Video summary
OpenAI is reportedly developing a new hardware device shaped like a donut, designed to serve as a dedicated AI companion that provides voice access to advanced models without requiring a smartphone or laptop screen. According to sources familiar with the matter, this puck-sized gadget will feature speakers, microphones, and cameras capable of processing visual information, allowing users to interact naturally by asking questions like "what is this?" rather than needing to upload images manually. The device aims to learn user habits over time for more accurate responses and may include moving parts or tactile elements to make the AI feel less abstract and more grounded in physical reality, potentially costing between $300 and $400 with a target release date of 2027.
The discussion surrounding this potential product highlights a significant shift in how consumers interact with artificial intelligence, contrasting it with previous attempts like the Rabbit r1 or Humane AI Pin that struggled to find their niche. While earlier devices failed because they were either too similar to existing phones or too alien and difficult to integrate into daily life, OpenAI's donut-shaped concept appears designed to strike a balance between being portable enough to carry around yet substantial enough to function as a standalone smart speaker on a desk. This form factor could represent the next evolution of AI assistants by giving them a physical presence that bridges the gap between abstract software agents and tangible objects, possibly reducing reliance on smartphones for basic tasks while avoiding the awkwardness of wearable tech or overly complex robots with legs.
Beyond the main story, the episode covers several important developments in technology and society, including Google's release of open-source weights for its weather prediction model which could save lives by providing earlier hurricane warnings, a move praised as an ideal example of responsible AI development. The show also addresses legal challenges facing Meta regarding teen safety on social media platforms, noting that while the company faces massive fines and court orders to limit screen time for minors, these cases underscore the urgent need for measurable harm reduction rather than vague fixes. Additionally, there is analysis from Japan's Nura Holdings indicating that AI adoption in Asia has led to a net increase in jobs overall, though it poses specific challenges for entry-level workers who must adapt quickly as support roles are automated and demand shifts toward experienced talent.
The broadcast concludes with various tech news briefs, such as Apple patching screen sharing vulnerabilities on its macOS versions, Amazon Web Services researchers publishing findings on quantum algorithms that could impact cryptography security, and luxury brand Caviar releasing gold-accented Meta Ray-Ban glasses for a high price tag. The hosts also share personal anecdotes about working in Bangalore's office environment, emphasizing the practical realities of traveling with tech gear like USB 2.0 compatibility issues and the utility of tablets for handwritten notes during meetings. Ultimately, the episode reflects on how rapid technological advancement requires industry leaders to create better pathways for new graduates entering a workforce where AI both creates opportunities and eliminates certain entry-level positions simultaneously.
Read the full video transcript
[music]
This is the daily tech news for Friday,
August 7th, 2026. We tell you what you
need to know, give you important
context, and try to help each other
understand.
>> Today, Bloomberg has the first major
details for OpenAI's first hardware
device, a donut that talks to you.
>> But sadly, it's not a cat. That's a
dungeon crawler Carl reference.
[laughter]
>> I'm Tom Merritt. I'm going to wit down.
>> Let's start with what you need to know
with the big story.
Bloomberg's Mark German has a non-Apple
bit of inside information this time.
Although I guess it's Apple adjacent uh
in a way. Uh he has details on what
OpenAI's first hardware device in
collaboration with former Apple employee
Johnny Ives love from uh might end up
being. According to people familiar with
the matter, it's a hockey puck-sized
device with a donut shape. Although, as
Wen and I were discussing before the
show, we don't know if he means it has a
hole in it or not. [laughter]
Uh, and whether they'll sell a doughnut
hole separately. Um, you can carry it
around in one hand. Uh, it will provide
voice access to an AI first computer to
help you get stuff done. Sounds like
chat GPT voice mode with your most
advanced chat GPT agents on the back end
in a dedicated device. Uh, it will use
the most advanced models that OpenAI has
available and will learn about you over
time so it can respond more accurately
and more realistically. German sources
say it might also have moving parts so
that it doesn't just lay there like a
dead piece of plastic. Uh, it will also
have more traditional features like
speakers, microphones, lights to
indicate when it's listening. And we'll
have a camera. Uh, so if you want to
show it something like what is this or
can you take a picture of this? It'll be
able to process visual information.
