluate its progress and found that GPT-6 Astra scores 87.4 % across 20 languages in MultilangBench, up from 26.8 % in GPT-5.6 salt.

36:57I know we have COBOL programmers in the audience. You tell me if 87.4 % sounds good enough. It actually got up to above 90 in COBOL in one benchmark, but I'm just curious. Spotify is joining Apple. YouTube, and Patreon and offering paid memberships for podcasts. Spotify's membership platform will let a podcaster offer perks like bonus audio, bonus video, fewer ads, not ad-free, and early access to shows, among a few others. Interesting. The U.S. Army is developing a program codenamed Echo Tribble to deploy agents that can access network infrastructure and neutralize digital targets. Yeah, Army agents in your network coming for your servers.

37:44Thanks to Motang for noting this one on the subreddit. TP-Link has signed a letter of intent with a U.S.-based Keytronic to begin assembling routers in the U.S. state of Mississippi by early 2027. That's aggressive. TP-Link needs that to be aggressive because it wants to get off the import control list in the United States. And having them assembled in the U.S. would be one way of doing that. That would be one way. All right. We end every episode of DTNS with some shared perspectives. And today, Jackson explains why he would want to own a self-driving car. Yeah, we had a good conversation about this on Friday's show.

38:19Jackson writes, loved the discussion last week. The autonomous vehicle ownership debate is something I just spoke to my spouse about. We both work remote for tech companies, about halfway between Denver and Boulder, 20 to 25 minute drive from either with minimal traffic or 40 to 60 minutes during rush hours. In the last couple of years, we've been regularly recruited by companies that have in office requirements three to five days per week. But the pay hasn't been competitive enough to justify the commute cost and time or giving up the flexibility of remote work. This weekend, my wife had such a recruiter reach out and for fun, we discussed what it would take to make that offer compelling.

39:00As we are in the market for a new car, likely an electric vehicle, we considered whether self-driving could ease the burden enough to justify commuting again. The answer, no. We agreed that if a company was willing to cover daily transportation expenses to and from the office, it might be worth considering. In the future, when autonomous transportation is widely available, it would be convenient to have a car show up daily to take you to work at 815, then waiting at the office at 1700 for your trip home. You could use the commute time to catch up on slack, get ready for the day, wind down on your way home, etc.

39:37I work with a well-known AV provider to support their vehicle routing algorithm development and asked this week about their timeline for expanding service to first and second ring suburbs. Unfortunately, we're likely more than three years out from being in their service area. So for now, that remains a pipe dream. But keep up the good work, Jackson. Yeah. I mean, hey, if that time in the vehicle, if it's a long commute, like I think about in the Bay Area, right? There's people that live all over the Bay Area. If they work at Google, you can drive to San Francisco and then get on a Google bus, and that'll take you down to Mountain View.

40:11It's like an hour drive or whatever. While you're on the bus, you have the ability to work. And so technically, that could mean that you cut down on the amount of time you have to spend in the office because it becomes work time. And I imagine that would be a huge benefit to having self-driving vehicles cart everybody around. I think that's a good point that I think we did see. That's a great way to look at it. Instead of 8.15 to 5, they should pick you up at 9 because that's when you're working. Right. And then they should pick you up to go home at 4, and then you're working on the way home. Love it.

40:41If it's not in the office, maybe that's the problem. But yeah, the idea is the autonomous car gives you that same advantage that you would have on the bus or the train, except a little more privacy. So you can play things out loud and not be upsetting the other passengers and stuff like that. Or you could choose to watch Netflix. That too. On the way home, especially if it does pick you up at five, you shouldn't be working anymore. You should just be relaxing and winding down, like you said. Anyway, good stuff, Jackson. Thanks for sending that. Love it. What are you all thinking about? You can share your insight with us into a story or any other discussion that we've had.

41:14Share it with us at feedback at dailytechnewsshow.com. Yeah. Thank you, Jackson, for contributing to today's show. Thank you all out there for being here for Daily Tech News Show. You could be the next Jason A. You could be named on the show tomorrow. Just support us right now. Patreon.com slash DTNS. The DTNS family of podcasts. Helping each other understand. Diamond Club hopes you have enjoyed this program.

41:50This 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. This is the story of the one. As a shipping manager, he's running the dock, stocking supplies, and making sure every order goes out on time.

42:28And Grainger helps him stay prepared. From shipping supplies to PPE to hand tools, he counts on Grainger to keep products moving and trucks rolling. So his operation runs smoothly, no matter the shift. Call 1-800-GRAINGER, click grainger.com, or just stop by Grainger. For the ones who get it done.