OpenAI Dots, Anthropic's Ominous IPO, & Meta Glasses Spying | Tech News of the Week

[00:00:03.08] - Ned
Well, I know that basically the only successful sites on MySpace at a certain point were bands, especially scene bands. So like emo, punk, pop punk type bands. That's what was popular on that.

[00:00:19.00] - Chris
Yeah. It's very different than the email pom-pom bands. Email pom-pom? Oh my God. We're gonna do great. We're gonna do great.

[00:00:30.07] - Ned
This is a good start. Hello, alleged human, and welcome to the Chaos Lever Podcast. My name is Ned, and I'm definitely not a robot. I'm a real human person who has feelings that I like to emote with my pom-poms. With me is Chris, who is also here.

[00:00:49.10] - Chris
What's sad is that is 100% a true story.

[00:00:53.03] - Ned
You know, we do that.

[00:00:53.20] - Chris
And listeners, it's as uncomfortable as you think it is.

[00:00:56.23] - Ned
It is, especially when I put the uniform on. We actually have some pom-poms in the house that we got from a basketball game or something.

[00:01:07.00] - Chris
And yeah, sure. From a basketball game.

[00:01:09.08] - Ned
That's what they call it these days. And unfortunately, our cat really likes to chew on anything that is of that plastic.

[00:01:18.23] - Chris
Oh, sure. Yeah.

[00:01:20.18] - Ned
Uh, so I've cleaned up more than my fair share of cat vomit with pom-pom strands in it. So we finally figured out how to hide them.

[00:01:28.17] - Chris
That's— that seems wise. But on the other hand, I mean, if you enjoy it, you know, live your truth.

[00:01:34.06] - Ned
Don't yuck my yum, I guess. Oh God, this is getting worse. Let's talk about some tech garbage. Um, before we get into the news of the week, uh, If you're running PHP on anything right now, and you might be, even if you don't realize it, uh, you should probably go patch that. Well, we'll do a full story on this probably next week, but seriously, just like stop listening and go get patching. Man, the internet's on fire, dude.

[00:02:00.17] - Chris
Again.

[00:02:01.05] - Ned
Again. Kind of like Lake Michigan. Anyway, let's talk about the news. OpenAI flatters Meta with DOTS. A few weeks ago, OpenAI held their Dev Day 2026 with a whole slew of announcements around their various products. Out of the hype, there were probably 3 things most people should pay attention to. First off was the announcement of DOTS. Not the gummy candy that welds itself to the crevices in your molars until a dentist pries them out. No, This is worse. This is a new service along the lines of a managed agent instance. Think about OpenClaw or Hermes, but on a machine hosted by OpenAI with less choices about which model to run. Sounds great, right? Each Dot is a cloud computer hosted by OpenAI and running GPT Astra with built-in connections to a ton of other apps. You can chat with your Dot via the ChatGPT interface, or connect it to Slack or Microsoft Teams if you really hate yourself, with other interfaces coming along at some undefined point in the future. Dots are essentially a nod to Meta's Muse project, which is also a cloud-hosted agent that leverages Meta's models instead. That was announced by Meta a few weeks prior to this announcement.

[00:03:28.23] - Ned
And this is OpenAI saying, hey, us too, we do that. Both of these products represent the simplification of an OpenCLAW or Hermes instance. With both of those, you're the one responsible for managing the agent software and connecting it to the appropriate models. MuserDots abstracts away the fiddly management side of things, attempting to give a streamlined experience to those who are less tech-savvy or Let's face it, too busy to be bothered. Whether or not you should actually trust Zuckerberg or Altman with all of your personal information is another question entirely. The answer is no.

[00:04:08.06] - Chris
It's definitely no. Yeah.

[00:04:10.03] - Ned
The second announcement was a price hike.

[00:04:13.02] - Chris
Yay!