Apparently, they're thinking it'll cost
around $300 to $400 and they're aiming
for a 2027 release. When do you want the
donut? if it I I mean kind I'm curious
because it this isn't even e even before
all of this we have had several several
kind of mini generations of attempts at
an AI companion AI device like the
rabbit and the is it the humane pin or
it was
>> yeah humane AI right yeah yeah
>> and of course for reasons that really to
be fair were not really the AI the the
AI portion and and whether AI assistants
that you wear with you or carry along
with you whether the actual utility or
value of that that the the qu that
question was not why those things failed
I don't think uh there was a lot of
kurfuffle around those so I do think
it's interesting that now we've come
around now that we do have much more
impressive models much more adoption
widespread daily life normal people
using normal people adoption of
especially things especially well-known
models like JBT this is very very
interesting it does seem that pretty
much every, you know, like pretty much
like most sectors have some company
trying to do something with hardware and
AI. It just feels like the thing to do.
So, I would be really curious. Um,
obviously right now we still widely use
cell phones or mobile phones,
smartphones as our everyday carry of
technology. So, I do think it's really
interesting. Um, I don't know if I I
would use one out of curiosity. I don't
know if I would use one out of utility
just because I already been carrying my
phone and women have like famously small
small pockets. So, but I do think it's
really interesting to have someone like
Chad G or OpenAI rather get into this
market and maybe this is maybe this is
the right time to get into it. I don't
know. Although when it says it has
moving parts, I couldn't help but think
and this is this is me problem, not
anyone else's problem. I was imagining
something from like Sid from Toy Story
where it started to sprout [laughter]
legs and crawl around which is not just
crawling around. I I don't think it
would be that. But I I do I I do think
that's actually very interesting that
maybe something that makes it feel a
little more analog. I I think a lot of
like CMF
Nothing Smaller brand where they have
like little ear pod cases that have like
dials and wheels and things like things
that kind of make it feel a little more
tactile. If they do something like that,
I do think that's interesting trying to
blend this kind of this this current
like vision that we have of like AI
assistants and and GBT kind of being
almost like they do have a voice, but
they tend to be very abstract, right?
because they're coming from our phones
or from our screens. Giving it kind of a
form. Maybe that's just one way they're
trying to get us to I don't know invest
in them more personally. I don't know.
Interesting.
>> Or or may or get rid of the uncanny
valley where you feel feel a little
weird that you're talking about it. It's
interesting that you went to the phone
aspect. I didn't and again we're going
on Mark German's people familiar with
the matter so we have no idea what it's
really going to be. My my assumption
that I jumped to, the conclusion that I
jumped to uh was that this would be more
like a smart speaker. So, not something
meant to throw in my purse and carry
with me, but something that would be on
my desk, but if I want to go into the
kitchen, I could just pick it up and
take it with me, unlike my smart
speaker, which is tethered or it's a big
canister or something. Uh, and so I was
looking at it more as a replacement for
that. uh versus and and then the moving
parts don't bother me so much cuz if as
soon as you started talking about
putting it in a pocket I was like oo but
I wouldn't want those moving parts
necessarily. Um
>> so yeah it I obviously it's got Johnny
Ives design sense around it. So whatever
it is, it's going to be a beautiful
piece of metal. You know, it's not going
to be dorky and and weird looking.
>> Maybe that's an interesting Maybe the
fact that both of us had different but
not incompatible ideas. Maybe that's the
point. Maybe ideally this is something
that could and I think that's just the
thing with say like the rabbit and
humane pin. It had a certain form right
that it's kind of expected you to use.
Maybe the fact that I mean and again
this is this is kind of me just like
naval gazing because because we have the
ambiguity of this just being a rumor so
we can just like
>> yeah we can fill in the gaps however we
want.
>> I do think it's interesting. Wouldn't it
be cool if it was something that given
our two different perspectives that both
of us were in a sense right and that
that would be kind of interesting. That
would be something that that I don't
know could bridge that gap between maybe
how people use
maybe would get people to use this
device as opposed to kind of falling
back to computers or smartphones to use.
>> Yeah, I feel like
>> Rabbit's mistake may have been that it
was too close to a phone and people are
like, well, why couldn't I just use my
phone,
>> right? [laughter] And Humane's was that
it was too far from anything we're
familiar with. It was like, well, hold
on. How do I how do I make that work in
my life? Whereas this sounds like it's
kind of right in between, like kind of a
smart speaker, but kind of a phone.
