Is Google's Gemini Winning?, Thinking Machines Drama, Claude Cowork’s Potential

Channel: Alex Kantrowitz

Published at: 2026-01-19

YouTube video id: ugV4vz4R3kc

Source: https://www.youtube.com/watch?v=ugV4vz4R3kc

As Google partners with Apple and
unveils deeper personalization, is
Gemini winning?
Thinking Machines Lab, meanwhile, is
falling apart. And Anthropic did it,
releasing Claude Code for non-coders.
All that and more is coming up right
after this.
Welcome to Big Technology Podcast Friday
edition, where we break down the news
>> [music]
>> in our traditional cool-headed and
nuanced format. We have a great show for
you today because
Google and Apple finally came to a deal.
They're going to fix Siri.
And has that catapulted Google to the
lead position [music] in the AI race.
We're going to talk about that. We're
going to talk about Thinking Machines
Lab falling apart. We're also going to
talk [music] about Anthropic on the
heels of our conversation about Claude
Code last week coming out with Claude
Co-work, the Claude Code version for
non-coders. It's a mouthful, but it's
meaningful. And joining us, as always,
on Friday to do it is Ranjan Roy of
Margins. Ranjan, great to see you.
Claude Code, Claude Co-work, the Claude
Code for non- coders. Was that it?
Put it all in the generative harness.
Put it in a harness. By the way,
>> god, we're going to
>> our conversation about harnesses last
week generated a lot of intense feedback
from our audience. People are really on
team harness. So, I'll take the
opportunity today to stand here and tell
all the haters
I still don't like the word, but I
respect you for enjoying it.
I'm glad harness is going to Imagine
when a harness is the word of 2026. But
But I I'm just [clears throat] excited
that to talk about our first story
because Siri has a shot. Siri has a
shot.
And Siri's talking in the background
because I just said that on my home pod.
But There's only one way to introduce
this story, and that is that Apple has
harnessed the power of Gemini and put it
into action in Siri. There is a new deal
that has happened between Google and
Apple that may lead may lead to the
inevitable improvement of Siri. And I
say inevitable because
goodness gracious, it can't get much
worse.
But I think to really think about what's
happening in the AI world and what this
story means, we'll of course get to what
Google and Apple's deal is going to look
like. We have to look at it in terms of
the context of the bigger AI race and
just how meaningful it is that Apple has
selected Google and put it Google's LLM
into Siri. And what that means for
Google itself and what it will what it
will do. So, David Pierce from The Verge
has a great article about this. And the
headline is simple, Gemini is winning.
And I basically decided to take that
headline and make it the first segment
of our story because of our show today
because the story really is I think the
probably the best articulated version of
the fact that Google has
is in the the pole position here in the
AI race. So, here's what Pierce writes.
And we're going to get to Siri in a
minute because it builds on this. If you
want to win in AI, you have to do a
bunch of hard things simultaneously. He
says, you need to have a model that is
unquestionably one of the best on the
market.
You need the nearly infinite resources
required to continue to improve that
model and deploy it at massive scale.
You need at least one AI-based product
that lots of people use and ideally more
than one. And you need access to as much
of your users' data as you can possibly
get.
Pierce argues Google is the one company
that appears to have all the pieces
already in order. Over the last year and
even in the last few days, the company
has made moves that suggest it is ready
to be the biggest and most impactful
force in AI. All right, before we jump
into the Apple stuff, um
talk a little bit about what you think
in terms of these factors, right?
Because it does look like Google has all
these factors. And of course, the fact
that it's not reliant on Nvidia and has
its own chips
along with the monopoly of search money
has really enabled the company to to do
seemingly take this lead as Pierce
argues. What do you think? I think
especially relative to the other
hyperscalers
Google, what they've done over the last
year is incredible. I actually was had
to look up and I think we probably had
entire segments on should Google fire
Sundar a year ago. I saw there's like
Jan 2024 and Citron Google should fire
Sundar Pichai. Like when all that
conversation was happening. And now
I I like this framework. You have the
model, you have resources, you have
products that can be kind of just like
ready to use AI at scale, which is going
to we're going to talk about personal
intelligence. So, I think all the pieces
are there
um to try to, you know, like see the
other side of this.
What you said is their benefit to like
having infinite capital because of their
search monopoly
can still also be the one weak link in
this. Is it I I actually give full
credit to how Sundar and the team have
managed this transition to actually go
after their
golden goose and actually try, you know,
AI search overviews completely destroy
their existing business model. Yet,
they're pushing it and they they know
it's the future. But that's that's the
one part that they're fundamentally
having to threaten their monopolistic
business model, which isn't great.
Right. But let me put it in context. So,