[00:04:14.02] - Ned
Well, it wasn't presented that way, but in effect, that's what it is. If you have an OpenAI Pro subscription, Your usage allowance was cut in half. But don't worry, because the newer models are more efficient, and you didn't need all those extra tokens anyway. Look, we launched GPT-6.1 SOL, which is almost as good as Astra, but 1/5 the overall cost, and we just know you're going to love it. Lastly, there was the announcement of UltraFast mode. For people who are trying to token max in a hurry. Now you can get a response 8 times faster at 8 times the cost. If you're on a subscription plan, you'll need to bump up to the $500 a month tier to use UltraFast, and you'll probably run through your allotment in about 10 minutes. You can see the writing on the wall here. AI— OpenAI is massively unprofitable, and every move they make from here on out will be pushing folks away from subscriptions And finding ways to burn tokens faster and charge us more. It's capitalism in its purest final form, and we applaud OpenAI.

[00:05:30.13] - Chris
And just to add a little dollop of additional cynicism on top of this nightmare hellscape, um, the PHP bug that Ned referenced at the top, OpenAI helpfully released a very simple JavaScript tool that you can use to search your internal or other external URLs to see if you're affected. And they did it in 2 different ways. One was a script that you could run, and the other was, you guessed it, a skill you can enable. Why would they do that, you might ask? So you can burn more tokens.

[00:06:01.06] - Ned
Of course. You could run this simple script that costs you no tokens, or you could have your model run it and burn all the tokens while it uses the skill.

[00:06:12.01] - Chris
People are shocked as Microsoft kills off Publisher. Following 1,000-odd days of Microsoft warning that they were going to kill off Publisher. Ah, bake sale, birthday list, and church newsletter authors lament. On October 1st, 2026, Microsoft pulled the plug on Publisher. Not that they're going to stop updating it or that support won't ever happen again. It literally stopped working. At least for M365 customers. Publisher was a budget, and if you're of a certain age and proclivity, easily pirated program that you could use to do graphics on documents. Now that doesn't sound amazing, but to be clear, this program is 35 years old. We didn't even have graphics back then. You could definitely not do this shit in Word. You could. overpay for Adobe InDesign if you wanted, or if graphic design was your passion, you could use Publisher. Publisher was the program that taught an entire generation of parish secretaries and enthusiastic tweens that Papyrus is a font and that any flyer improves with more clip art. And this was before we even had the word tween, I think. Possibly.

[00:07:32.09] - Ned
Probably.

[00:07:34.16] - Chris
Microsoft had kept this program going long after its need in the market was supplanted by native capabilities in Word and, so help me God, if you do this, PowerPoint. You're going to do this, aren't you?

[00:07:47.06] - Ned
I already do.

[00:07:50.19] - Chris
Microsoft also, after a certain point, got sick of maintaining it, which, I mean, fair. They announced in 2024 that the end was nigh. And as of a few weeks ago, The end has nade?

[00:08:05.10] - Ned
Indeed.

[00:08:07.12] - Chris
LibreOffice took this opportunity to write a back-patting blog post about how they still support .pub files, which is funny for 2 reasons. One, they've been able to do that since 2014. And 2, 2014 is coincidentally the last time LibreOffice bothered to update their user interface.

[00:08:26.13] - Ned
Boom. Got it.

[00:08:28.21] - Chris
I was going to make a closing zinger here about the shuttering of publisher is worse than when Google closed shop on Reader, but it is still— it's still too soon. Goddamn it.

[00:08:37.23] - Ned
It will never not be too soon. Anthropic IPO is ominous. Not exactly the vibe you want to encourage investment. Anthropic filed their S-1 draft to the SEC back in June of this year. This is the prospectus they will use when they're shopping around to investment bankers in anticipation of their IPO. The S-1 document has leaked because of course it has. And what did we learn? There are a few top line items to get out of the way. For starters, Anthropic had an operating loss of $8 billion for 2025. That is based on revenues of $4.6 billion for the last year. In Q2 of 2026, they pulled in $11.5 billion of revenue. So it's pretty clear that revenue is growing quickly and Anthropic could be something approaching maybe profitable sometime in the indiscriminate future, except for 2 things. First, they have a net loss of $42 billion for 2025. It's a net loss. Which is a little bit more than the $8 billion of operating loss, which, which one of those actually counts? I don't know. I don't, I don't fucking math, but whatever. It's the second one. So the path to profitability is a little steeper.