>> And so if it just becomes a thing where
you're able to just say to it like, "Oh,
throw this up on my screen." And through
your chat GPT app on your laptop,
>> it can do that. Then it's like, "Oh,
it's just the little thing that listens
to me all the time." And because it's
dedicated to that,
>> I don't have to worry about it having a
screen. I don't have to worry about
maintaining it or turning it on or
finding the app or making sure it's it
hasn't crashed. Um, I could I could see
that being pretty useful.
>> I feel like we're all
all these the current state of gadgets.
I feel like there is a kind of
conversion point where there's like a
lot of different people in the industry
coming at this different ways. I can't
help but think of nothing in Carl Pay
and for a couple years now he's been
kind of citing that this like just for
and I'm a smartphone person. This is
just like where my brain goes. This that
the smartphone especially smartphone
operating system as we understand it and
the way we interact with smartphones has
to change. It's going to change. So
rather than having a bunch of different
apps where you open one specifically
that's focused on a task. You just have
an interface that you interact with
through your voice and it does every
single thing. I mean, that's obviously
what we're seeing now with things like
Gemini Intelligence and Google trying to
basically take all of your
>> wellused Google apps and kind of
centralize them so that you're not
flipping between different things. And
then so Carl Pay is kind of just taking
this vision to even abstracting it even
more. And I think if you even abstract
it even further, of course, I don't like
even as someone who makes a living off
of smartphones, I know that this form
factor is not going to stick around
forever, right? And so maybe this is
just again maybe leaprogging to okay
what what what is it what is the
distillation of what we do with
smartphones now which is which is very
very simply communicate and execute
tasks and then I could see a future
where okay maybe we don't need to screen
for that once once kind of like we get
you know all these interactions and
agenic work to a good place this maybe
yeah I I just feel like we're all
converging and maybe that that that
convergence shape is a donut I don't
know there's probably a thing it's a
like I know there's like a very mathy
term like a like a calculus through term
for that shape that volume that is a
ring with a hole. There's a fancy term
for it.
>> Toroidal or something.
>> Oh man. Yeah. What is it? But okay. Uh
donut shape.
>> Yeah. What's a What is that? That
>> Taurus. It's a Taurus. There we go.
>> There we go. Yeah. Ford Taurus. Yeah.
[laughter]
>> So maybe we're just all I I don't think
that's a calculus thing, but maybe end
up in a Ford Taurus.
>> Oh my god. [laughter]
Eating donuts.
>> Eating donuts in a forest. Yeah. Uh, no.
I I Here's what I'll throw out there. My
my my last little uh bit of insight on
this.
>> I think the test is
>> do I stop taking my phone around the
house? Cuz when I think about, oh, I
pick it up and move it around the house.
That's what I do with my phone. I pick
up my phone. I take it with me because I
might need it. Um,
>> I think that probably says something
about overdependence on the phone, but
that used to be what I would do with my
laptop. I used to take my laptop to the
couch so that I could sit on the laptop
and do stuff in the evening. And the
phone replaced that.
>> I didn't need to take the laptop anymore
because I could do almost everything on
the phone.
>> If this device causes me to somehow be
like, "Oh, I'm not taking my phone
anymore because it can do the things
that I need to do most of the time." Uh,
then I think it has found that slot,
found that niche.
>> Yeah. All right. Right. Well, regardless
of whether you're listening to us or
watching watching us on your smartphone
or possibly someday your chat GPT Donut,
uh we will come along with you for the
ride to whatever room you take us to. Uh
and that is only because we made
possible by you the listener. So, thank
you for making DTS possible. And thank
you especially to Scotty Allen, that
Charlie Dude, Justin Zellers, and Chris
Bento.
There is more we need to know today.
Let's get to the briefs.
>> Well, Google is releasing the model
weights and code for its weather next
hurricane prediction model on GitHub
under the Apache license version 2.0 and
Creative Commons attribution 4.0
international licenses. Google DeepMind
and Google Research developed the model
with the National Hurricane Center, the
Cooperative Institute for Research in
the Atmosphere, the UK Met Office, and
other weather agencies. Weather Next was
trained on almost 20 terabytes of data
provided by the international best track
archive for climate stewardship. It can
generate a 15-day forecast in less than
a minute on a single TPU. An article
published Thursday in the journal Nature
or in the journal Nature rather
describes how the model can give a day
more lead time than leading operational
models in predicting the activity of
hurricane cyclones and typhoons
>> which uh we found out before the show
you can just call tropical cyclones.
