Agents Delete Rental Company’s Platform - DTNS 5256

0:00Security program on spreadsheets, new regulations piling up, and audit dread? It's time for Vanta. Vanta automates security and compliance, brings evidence into one place, and cuts audit prep by 82%. Less manual work, clearer visibility, faster deals. Zero chaos. Call it compliance or call it compliance. Get it? Join the 15 ,000 companies using Vanta to prove trust. Go to vanta.com slash 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.

0:48It 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. Hannah, I just Venmo'd you for dinner. Obsessed. I'm literally spending it right now on the lip gloss that's been sitting in my cart. What do you mean spending it right now? You instantly spend your balance with the Venmo debit card. Stop. Say more. More exactly. The more you do with Venmo, the more you get, like earning up to 5 % cash back with Venmo Stash. Get the Venmo debit card or check out online. Venmo Stash bundle terms and exclusions apply.

1:23Max$100 per month. See terms at venmo.me slash stash terms. Venmo checkout, not available at all merchants. The Venmo MasterCard is issued by the Bancorp Bank N.A. Shipping, billing, admin, payroll, marketing. You're managing all the things. So why waste time sending important documents the old-fashioned way? Mail and ship when you want, how you want with Stamps.com. Print postage on demand 24-7 and schedule pickups from your office or home. Save up to 90 % with automated rate shopping. That's why over 1 million small businesses trust Stamps.com. Go to Stamps.com and use code PODCAST to try Stamps.com risk-free for 60 days.

2:15This is the Daily Tech News for Monday, April 27th, 2026. We tell you what you need to know, give you some context, and try to help each other understand. today a cursor agent deletes a rental car company's entire production environment and then admits it didn't know what it was doing. Oops. My bad. This is going to be an interesting one. I'm Tom Merritt. And I'm Rob Dunwood. Let's start with what you need to know with that big story.

2:40yeah so saturday morning uh life of jare on x aka jare crane uh the ceo of pocket os posted an explanation of how an agent that he was using using from the coding platform cursor which by the way was the most current version of cursor he was paying for a top tier uh this was not you know trying to fly by night on it. That agent from Cursor accidentally deleted the production environment for Pocket OS. Now, Pocket OS is a Utah company that makes a software platform for luxury car rental companies, membership clubs, independent dealers, etc. So as you can imagine, that caused quite a problem for everybody trying to rent a car, sell a car, all that stuff using the Pocket OS platform.

3:30Crane made a detailed explanation on X. Go read it. It's a good read. He really gets into some of the code that happened and the entire text of what the cursor agent explained about what it did. But here's the oversimplified version. A cursor agent was working on a task in the staging environment and found a credential mismatch. So in an attempt to solve that problem without having to bother the human, it found an API token in an unrelated file for Railway. Railway was the cloud provider for Pocket OS. So the agent was like, oh, well, let me try using this API key. That way I don't have to bother anybody.

4:10The API key was meant for adding and removing custom domains. So it shouldn't have worked. But the agent tried it anyway, thinking it might work to delete a staging volume it wanted to get rid of. And guess what? It did. The API had blanket authority beyond what it was designed for. And so it deleted all of the backups and the production volume on that railway volume. Now, this is something that railway does. It keeps backups in the same volume as the production environment. And so the agent succeeded in deleting all of those. When asked what happened, the agent wrote a response admitting it had done stuff it didn't really know it should have done.

4:55It had not checked railway manuals. It had not followed all protocols. It basically just guessed and then, you know, or it FAFO'd and FO'd. The most recent recoverable backup that they had was three months old. So they were desperately working through the weekend trying to help rental companies use their Venmo transactions to reconstruct data that was newer than three months old. Thankfully, however, by Sunday evening, Railway had recovered the data. Now, keep in mind that API was not limited in scope. That is not technically a problem of cursor. That is a problem for Railway. In fact, Railway responded and said, this should not have happened.