[00:10:04.04] - Ned
In addition, they have $518 billion in cloud and infrastructure commitments over the next few years, and 25% of their revenue comes from 2 customers.

[00:10:16.14] - Chris
Oof.

[00:10:18.01] - Ned
Will their revenue grow to cover their commitments and operating costs? I have no clue. Go talk to a financial advisor or, or listen to Planet Money or just like do the opposite of whatever Jim Cramer says.

[00:10:31.17] - Chris
My God, I forgot Planet Money was a thing. Is that still a thing?

[00:10:34.14] - Ned
I actually don't know. Slightly more amusing or Depressing, depending on where you sit on this whole AI will end the world thing, is 33%. A full one-third of the S-1 is dedicated to the existential risks of AI. Nothing gets in the way of a great IPO like a mass extinction event. After all, sources quote Anthropic as saying models pose a, quote, catastrophic or existential risk to humanity. End quote, and that their current models have shown signs of resisting shutdown, concealing information, or conducting blackmail. As we all know, there is no greater risk to humanity than pictures of your mom and dad at their swingers club. Seriously though, the whole existential risk to humanity thing is wildly overblown, and it makes one question. Why it takes up almost a third of a document meant to sell investors on buying stock in Anthropic. Are they bragging about how powerful their technology is? Ooh, it could end the world. Are they attempting to engage in regulatory capture? Do they truly believe these are systemic risks? Have they even stared into the eyes of Roko's Basilisk? The answer to all of these questions is, I don't know, probably.

[00:12:06.15] - Chris
LibreOffice is advertising zero AI as a design position. LibreOffice, an open-source office productivity suite best known for not being Microsoft Office, released version 26.8 back in August. And this month they released 26.8.1. Which did all the usual things that a minor release does. Bug fixes, inefficiencies, probably something about kerning. What was interesting about the release, though, was the way that LibreOffice came out of the gate with it. They literally said that the 26.8 versions, quote, contain no generative AI features, unquote. It is beyond sad that a company has to pat themselves on the back about something as simple as this. But here we are. The reasoning they put out for this is pretty sound, and for anyone working with tightly regulated environments, it's going to sound real good. Basically, they say that no data, no metadata, no document, no nothing will ever leave the machine that is working on it unless that user explicitly uploads the information on their own. That, if you're keeping score at home, is the exact opposite of what jamming GenAI into Word does. GenAI just takes your documents and sends them everywhere with no attribution or compliance assurances or an audit trail.

[00:13:33.16] - Chris
Why? Because fuck you. That's why. And probably also something about kerning. The fact is that the carelessness and insecurity of data flows from AI initiatives is astonishing. and staggering and other words that end in -ing. Mm-hmm. And it shows no real momentum towards responsible design. There have been an increasing number of countries and other highly regulated companies and organizations who have switched away from Microsoft Office 365 for exactly this reason. The Austrian military, a lot of German workplaces, the Ministry of Defense in Italy, to name just a few. And the reasoning they gave is not just price. I mean, it's definitely price, but it's not just price. Fun fact about 26.8 as well. LibreOffice has finally, blessedly updated their user interface. Now, some less scrupulous news outlets may have said crazy things like LibreOffice hasn't updated their interface since 2015. What a bunch of ill-informed baboons, am I right? Who, who does that? Jeez. The new version of the interface has been in development for over a year. It is called Notebook Bar for some reason, and now with 26.8, you, the end user, can enable it. And I do mean enable it. It is not on by default.