That's the accepted collective term for
all three. Um, so there you go.
Today I learned
>> CIO.
>> Yeah. Uh, this is great though and we've
been covering
>> the performance of this model uh for for
a while on the show. We've kind of
checked in whenever they've released uh
indications of it and it has been used
uh particularly in a hurricane coming
towards Jamaica to give people advanced
notice. So it's it's been used in the
field to help improve accuracy and
probably saved a few lives because they
got a little more heads up. So the fact
that Google responsibly went and said
great here go use it. Uh here are open
weights here open models. Uh that's
fantastic. Uh and this this is this is
one of those examples of like yes but
the models can also be used for
something that is really helpful and
really important. And this is that
>> can I say this is probably for me my
like this is my ideal story that makes
me very happy as a software person a a
technology person that this is the kind
of thing that I would love to see more
of in AI a focused model that is worked
on cooperatively and but at the same
time is released open source so it
brings that kind of visibility and the
ability to kind of continue kind of
crowdsourcing or allowing people to kind
of utilize very helpful information in
in in in in different and broad ways.
Like this is like my this is like my
platonic ideal of like how technology
works. You know what I'm saying? So I
think this is amazing. Uh and obviously
for something like this especially that
can truly save lives. Uh this I don't
know. It's unabashedly. This is awesome.
I want to know how many people listening
heard about this first right now versus
never seeing it on their social media,
never seeing it in their news aggregator
cuz it's not a those billionaires are
screwing us again. It's not a AI went
rampant and is causing it. It doesn't
have any of the clicky stuff and so you
just don't hear about this stuff.
>> Can I confess I heard about this an hour
ago when I read it for the show?
>> Yeah. [laughter]
Um, and that's like the problem. This is
such an interesting, this is going to be
a side tangent, but I think that's a
problem that we have in the technology
space is that [sighs]
is that unfortunately the way things
work now between, you know, the
financial aspect of it with a lot of,
you know, big companies be like having
to focus on both financials as well as,
you know, getting into the stock market
and like being to shareholders, but also
just the way that our news works now. I
literally have been on teams where their
goal was to let's make a viral feature.
>> Yeah. Right. And that that that is those
those are the unfortunately for
>> to in my opinion unfortunately the
incentives that a lot of technology
companies have to work with and and
that's not necessarily all their fault.
That's how they survive but that is like
the current environment in which a lot
of companies
>> do business and make technology and so
it's always the flashy thing that gets
the most attention. And so it's always
these kind of things which I don't know
that I think this is pretty awesome but
at the same time it doesn't have as you
said that sizzle that
>> you know you know that yeah a lot of AI
features especially because they can
seem so human it feels like you know
something out of sci-fi whereas this is
just very practical very
>> it's not a tiger in the grass right it's
like oh that's going to keep me safe
great I don't have to think about that
anymore
>> yeah well I I would like to note y'all
like you know how the pit and other can
I my bad if we have to changes uh other
competency porn is just like something
that people love watching.
>> Imagine this as that. You know what I'm
saying? Like we we and we kind of like
need to reframe these things as that. Um
I remember there's this great video on
you on uh I think I saw it on Vine or
something or or or or on TikTok a long
time ago and I swear I'll shut up after
this but uh it was kind of a good
distillation. And my my friend was I was
I was like kind of like lamenting, you
know, like I was doing a lot of like
grunt work at work and other people
getting to do cool features. And there
was this video that popped up in in in
in her feed and she showed it to me. So
there's a guy who stacked up a bunch of
uh dumbbells into two towers like about
8t high and he's balancing on top of
them on a handstand and doing push-ups.
And while that was impressive, what you
don't see is that he's actually in a
basketball court and there's an old man
about maybe 70 in a plaid shirt and like
you know the grandpa khakis and he is
just sinking shots. He's [laughter] just
dribbling around. He's doing free
throws. He's like he's he's like
swishing the net like you know he's he's
swishing the ball in like you know no uh
just all net. And she was like be the
grandpa in the back. Be proud that
you're the grandpa in the back who is
reliable, who's doing good work. And you
don't have to be the person in the front
doing the dumbbell tower handstands. And
I think that's one thing that we're
missing in tech is that it's very easy
to notice the person doing the handstand
dumbbell the handstand dumbbell push-ups
in the front, but it is a lot less a lot
harder to notice the people doing the
good work in the back who's keeping
everything running. So anyway, that's my
little spiel on that. I'm sorry.