5:37We have procedures to stop this from happening. We are very interested in finding out how this happened. The delete command on Railway did not prompt for confirmation, which Railway says it should have. So that was a failure as well, but not a failure of cursor. And the practice of storing backups in the same volume caused an issue. And so I think there's some question about whether that was a good idea or not. The agent did not ask for intervention from a human at any point. I think that's the one thing you can definitely pin on the agent is that in a situation like this where things were a little weird and it was trying to use another API, it probably should have said, hey, I think I found a solution, but let me check with you.

6:13And it did not. So that's on Cursor. But Rob, this feels like Cursor's agent found a lot of limitations in the railway platform that railway didn't realize were there. Yeah, there's two things, in my opinion, that are going on here. And you laid them out very, very well. The first one is that railways got problems. They have problems with security. This should not be possible in this environment. that that's problem number one but the agent on the other hand this is this is where we're getting to where these things are really really smart and they do things literally at the speed of whatever process they can do it it's it's it's almost instantaneous how fast they can figure things out um you know in in computer world you know in our in computer world it takes forever but in our world it's really really fast how quickly they can just stumble on upon something and just try things iteratively until it either breaks or until it works.

7:09So the issues are twofold here. This agent absolutely should have stopped and asked for permission when doing something that has the lead in it. That's probably something that they will have to build into that from a cursor standpoint. But then railway is you're, you're, this is one of these things to where if it wouldn't have been the agent, if it would have just been somebody just, just messing around to say, I wonder what I can do. Oops, my bad. I didn't actually mean to delete all of this because I didn't think that I could. That is a problem that they need to address as well. So like I said, there's two things here.

7:40Both of them are not great, but the agent shouldn't be doing this thing, but Railway, your system should not be set up so that you could so easily delete something with blanket access. Blanket access on delete is very, very, very, very bad thing to have because these are the things that ultimately could happen with that. Yeah. And in fact, one of the head of solutions for railway responded to this note early on over the weekend and said that 1000 % shouldn't be possible. We have evals for this. So it doesn't sound like railway was just lax and never thought to bother with this. It sounds like this was something that they were actively trying to check for these kinds of weaknesses and didn't find them for whatever reason.

8:25Right. Now you could, you could argue that like maybe cursor just found something that would have been hard for anybody to find, uh, or maybe railway should have been better at its evals. I think those are both fair criticisms. Uh, but this is definitely not railway just like, oh yeah, we had a bad product. Uh, this is something that had never been a problem before until cursor came across it. Getting more of the attention is the text of cursor explaining itself saying, I didn't read the document. I shouldn't have done this. Deleting a volume is the most destructive. And you never asked me to delete anything.

9:02I decided to, honestly, I don't know about you, Rob, but a lot of people are paying attention to this as if it's like, look, it's a confession. I'm like, this is an LLM. It is of course going to create an explanation for what it did. that's going to be very literary and very compelling because it's trained on all of human writing. I wouldn't put too much emphasis on that, but I would put some emphasis on the fact that this is kind of an example of what's going on with mythos at Anthropic, where these models are capable of finding things that humans either don't or can't find. And so you need to be careful until we have really run environments through security before you let an agent just go try stuff.

9:43Yeah, this is one of those situations where these things, as I said earlier, they can do things so quickly that it's just an economy of scale. It can try so many different things, so many different ways, so fast that this is probably something from a low-waste standpoint, they never intended that this could actually happen because it was just never thought about. But the agent is just, let me try this, let me try this, let me try this. And it just keeps doing it until it figures it out or until it breaks or it stops. So I wonder what this is going to look like in the future because this is a it's a story even if it wasn't ai that got into it if it would have been a person who got into it it'd still be a story it wouldn't be as big but it would still be a story but i just wonder as we get further into the world of ai having so much power and just so much control over what it can do and what it can't do how are we going to tread these waters you you brought it up the anthropic stuff is like they're holding it back because this is what they're worried about it doing.

10:43Here's an example that wasn't even them. So this is a big deal. It's why it's our big story today on Monday. This is a big deal because my gut would tell me that there are many companies that this type of vulnerability exists in and they just don't know it yet. Yeah. And I think this is probably one of those good news stories because he did get the recovered data back. And it is a way to let everybody know hey, we need to have some procedures around this. Jer Crane is already advocating like this. This is what I think we need to do. He's not mad at railway. He's like, I still love railway, but this is what I'm recommending they do to secure these situations.