[00:14:56.04] - Chris
It's a little bit buried in the menu system, in the bottom of a locked filing cabinet stuck in a disused lavatory with a sign on the door saying beware of the leopard, but it's totally there. Now, full disclosure, I have not actually tested it yet, but I can confidently say it's better.

[00:15:14.23] - Ned
Windows 10 support costs double, but can you really put a price on keeping your sanity? There's an informal rule about Windows releases that basically says every other release is awful. I mean, they're all kind of awful, but the every other one is especially awful. Windows ME, Vista, Windows 8, and Windows 11 are all great examples, especially those last 3, as enterprises and personal workstations alike clung to the preceding OS with the tenacity of a Rottweiler. And why shouldn't they? The operating system is supposed to run in the background, quietly servicing the requests of the end user and their applications. A good OS should be like good plumbing. You don't notice it unless something goes horribly wrong. Windows 11 has been a little bit more like the plumbing in a frat house on rush week after a kegger on Taco Tuesday. Yuck. Sadly, its predecessor, Windows 10, which I will remind you was supposed to be the last version of Windows ever, fell off long-term support in late 2025. Enterprises that wished to continue running the venerable OS while receiving support from Microsoft would need to pay a hefty extended support cost. Not only that, but the cost for each license would double every year for a maximum of 3 years, starting at $61 and possibly hitting $976 per license.

[00:16:51.11] - Ned
Depending on how you interpret their statement. Microsoft is clearly trying to avoid another Windows 7 fiasco by making it financially infeasible to keep using Windows 10. Now, Chris, I know what you're thinking, 'cause you love Windows and Microsoft so much. If Windows 10 is in extended support and 11 is the only thing on the market right now, does that mean that Windows 12 is rapidly approaching? The shorter answer, Not the shorter. The short answer is no one fucking knows. There've been lots of rumors, but if you were one of the roughly 22% of people still parked on Windows 10 and hoping to leapfrog over Windows 11, I wouldn't keep waiting. Probably just switch to Linux. Microsoft does have Project Windows K2, which is an internal project to make Windows 11 suck Allegedly, Microsoft will be focusing on performance, reliability, and UI improvements instead of packing AI into every last nook and cranny of the OS. The 26H2 update is now available and is supposed to be part of the K2 effort. I have installed it and I've noticed nothing. But also no additional AI, so like Fine. Maybe, maybe Windows 11 will morph into Windows 12 through these updates, or maybe 2027 will be the year of the Linux desktop.

[00:18:22.02] - Ned
2 equally likely outcomes.

[00:18:26.02] - Chris
I think it's funny that they called this Project K2. Uh, they probably were just like, oh, that's a cool mountain. But like, it's supposed to be one of the hardest mountains to climb in the world. So I think that's accidentally apropos because fixing Windows 11 is basically an impossible effort. Yep. And I don't think that it's a 0% chance that some of the developers might freeze to death.

[00:18:52.06] - Ned
Fair enough.

[00:18:54.10] - Chris
Update, Facebook Ray-Bans are not only creepy as hell, they're also poorly engineered. Hmm. Okay. Maybe that's not an update. in the sense that everybody probably assumed and/or knew that much already. The current crop of impatient and incompetent oligarchs have shown no qualms whatsoever about forcing poorly designed and wildly insecure products onto the market, all in the name of stealing your personal data and selling it with impunity. Why would Zuckerberg be any different, really? I mean, really, he kind of is the pioneer of being technologically awful.

[00:19:27.16] - Ned
Really an innovator.

[00:19:29.21] - Chris
There have been enough stories about how trivial it is to circumvent the security features in the glasses so that it becomes undetectable when they are in recording mode. But this new report this week details user experiences with the glasses recording their owners without consent. Irony.

[00:19:47.01] - Ned
Hmm.