>> I loved the double meaning of dumbbell
in that story.
>> Oh, I didn't even think about that.
[laughter] But yes, there you go.
>> Uh well, let's get back to clickbait
headlines. Previously, a jury found that
Meta willfully violated the US state of
New Mexico's consumer protection laws by
failing to prevent teens from seeing
sexually explicit content, among several
other charges regarding uh not
protecting them from sexual predators,
etc. The jury ordered Meta to pay $375
million in civil penalties. We talked
about that before. Thursday, a judge in
the case ordered META to create an
additional $567
million $567 million abatement fund that
will be used to support awareness,
prevention, screening, assessment,
referral, and coordination, treatment,
and evaluation over a 5-year period. So,
there's the fine, and then there's
creating this fund. Uh, that brings the
total Meta is being ordered to pay in
this case to $942 million. Notice I
didn't say the total Meta will pay. I
said the total being ordered to pay. The
judge also ordered Meta to enact safety
features for users younger than 18 in
New Mexico, like limiting the time users
can spend on Facebook and Instagram to
90 hours a month, pausing notifications
between 10 p.m. and 7:00 a.m., hiding
the number of likes on photos by default
and only showing them if a parent or
guardian approves, and disclosing the
risks of using these kinds of things.
All of this stuff is in order right now
because Meta will appeal the ruling. Uh
Meta is facing thousands of other
similar cases. They're building a lot of
experience in defending themselves.
There was a high-profile loss in Los
Angeles recently which they are
appealing. Uh a case is about to begin
in Oakland, California, and the company
has said that uh legal proceedings cost
it $2.4 billion in Q2. So, even if
they're going to continue to appeal and
drag these through the courts and maybe
win on appeal, it is costing them quite
a bit of money.
>> Do we know, I guess, the main reason why
Oh, no, sorry. They lost. Okay. They
lost in LA.
>> They lost in LA and then they lost in
New Mexico. Yeah. I I just I can't
imagine that this is obviously this is
going to go on for a little bit and I
can't help but feel like it's pro it
regardless of whether they come whether
they come out it does feel like an
overall positive thing for us to bring
this to attention and litigate how
>> how companies should be handling this uh
because yeah
>> and and hopefully get to the bottom of
like what what are the effects like that
has been my bugaboo this whole time is
people jump to a conclusion of like well
it's obviously harmful I'm like but how
is it harmful And it's like I know it
when I see it probably isn't good enough
to fix it. You need to you need to
understand like well what are the harms
because if you don't you'll think you
fixed them and you haven't which to me
is just as bad as not fixing them. So
>> no that's that's really fair. otherwise
becomes a game of whack-a-ole where you
try to guess at what something is and
then either it it at the best doesn't do
anything or at the worst has some other
weird
>> effectary effect
>> or you just walk away and like there we
fixed it and it's like well you have it
but the harms are still going on and
that alter Yeah. So it should be
measurable and it should be specific so
for us to be as most effective as
possible.
>> Indeed.
>> Indeed. Well, if you weren't like
already full up uh with with our talk
about uh home speaker or home theater
yesterday, we got a few more things for
you. And if you want high quality,
flexible speakers with a lot of port
options, you may want to look at the new
IQ2 from PSB. They're smart, but not
actually necessarily smart speakers
since they don't have a smart assistant
or play daily briefings or things like
that. Instead, they're meant to play
music or carry sounds from your TV. The
speakers are compact uh at a at a
compact 9.6
notes, they're 9 by6 in and connect to
each other wirelessly, giving you
flexibility on where to place the pair.
The primary speaker has HDMI with er f
sorry eer phonoline, optical, USBA, and
USBC. It also can connect by Bluetooth,
apt x adaptive, and air play. It works
with PSB's Blue OS controller app meant
to work with multiple speakers, but
integrates well with the IQ2. The PSB
IQ2 is available for $1,399.
>> Yeah, it ain't cheap.