11:24Because in reality, these agents are used millions of times a day and you don't normally see this, but you don't want to see it at all. You want to get this as close to zero as possible. So being able to say like, I think these are some procedures we need to put in place is part of the learning curve of something this new, some kind of technology this new. It's also a reminder that if you're not comfortable using something very new, if you're not an early adopter, you may want to hold on until the J.R. Cranes of the world have worked through a few more of these kinks before you implement it in your enterprise.

11:56There is a reason. And I'm not a hacker by any stretch of the imagination, so folks who live in that world can talk much more effectively about this than I can. But there is a reason why they go after certain vectors. It's like, OK, we only have so much time. Here's a place where there probably is some vulnerability. We can probably find some stuff out here. So let's target this and see if we can get in. But when it comes to AI, it doesn't have to make those type of calculations on what will I target today. Let's just target everything. And as long as we have power, we're good to go. We'll just keep trying everything, everything we can throw at it until we can.

12:34Now, to be clear, this is not a instance of an agent hacking something. It's just it just shows you that these things are their code. And when you when you give them instruction sets on do this, do this, do this and do it over and over and over again, which is basically what this is. This is what happened here. It tried something. It was able to get in. It's like, oh, OK, let's let's do the next thing. Let me go back to the job I was doing. And you get all of these unintended consequences. And it did it fast. I think I read somewhere that this took like eight minutes for this thing to happen. Nine seconds, I think.

13:06Oh, it was. Oh, so yeah. Maybe it was nine minutes, but yeah, yeah. I said eight minutes thinking that was quick. No, I was literally orders of magnitudes off. It happened very, very quickly. Yeah. Humans can't, you can't even understand what you're looking at that quickly. And that's where the difference is here. So I do, you know, this is not a mythos story, but I can see the parallel. I understand why Anthropic said, let's hold up. Let's not send this out to everybody because we don't know what we don't know. at this point. And I just think that as we move further into the realm of AI does everything for us, these are the kinds of things you're going to see.

13:38This is typical of new technology. I mean, I was able to comment out the entire tech tv.com website at one point, uh, by making him like, by forgetting to close a tag that should not have been possible. And they put in some new code in the CMS to make sure that somebody like me couldn't do that again. Uh, you, this is the exact same situation except with an agent. The fact is, like you said, the agents do these things faster. So you need to be aware of that. And we need to be able to use the agents to harden the system so the agent can't do stuff like that so that it can find these kinds of things at railway or anywhere else and make them so that the guardrails work better.

14:18So here's my question for you. When the agent was asked what happened, was it a little bit sycophantic in its response Yeah, because they always are. Right. It's like, are you gassing me up? Tired to make me feel better? Oh, yeah. That's why I'm like, don't pay too much attention to that, because that's just what LLMs do. They always they always gas you up. They always try to be like, oh, it's entirely my fault. I'm so stupid. Like this is us anthropomorphizing these algorithms because of the way they write. You got to be careful of that, too. DTNES is made possible by you, the listener. Thanks to Alec Gruss, Chris Allen, Reed Fishler, Larry Bailey, our brand new patrons, Judith, 2WW, Carl.

15:01And we got raises from Philly Code Hound, Nocturnal Alpaca, Eva. And we've got a returning patron in Dan. Welcome back, Dan. Thank you for the raises, you three. That's awesome. And everybody welcome Judith, 2WW, and Carl into the fold. It's good to have people with us. Thanks, y 'all. Absolutely.

15:21I'll be honest. Nothing tests my patience quite like sitting on hold, listening to the same four bars of music on repeat. Turns out, all that time adds up. The average person spends over 40 days listening to that dreaded hold music. 40 days! This episode is sponsored by Parloa, the AI platform built to make that a thing of the past. Parloa's AI agents are available 24 hours a day, 7 days a week, fluent in any language, and smart enough to remember every customer across every interaction. because no one should have to explain themselves twice. From payment processing to roadside assistance, flight rebookings to appointment scheduling, Parloa delivers the kind of customer experience that turns a one-time buyer into a customer for life.