to put Google Search on Apple phones
Google has to pay Apple something like
$20 billion a year.
Now, Google
>> Get Get into it. Explain the
>> Gemini into Siri and it's getting paid.
And this is really building up to like,
okay. So, So, Google has all these
elements. And now the question is, what
do you do with it? Here's what Pierce
writes. What do you do when you have all
this tech in place? You put it in front
of people and put it to work. On Monday,
Google and Apple announced that Gemini
will power the next generation Siri
that's coming this year.
Pierce writes, Apple saying this is the
best technology available is obviously a
powerful signal to the market. But even
more than that, Siri immediately becomes
one of the most popular ways people
interact with Gemini. And the deal the
deal matters even more because every
user matters. The more user activity and
data these companies can collect, the
better their models and products can be.
And
it creates a flywheel a flywheel with
search and the same will hold true with
AI.
I think these are very big points.
Hold on. Explain to me the flywheel with
search. Like we have personal context,
we have personal intelligence. But
charging advertisers to show up in your
search results, which are presented in
some kind of order as blue links. And to
be higher on that page, that is the
search business model. It's not actually
just searching for stuff. That's the
business model. So, how is it a flywheel
for that? Well, I think the Okay, the
the there's the business and then
there's the product. And so, this I
think Pierce is talking specifically
about product. Now, one interesting
thing about generative AI is that the
queries are much more varied than
search. And I think something like 1/3
of searches are brand new to Google
every day. I would argue that many more
of these questions to generative AI
models are novel. And so, I think to
become good at doing what you do, you
have to be able to really appreciate the
broad range of queries and figure out
where you're doing well and figure out
where you could get better. And so, now
Google is going to put its its um
its product in front of the billions of
Apple users who are now going to be
giving it data back. Now, maybe Apple's
going to build some sort of fly
some sort of firewall. But the fact is
that this partnership is going to make
the Gemini product better because it's
going to really understand a much
broader range of user behavior and
optimize based off it. Whereas again,
with ChatGPT
they got to fight for every single user.
They don't have the distribution of
their own products like Google does. And
they certainly don't have this Apple
deal anymore. In fact, Google is
supplanting ChatGPT
in Siri.
>> Did you Did Did you ever use ChatGPT in
the kind of like embedded Apple context,
the iOS? I think I used it when it came
out a couple of times where you had to
like
click tap twice extra to get a
half-baked ChatGPT search?
Did you ever use it or
>> Yes, I think I used it once just to feel
how painful it was to go through.
[laughter] Well, actually, no. That
wasn't That's not really true. Now, I'm
done with When I look back, I actually
was eager to use
ChatGPT within Siri. I think the pain
came from actually realizing how many
steps away it was. And and that So, I
think this if you're if you're signaling
that we don't really know what the
extent of this is going to be, maybe we
can't trust Apple to get this right,
then I'm I might be buying that. Yeah,
okay. So, couple of things on that.
That it is amazing that cuz I I had I
think I've said before, I have like a
Pixel 9, I think it is, just sitting
there. So, sometimes I like loaded it up
again and started just testing what
Gemini natively feels like. And my god,
it's so much better. And
if if Apple screws that up with Siri,
I I just don't know where this company's
going to go in the whole world of AI.
It's kind of just terrifying. But But
what you're saying earlier is
interesting to me cuz
cuz I agree from a product perspective
the amount of data and access and
context this gives Google is just like
exponential in an already exponential
company in terms of data, but still how
do they I mean Apple's going to be
paying them? That's one business line,
but like
that they no one has figured out AI
search LLM search based advertising.
>> advertising. Actually, who do you
think's going to Who would you bet does
it first? Fiji Simo over at Open AI or
the masters, the old-timers, Google and
they figure out how to actually charge
people to show up in LLM search. If you
had to take take a bet.
Oh, it's not even a question. I mean it
is the one company that's built a
multi-trillion dollar business out of
search advertising, right? I I I think
that Open AI may do it, but remember
even with the code red that they had
recently trying to fight against the
product, the business model
was put on hold. They put on the the ad
rollout on hold even though we're seeing
some of it. I I think there's there's no
doubt, you know, and we're going to talk
a little bit about the fact that
Google's going to be able to share data
across different products,
but I think there's no doubt that Google
will take some of its targeting, some of
its expertise from the other sides of it
its business. So it's not just starting
at zero
with this LLM. But actually now I'm
curious to hear your perspective,
Ranjan. Why? So we're talking obviously
about the product side of things.
Um
why have you zeroed in on the business