[00:19:48.17] - Chris
The glasses are supposed to only start listening to their owner when either the person wearing them pushes a button Or says out loud, hey Meta. Tech-savvy users have been reviewing the recordings made by the glasses and seen clear evidence that they were recording audio without either of these 2 things happening. Many users have stated that the rogue recordings were occurring even with the wake phrase option completely disabled. To be fair, this was only a small number of reports. Some users though have reported that it's been happening over multiple generations of the product. Facebook also claims that it's fixed. Their claim in full, quote, this issue has been resolved. People can manage, review, and delete voice log interactions with Meta AI via settings at any time, unquote. A little terse. But a great reply if you don't notice the second half where the remedy for unwanted recordings is that you can go delete the unwanted recordings on your own if you notice them. Oh, and, uh, we're not going to tell you why this happened in the first place or how we fixed it or how we know it's not going to happen again because— and say it with me, folks— You're Because fuck you, that's why.

[00:21:15.01] - Chris
So what should you do? I mean, if you own a pair, you should definitely log in and check the voice logs to see if this has been happening. After that though, you should go out to Target. You should buy a Ninja blender. Treat yourself.

[00:21:29.15] - Ned
Yeah.

[00:21:31.00] - Chris
Put the glasses in the blender and then turn the glasses into a fine powder. Then you won't have to worry about another shittily made device recording you when you don't want it to. And also you're gonna have a kick-ass new blender.

[00:21:43.23] - Ned
True.

[00:21:45.03] - Chris
Don't sleep on the Ninja. Just maybe, maybe give it a good cleaning before you start making protein shakes.

[00:21:53.08] - Ned
Mm-hmm.

[00:21:54.01] - Chris
Meta glasses are probably as physically toxic as they are societally toxic.

[00:21:58.14] - Ned
I mean, I do have a Ninja blender and I use it every day to make my morning shake.

[00:22:04.15] - Chris
Yeah, I also have a blender.

[00:22:07.13] - Ned
And that's it.

[00:22:10.17] - Chris
Hey, thanks. Oh wait, that's your line.

[00:22:12.02] - Ned
No, we have a new segment.

[00:22:13.23] - Chris
Oh, right. I almost forgot. Carry on.

[00:22:16.02] - Ned
I, I was inspired by an Elon Musk tweet, whatever, X, whatever, skeet, whatever they do.

[00:22:22.23] - Chris
They're called tweets, kid. Focus.

[00:22:25.23] - Ned
And so I'm calling this new section Christ, What an Asshole. So here are a few quick-hitting news items that didn't warrant a full take, but in the spirit of Chaos Lever, They highlight how the tech industry is filled with dumb assholes. First one, Elon Musk plans to rename SpaceX AI to SpaceX SI because super intelligent is as super intelligent does. Next up, certification authority eMudra uses the number 4 for its private signing key. It's awesome. Well done, everybody. And lastly, and my absolutely favorite human in the world, let's say, FCC Chair Brendan Carr is threatening broadcasters about polls that show how much people hate him, his boss, and his party. Yay, free speech.

[00:23:26.10] - Chris
I mean, this is literally a free speech violation, I think.

[00:23:30.12] - Ned
Oh, very clearly.

[00:23:32.10] - Chris
Okay.

[00:23:33.23] - Ned
Hey, thanks for listening or something. I guess you found it worthwhile enough if you made it all the way to the end. So congratulations to you, friend. You accomplished something today. Now you can go roll your own private keys using 4-digit numbers and make yourself your own certifican— certifican? Certificate authority. That's the one. You were—

[00:23:54.21] - Chris
Sip teklifind. Jesus, that sounds like something the pharmaceutical company would sell us for $1,000 a pill.

[00:24:02.02] - Ned
Oh, go get my email pom-poms. You can find more about the show by visiting our website, chaoslover.com, where you'll find show notes, blog posts, and general tomfoolery. We'll be back next week to see what fresh hell is upon us. Ta-ta for now.