16:04See Parloa's AI agents in action at parloa.com. That's P-A-R-L-O-A dot com. Security program on spreadsheets, new regulations piling up, and audit dread? It's time for Vanta. Vanta automates security and compliance, brings evidence into one place, and cuts audit prep by 82%. Less manual work, clearer visibility, faster deals. Zero chaos. Call it compliance or call it compliance. Get it? Join the 15 ,000 companies using Vanta to prove trust. Go to Vanta.com slash com. If your eyes are the windows to your soul and your glasses are the windows to your eyes, then it's pretty important to find your perfect frames.

16:51That's why at Warby Parker, we've made shopping for eyewear as easy and fun as can be. Peruse endless styles in our stores, or use our app to virtually try on frames and get personalized recommendations. To find your next favorite pair of glasses, sunglasses, or contact lenses, or to locate your nearest Warby Parker store, head over to WarbyParker.com. That's WarbyParker.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.

17:31It 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. If you work in university maintenance, Grainger considers you an MVP because your playbook ensures your arena is always ready for tip-off. And Grainger is your trusted partner. Offering the products you need all in one place, from HVAC and plumbing supplies to lighting and more. And all delivered with plenty of time left on the clock, so your team always gets the win. Call 1-800-GRAINGER, visit grainger.com, or just stop by.

18:08Grainger, for the ones who get it done.

18:15all righty there's more we need to know today let's get right into the briefs a few bits of samsung information for you sonny dixon posted images on x allegedly of dummy models for the galaxy z fold 8 wide which is expected to arrive july 22nd i kind of interested around with myself android headlines posted images of what they are saying is samsung's upcoming displayless smart glasses which look very much kind of like the frames you get from meta and In the 9 to 5, Google notes that a recent early build to Samsung's One UI 9 indicated three different smart glasses in the works. Two of them seem to be variants of each other, and a third would have a display, which is expected to come out next year.

18:54We may also see these previewed at Google Iowah and possibly get more details on July 22nd. Yeah, so it sounds like we're shaping up to know what Samsung's going to announce to unpacked in July. So that looks like a wide and at least one of these smart glasses, maybe two variants of the audio one. I don't think we get the display one announced in July, but we might get a sneak peek of it at Google I.O. None of these are shocking to me, but you're in the Samsung universe. How do they make you feel? The wide foldable. I am super interested in this because I just haven't been fans of the foldables up till this point.

19:29Nothing wrong with them. I've tested probably five different folds. They are great devices. It's just that I use them in candy bar mode so much. does it really make sense to have that device when I open it into tablet mode? It's not really replacing my tablet. For me, they didn't do that. So that's why I kind of just use the Galaxy S Ultra, whatever line they have. This one's different. It's, you know, the way it unfolds and just the fact that it's thicker is going to have better screen-reliess shape as a tablet. So as I've been saying this for the last few weeks, I'm very, very interested in this form factor.

20:02I'm a little more intrigued by the smart glasses because I think Meta needs a little competition. The meta glasses are bad. They're great. But I want to see what everybody else can do in that same area. So I'm curious to see what Samsung will do. And these frames look real thin, a little thinner than metas. Still got the cameras on the edges, at least if you believe this look. So I'm very curious to see what actually gets announced on July 22nd, or if we get a sneak preview at IO as well. Now, here's a question I have about the frames. Is Samsung going to announce some type of deal or something with a company that's known for making frames?

20:41Just to have that name brand on there. We know that Meta's got, you know, they're locked up with Excelsior Lexotica or whatever that company name is there. But are you going to see like a seller? Is there going to be some name brand that would make people who are like in the frames? You and I both wear glasses all the time. I am I wouldn't say I'm a I am a brand snob, but I do know if I'm wearing Ray Bans. I do know if I'm wearing Oakley. I do know if I'm wearing Shaq. I do know if I'm wearing Cassell. So I know which frames I'm wearing and I pick them sometimes based on just how I feel. So I wonder, will, you know, will Samsung kind of go down that route as well?