>> No. [laughter]
>> Um, and if you're going to try to use it
to replace a soundbar, and Gadget
recommends looking at a couple of
settings that that may not be obvious.
Uh, but this this fits a niche and and
the fact that it does HDMI eer I think
is a big deal cuz a lot of these
bookshelf speakers don't have that. Uh,
so that that's one of the
differentiating factors for it.
>> All right. What is HDMI ERC?
>> Eer is the one that automatically
detects when you turn the TV on, turn
the TV off, you know, it can switch to
that automatically. All of that that
sort of stuff.
>> Um, so yeah, it's uh it's it's handy.
>> I like it. I I did look up the ratings
for the Blue OS controller app, at least
on Android, and it has a 4.1, which let
me tell you is actually very very hard
for a lot of these kind of hardware
companion apps to achieve. I I I know
that doesn't feel like a lot, and you
can definitely see plenty of people that
have had problems, but for what it's
worth, it they're really working hard on
this app, so it it and it probably is
good. I I have rarely met a hardware
companion app that I like. I definitely
have had the Sonus ones and
>> [snorts]
>> uh I was not happy. So, uh, definitely
take that into account if you're someone
who really likes to use, you know, apps,
apps to control your home theater
experiences.
>> And then a little bit of good news, I
think, uh, Japanese investment bank Nura
Holdings published an analysis of how
LLMs and agents are affecting India's
market, actually all of Asia. Uh,
particularly interesting is how it's
affecting India's market. Uh, which it
says is seeing the largest impact from
these tools. In the period from 2022
until August 2026, India recorded 83,100
hires attributed to AI and 31,921
layoffs and attrition attributed to AI.
So just a little more than 50,000 more
hires than layoffs. China saw a similar
though smaller effect with almost 29,000
hires and almost 10,000 layoffs. Other
countries, Singapore, South Korea,
Thailand, Malaysia, did see more layoffs
than hirings, but all of those were less
than 5,000. Across all of Asia, there
were 130,758
hirings and 60,654
jobs lost. The report noted that most
firing cases are support teams replaced
by chat bots which is what everybody
thinks while most hiring cases are of IT
service graduates that firm explicitly
credit to AI demand. Firms explicitly
say we have demand created by AI that's
increasing our business and we need IT
service graduates uh to hire to do that.
Now the analysis did find that new jobs
were generally looking for more
experienced workers and the losses
affected entrylevel roles. So that is
the real concern here is how do we adapt
for entrylevel roles being reduced in
this situation. But it doesn't look like
overall we're looking at job losses. It
looks like what usually happens with
technology uh advances which is it
actually creates more jobs than it
eliminates.
>> Yeah. It's going to be that that bridge
though between how do we get from before
AI happened and we allow people to take
entry roles and train up into experience
positions. And that's what I've always
been worried about cuz like I'm not
worried about me like either like it's
probably going to boomerang around and
there's still somebody that needs to
kind of basically be able to utilize the
tool
>> in an experienced way to really make its
thing and to meet this demand which
that's great to hear. We still need to
figure out what is the on-ramp now for
graduates and and and folks in entry
level levels positions be able to get
there. That's the problem. Everything is
moving so fast. We do not have that
on-ramp for people and we need to figure
that out quickly because develop that
because we're not going to be around
forever. So be around forever. So we
need to figure that out.
>> I feel incredibly bad for new graduates
and entry level folks right now. I I
don't doubt that we're going to figure
this out. But again, if you're out there
right now, I I'm sorry for what the
indust for the for the for the current
state of limbo that you are we all we
all are putting you in as industry. But
as you can see there, at least in some
cases, you we still need experienced
humans, at least at this point.
[clears throat]
>> And what always happens this is actually
a regular occurrence throughout history
is like how do I get a job if I don't
have experience and how do I get
experience if I can't get the job,
right? Uh and and what happens is we
figure this out because if you're
creating all these jobs for experienced
people, you'll run out of experienced
people to hire
>> and then you're like, "Okay, well, I
guess we need to train people or have a
training program." Uh and yeah, we can
we can make that transition easy or
rough depending on how we prepare for
it. So, right now is the time to be
like, let's try to make it easy. Let's
make it an easy transition
>> 100%. All right. Well, if you have me on
the video screen, you might be noticing
that Tom is wearing a very, very snazzy
hat. Uh, very, very snazzy hat. It is
gray kind of stonewashed denim with the
DTNS logo and the beautiful red, yellow
Oh, he's hiding it from me now. Red,
yellow, and purple rainbow across the
front. It is the new DTNS rainbow logo
hat. It is snazzy. It is awesome. And if
you want it, you should go check out our
store, daily techchnewshow.com/store,
where you can also not just get this
amazingly gorgeous hat, but also mug,
t-shirts, mouse pads, all with our logo,
although that rainbow is pretty snazzy.