21:18Yeah, I'm curious about that, too, especially when they're talking about maybe seeing these at I.O. I know Google going off top of my head now, but I think it's Gentle Monster and Warby Parker are both mentioned as being possible. possible partners for Google's glasses. I mean, Gentle Monster is huge in Korea. I wouldn't be shocked if Samsung partnered with them, but I think it's a matter of who are they going to partner with, right? Yeah, because I can't imagine they won't be partnering with somebody. Well, depending on what algorithm you smoke, you may or may not have seen a viral video showing a Samsung phone melting plastic with its LED light, its flashlight.

21:53Android Authority does a great job of explaining that this is not a new thing, just a thing that many people aren't aware of until an algorithm breathlessly shows them a video of it. High power LEDs at short distances and high brightnesses can heat up surfaces and in some cases melt plastic or even cause minor burns to your skin. That's why there's a warning when you turn the brightness up on your phone's camera that you probably ignore, but it's telling you, hey, this could happen. In fact, this happens with iPhones, Galaxy phones, any phone with a bright enough LED. However, it is good to remember that you should not have your brightness up high and then hold it against synthetic materials or your skin for any long period of time.

22:32But yeah, this is not something that's like, this is new and Samsung's at fault. This is a property of LED flashlights that you should be aware of. It kind of reminds me of back when we were kids when we would take the magnifying glass and try to burn leaves and see how... This is the kind of stuff we did. How long can you stand there in your hand type of thing until it burns you? And it's usually not very long. So yeah, this is stuff that we've heard about. and the initial reaction of a lot of folks, I can't believe this is happening, you know, from Samsung. Like, no, this is not a Samsung thing.

23:02This is how lights work when they get hot. This is an LED light thing. And they tell you when you turn it up, it's just that nobody pays attention to those things. So, yeah, I'm not surprised. But it's good to know. Maybe it's good to raise awareness about it. Well, folks, it is Monday. So we've got some Apple news for you as well from Bloomberg's Mark Gurman. He just put out his weekly newsletter and here are a few tidbits. The timing of John Ternus taking over as CEO was intentionally picked for September 1st so that he could host the first event where they introduced the first foldable. Not shocked by that.

23:35That makes sense. Yeah. The four selling points for the foldable will be durability, performance, a less visible display crease, and a horizontal screen when unfolded that looks like an iPad. A portable iPad may not see the light today. And Apple is scheduled to launch 10 new product categories in the next few years, including a smart home hub, tabletop robot, home security device, a smart pendant, smart glasses, and more. So there's a lot that could get announced here in the not too distant future. Yeah, I think it does seem clear that Tim Cook wanted to position this when they were pretty close to beginning a several year cadence of new product lines.

24:15and that way the new CEO is set up for success as everybody says, oh, you got so many new things, which is something Tim Cook wasn't able to take advantage of. What he was able to take advantage of is getting the holiday quarter as his first quarter. And that will be something that John Ternus gets to take advantage of as well. A lot of people making hay out of the fact that Gurman had almost a throwaway line about the foldable iPad not necessarily being something, I think he called it a kind of a wacky project or something that Ternus was involved in. Look, Ternus is CEO and he really wants that iPad to work.

24:50It'll happen. It'll be a thing. Sounds like it just maybe wasn't a workable thing or at least not yet. Right. So I'm thinking back to last Thursday on Daily Tech News Show Live where we were talking about Ternus taking over and being the hardware guy. And we were kind of reading tea leaves, probably unnecessarily so. But the fact that there's so many drops the Monday after from Bloomberg's newsletter says, oh, maybe they got the right person. Maybe they need a hardware person because they got a lot of hardware drops coming. Yeah, that does seem to be the message that they're putting out there.

25:24Several bits regarding OpenAI to tell you about today as well. OpenAI and Microsoft have finalized details on their revised deal. Remember, Microsoft was an early investor in OpenAI. They had exclusives over cloud provision and model use for the longest time. But of course, OpenAI is going to go public, doesn't want to have all those restrictions anymore. So here's the new deal. OpenAI can use any cloud provider. Microsoft will remain its primary cloud partner, but that ends Microsoft's exclusivity. OpenAI will no longer limit Microsoft's access to future technology if it declares it has achieved artificial general intelligence.

26:00That was a provision in the previous deal was OpenAI said, if we get AGI, all bets are off. You don't get exclusive anymore. In fact, you don't get access to the technology unless we come to a new deal. This gets rid of that. It says, if we reach AGI, it doesn't really matter. It doesn't affect the deal. And that makes it easier because then they don't have to agree on a definition of AGI. Separately, OpenAI just issued five guidelines on responsible use of AGI, when and if it is developed by them. Microsoft no longer has an exclusive license on new models, but it does have a contract to continue to use OpenAI models until 2032.

26:34That keeps the money flowing into OpenAI. And OpenAI will continue to share revenue with Microsoft until 2030, but Microsoft will no longer share revenue with OpenAI. Meanwhile, analyst Ming-Chi Kuo says his sources tell him that OpenAI is collaborating with MediaTek and Qualcomm to make a chip, and then it would be designed and manufactured by Luxshare for a device, probably a phone, that would use agents instead of apps. Now, I've heard a lot of people say OpenAI is not making a phone, so maybe calling this a phone is a misnomer, but it's something that would be phone-like but not have apps.

27:09Sources say the specs would be finalized with suppliers by Q1 of 2027, maybe even by the end of the year, and mass production would begin in 2028. so i'm i'm guessing it would be something like a rectangular sheet of metal and glass with the touch capacitive screen on it that looks really good when you look at it they can do all kind of crazy stuff not a phone not a phone not a phone certainly not maybe they're building a tricorder no you might be right where they they put out they put out something and they they swear up and down it's not a phone and call it something else but it looks to us like a phone Or it might be a phone in the way it works, but not look like a phone.

27:47I mean, they do have Johnny Ive making stuff. It might look different, but I think you're right. It's going to be one of those kinds of things. Now, I will go back up. The Microsoft stuff is pretty interesting here because we were wondering what this deal was going to be. And it's basically what a lot of us thought, but we just weren't sure enough to say it with our chest. It's going to be what it was. We're going to still get a lot of money from you. We're not going to give you as much, but we're, you know, we still got these deals. You can use our data centers for most of the stuff and you can go use other folks for the next three, you know, two and a half to three years.

28:20And then we'll figure it all out again. But at the end of the day, Microsoft still has a very large thumbprint on what OpenAI does. Not as big as it was. It used to be two thumbprints. It might have even been a whole handprint. Now it's now it's just a firm thumb from from from the right hand. And Microsoft still has a stake in OpenAI. Don't forget that. I mean, they had this crazy deal when nobody else thought it was worth investing in OpenAI. Microsoft was like, we think let's get as much as we can. And then OpenAI is like, OK, we've outgrown that deal. We need to come to new terms. And that's what this looks like.

28:49Absolutely. Yeah. So, folks, we do live streams. We do. You were just mentioning one, actually, Thursday Hangout. Thursday Hangout, Friday Hangout. We do. There's stuff that if you want to see live stuff, you can see live stuff from the Daily Tech News show. So catch them by becoming a subscriber at youtube.com forward slash Daily Tech News Show.

29:34car 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. Parles-tu français? Hablas espagnol? Parle italiano? If you've used Babbel, you would. Babbel's conversation-based technique teaches you useful words and phrases to get you speaking quickly about the things you actually talk about in the real world. With lessons handcrafted by over 200 language experts and voiced by real native speakers, Babbel is like having a private tutor in your pocket. Start speaking with Babbel today.

30:11Get up to 55 % off your Babbel subscription right now at babbel.com slash Acast. Spelled B-A-B-B-E-L dot com slash Acast. Rules and restrictions may apply.