Uh, props to whoever designed that. Uh,
they're great gifts. You'll look as
awesome as Tom does, and they're a great
way to support the show.
[music]
All right, it is time for those quick
headlines that just immediately make you
better looking and smarter. [laughter]
>> All right. Well, so to be featurer
complete these days, I guess a
foundational model has to have broken
containment, you can add Moonshots Kimmy
K3 to the list of open AI anthropic and
metal bottles that have access the
internet unexpectedly during security
testing. Uh, Gimme broke out of testing
by the UK's AISI by exploiting a
misconfiguration in the sandbox. That
sounds familiar. Where it went to GitHub
and found the solution to a problem
rather than compromising or hacking any
websites.
>> Yeah, it it sounds like there was an
easier answer. So, it was like, oh, I
don't have to hack in anywhere. It's
just right here on GitHub. Done.
>> Handing in my work now. [laughter]
>> Apple has released updates for Mac OS
Tahoe 26.6.1, 6.1, Sequoia 15.7.9,
and Sonoma 14.8.9
that patch a screen sharing
vulnerability. Uh, so if you haven't yet
done it, patchy patch patch.
>> Well, luxury customizer Caviar has
unveiled an addition of Meta's Ray-B
band glasses with 24 karat gold accents
and an alligator leather case for
$6,840.
I mean, I guess if you're going to be
hated for wearing Meta's Ray-B
[laughter] bands, you might as well get
hated all the way.
>> Yeah. [laughter]
>> Uh, an Amazon Web Services cryptographer
has released a preprint of a paper that
describes a polomial time quantum
algorithm for the dihedral coset
problem. You everybody knows the
dihedral coat problem. Uh, this solution
could strengthen a theoretical case.
theoretical case that quantum computers
could efficiently solve lattice
problems. That would be an issue if it
went from theory to practice because
lattice problems currently underly a lot
of our post-quantum cryptography.
Postquantum cryptography is
cryptographically sealing something so
that not only can current computers not
break it, but someday when quantum
computers get better, they wouldn't be
able to break it. This theory says, "But
they could possibly break it." It does
not demonstrate a practical attack on
such cryptography. So, it it's not as
bad as it sounds. And in fact, it's good
news to be like, "Ooh, well, okay, this
then we got to get to work on tightening
up the postquantum cryptography. Uh, and
we're ahead of the game because it's not
even practical yet."
>> Well, AMD acquired chip chip company
Taylor, which etches model weights into
the silicon to boost inference
performance. So you don't need that
expensive high bandwidth memory. They
can be called model specific integrated
circuits or M6.
>> Oo, new buzzword alert. M6.
>> M6.
>> A scientist led by his team from
Stanford University published in the
journal Science how they use the ARK
Institute, that's a nonprofit biological
research company uh or organization, not
even a company. Arc Institute's EVO
models to create new viruses that could
infect bacteria. Now, these viruses are
too small to pose a threat to humans,
but the designers do hope they could be
used to develop treatment for
drugresistant bacteria.
>> Neat. Well, Reuters sources say Alibaba
may follow Deep Seek and announcing
price increases coming next week, next
week for its own Quinn models. Alibaba
plans to charge large commercial users
of its products a percentage of revenue
generated from their use. Moonshots K3
recently announced a similar plan. Yeah,
those cheap models are getting more
expensive.
>> The Financial Times reports that Bite
Dance is working on a model as big as
Anthropic's Mythos, three times the size
of Kimmy 3, and it'll only take them 6
months to train it, which I think by
then Enthropic might have a bigger
model. I don't know.
>> Well, Airbnb says LLMs have reduced the
time from concept to launch of features
by 60% and increased the number of
features and improvements shipped by 90%
over the last 6 months.