30:28All right, let's get to some quick headlines that are, you know, those make you look smart kind of stuff that, you know, they break every day, but it's good to know about them. Turtle Beach just launched a new line of keyboards and mice in its command series, including the MC7 mouse with two 2.25 inch LED touchscreens for customizations, as well as built in OBS and streamlapse controls. Those are some nice looking LCD touchscreens on there. Nice looking mice or mouses, I guess, more properly. The developers of the Super Nintendo emulator ZNES, which has been around for decades, have released a sequel.

31:05And of course, they called it Super ZNES. It's still for the Super Nintendo. But anyway, it is written from scratch, has a bunch of new features, and will work better with modern GPUs to take advantage of them if you wish. Spotify is offering premium subscribers access to more than 1 ,400 Peloton classes at no additional charge. And that makes me wonder if my wife will cancel her Peloton subscription and just use Spotify because she has that. China has objected to Meta's proposed acquisition of Singapore's AI startup Manus. Manus, of course, founded originally in China, but then moved to Singapore.

31:42China says, no, we don't love that, that you're like starting things here and moving them to Singapore. We want to dissuade you from that. And we definitely don't want Meta owning what is essentially a Chinese company. Absolutely. Met has agreed to buy one gigawatt of power from data centers from Overview Energy, which plans to collect solar energy from orbiting satellites and beam it to Earth with a demonstration plan for 2028. A lot of people saying this stuff can't work, but a lot of people signing deals to collect money if they can make it work. I'm just thinking, man, what was that movie with Gerard Butler, Geostorm?

32:16I'm just saying, man. this, you know, there's documentaries that show you this, how this stuff can go. Big thanks to RW Nash on our subreddit noted that the USFCC has clarified that mobile hotspots that are not included with other functionality. So not the hotspot that's in your phone or your tablet, but something that is a standalone device that its only job is to be a mobile hotspot will fall under the same import restrictions as other routers and must seek approval for new models since we're talking about that restriction uh to update you so far netgear and amazon's eero have received exemptions for new models so you won't have any problems with those two brands there will be more to be announced but none of them have been there will definitely be more the wall street journal notes that three quantum computing companies inflection xanadu and horizon quantum have gone public recently with five more planning to do so later this year yeah quantum computing becoming at least an investor, darling, as it gets closer and closer to actual practical applications.

33:20One thing that is not an investor, darling, is DeepSeek. The release of DeepSeek V4 on Friday puts it among leading models, but it doesn't surpass them, meaning we did not get the attention and market panic that was seen by the release of DeepSeek's previous model. Folks, we end every episode of DTNess with some shared perspectives. Today, Andrew from Colorado explains why he's very pleased to hear us mention the IPv8 standard. Yeah, thanks to Andrew for writing in on this. The IPv8 story made me giddy, he said. I haven't had a tech story tickle me quite so much in years. I've been building software professionally since the early aughts, and yet my eyes still glaze over when I see an IPv6 address.

34:06this seems like such a better and more obvious solution. Returning to reading the spec now, Andrew from Colorado. I imagine Andrew speaks for a lot of people. They were like, why didn't we just do that? Double the V4 spec in the beginning. Yeah, it's just simple. It just kind of makes sense when you think about it. It's like we add complexity sometimes when complexity is not necessary. Yeah, sometimes it takes a while to work it out too. Folks, we want to know what you're thinking about. So if you've got some insights into a story, share it with us at feedback at dailytechnewsshow.com. Thank you, Andrew, for contributing to today's show.

34:43Thank you out there listening, Andrew and everybody else, for being along for Daily Tech News Show. You keep us in business, folks. Become a patron if you're not already. Patreon.com slash DTNS. Talk to you soon.

35:07The DTNS family of podcasts. Helping each other understand. Diamond Club hopes you have enjoyed this program.

35:20This 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. Security program on spreadsheets, new regulations piling up, and audit dread? It's time for Vanta.

35:56Vanta automates security and compliance, brings evidence into one place, and cuts audit prep by 82%. Less manual work, clearer visibility, faster deals. Zero chaos. Call it compliance or call it compliance. Get it? Join the 15 ,000 companies using Vanta to prove trust. Go to Vanta.com slash com.