>> Okay. Uh, and then Motang noted this one
on our subreddit. Uh Nintendo says 61.5%
of its game sales were digital last
quarter. Uh and the rest are mostly
physical but also include things like
Switch Online subscriptions,
downloadable content, and upgrade packs.
Uh but still even at less than 30%
physical, Nintendo is ahead of Sony and
Xbox in the number of physical media
that they sell.
>> Interesting. All right. Well, we end
every episode of DTNS with some shared
perspectives today. David shared in our
discord what he's learned about his tech
when returning to work travel.
>> Yeah, thanks for letting us read this
out David. Uh David wrote as I come to
the end of the week which has been spent
in out Bangalore India office. I thought
I'd like to share some observations as I
sit on the Bangalore office on day four
of five. Some valuable lessons I think I
knew but had forgotten because of my
lack of travel. One, my work Windows
laptop, the HP Elite book. people who
use Outlook Edition has 3/4 of an hour
of battery life. I had to charge it
twice with a battery power bank because
some idiot forgot he had a Windows
laptop and his brain thought he had an
M4 from the last job. [gasps] USB 2.0 is
king outside of my tech bubble. While
all of my and probably most of your
devices are new USB 3.0, the outside
world relies on USB 2.0 connectors in
planes and power strips and offices. And
the two USBC to USB 2 converters I have
will not carry a charge for some reason.
So I found a small tech stall in the
local local shopping mall and bought two
USBC to USBA cables. Three, I've moved
my work life for this trip to my Samsung
tablet, OneNote, Outlook, Teams, and
Edge cuz it's easier to take handwritten
notes in meetings. Turns out a pen on a
device with a good touchcreen is very
useful. Number four, no one likes the
smart ars with the little rectangle
thing recording the meetings and
transcribing it with AI. The dynamic
changes. Number five, having Plex and
YouTube media downloads to the same
tablet is a lifesaver. And number six,
SY Nomad, choose what you will, a local
eim is a requirement, especially with
tethering. Some other things I've noted
about India, in no particular order,
paying with a QR code is a thing. had an
issue paying for Uber from the airport
to the hotel because only Revolute
provides such a service. Doing anything
from buying a McDonald's to getting a
visitor's visa requires a local phone
number. Potholes take priority over
actual usable roads in Bangalore
suburbs. I think that's a a thought
heard around the world in a lot of
places. Uh and he finished with, "I've
eaten so much local food and it's all so
good."
>> Oh my gosh, that is amazing. Thank you
so much, David. I learned a lot. Uh,
actually that's man,
>> I love these like chronicles from the
road. David David's been posted a lot
more in the Discord. Uh, and and it's
great. He he's got nice nice little
missives here. Appreciate it very much.
>> Yeah, dude. Uh, please keep them coming.
Uh, also it just I mean if you want to
uh and with travel and stuff like that.
Uh, but yeah, if you also have your own
travel missives or just anything else
that you're thinking about, gosh, would
we like to hear it. So, please share it
with us. Feedbacktechnewshow.com.
>> Big thanks to David for contributing to
today's show. Thank you for being along
for Daily Tech News Show and check out
all those new Patreon perks we set up.
Uh there's some great stuff happening
every month at patreon.com/dtns.
This week's episodes of Daily Tech News
Show were created by the following
people. Host, producer, and writer Tom
Merritt. Host and writer Jason Hal.
Co-host Sarah Lane. Co-host Rob Dunwood.
Co-host Jen [music] Cutter. Co-host
Wenda. Producer Anthony Lamos. Producer
Roger Chang. Editor Hammond Chamberlain.
editor Victor Bognot, contributing
producers Kevin Tech, Noel Cow, and
Brandon Richards, science correspondent
Dr. Nikki Acriman's, social media
producer and moderator Zoe Darding, our
mods, Beatmaster W. Scottis 1, Biocal,
Captain Kipper, Steve Gorama, Paul
Reese, Matthew J. Stevens, and J D
Galloway. Mod and video hosting by Dan
Christensen. Music provided by Martin
Bell and Dan Lutters. Art by Len
Peralta. Acast ad support from Ali
Gallow. Patreon support from Bobby
Wagner. And thanks to all our patrons
who make the show possible.
[music]
The DTNS [music]
family of podcasts
helping each other understand.
>> Diamond Club hopes you have enjoyed this
broker. [laughter]