side of it? Oh, wait.
>> about the business side and this this
deal that's making you, you know, making
something light up in your head? Well, I
I think and that way we always debated
is it the model or the product? Now we
have is it the product or the business
model, not actually the foundation
model.
Um
I think it's cuz
like I I agree with this article.
Everything appears to be going right for
Google and maybe it's just trying to
kind of bring some nuance to it, but but
to me it was more when you said that
infinite capital that like that search
monopoly, it just reminded me that that
is going away. They recognize it's going
away. I give them full credit for
actually trying to fight for whatever's
next, but still the one of the greatest
business models in the history of
mankind is going away. And that's that
like you can execute on everything else
and you can still not survive that.
That is a very interesting perspective.
So I think we've almost come full circle
here with Google, right? It was like
Google at the beginning. Oh, Google is
slow playing generative AI. I remember
they invented it effectively with the
transformer. They had the living
breathing lambda
inside Google, never released it.
Um
and everyone said Google is too worried
about his business model
to
release this to the public. Then Open AI
of course jumps the gun. They have no
choice. Now they're like, all right,
we'll play that game. They're playing
it. They're leading it and they might be
accelerating their own cannibalization.
Yeah, I think
I mean I think I'm being just trying to
see both sides of the argument here.
They're in a very good position right
now. I recognize, but actually I am
going to take the other side. I think
Open AI figures out LLM based
advertising ahead of Google.
Really? Hear me Hear me out here.
I think when I think about
like search based advertising, Google
has been on a steady
state of degradation over the last 10
years and gotten worse and worse and
worse and worse. Meanwhile,
Facebook and Meta, God bless them, their
advertising has gotten better and better
and better and better. So like the
actual at the core of taking what are
new technologies and actually turning
that into a compelling advertising
offer, Google's still working off of
something that they built a monopoly off
15 years ago, but has been degrading.
Whereas Facebook, to their credit, has
been I mean they've been Apple tried to
kneecap them like and they actually just
made better and better ads. So I'm
taking Fiji and uh
the Open AI folks on this one.
I mean Google only grew its ad business
by like 15% off a base of 60 billion in
the last quarter, so Only. Only.
>> you'll forgive me by being a little
skeptical of this idea that their
advertising has gotten worse and they
can't figure this out. Do you One thing
I had to bring out. I was thinking about
this. This came up a couple of weeks
ago.
Like is it Sergey or Larry who's back?
Sergey. Sergey, yeah. Like I've been
seeing so many credits that like
co-founder comes back. That's what's lit
the spark and that's what's causing this
like and like you know, like kind of the
hero returning. I still like the idea
that it's Sundar's McKinsey days that
taught him the key to organizational
like transformation and reorgs and the
move to bring in DeepMind and like, you
know, like actually centralize like the
AI development across the entire It was
the most boring MBA McKinsey style reorg
that was right that has been the key to
the success. I'm That's my take. I think
we can try to get to the bottom of this.
So,
you know, it's a We have a long 2026
ahead.
Maybe maybe we'll get a chance to sit
down Sundar and talk about this. I know
at Davos next week. And folks, by the
way, if you're watching on video, I'm in
a different location. I'm in Germany. En
route to Davos,
I'm going to have a series of
conversations there and two of them are
going to be with Google AI leadership.
So Demis Hassabis is going to be back on
the podcast next week. We're going to do
a live conversation from Davos that
should air on Wednesday. And then I'm
also going to do a conversation with the
COO of DeepMind, Lila Ibrahim, which
will air
a couple weeks later. So maybe I can ask
about this McKinsey thing. Maybe I can
also ask about the advertising thing
because
Ranjan, another thing happened this week
that we should discuss which may
actually set up Google for the perfect
generative AI advertising play. Here's
I'm going to keep reading from Piers.
This is good. Let's keep going with it.
Google's other announcement this week is
an even bigger flex. It announced an
opt-in feature called personal
intelligence which connects Gemini to a
vast ocean of information Google has
about you in order to give you better
responses. Every time you ask it a
question, Gemini can now answer it by
looking at your recent searches, the
videos you watch on YouTube, your
emails, your photos, your files and
more. You really can't overstate how big
of a deal this is. Google no longer has
to ask you to give it lots of context,
hope you provide excellent and detailed
prompts every time, or build out
complicated custom instruction systems.
Google already knows a scary amount
about you and now Gemini does too. And
you know, I initially looked at this
from a
product standpoint about how oh, this
would be actually a great product to use
and I think it will. But given the
direction that our discussion has gone
today, you know who would also like to
know a scary amount about you?
Advertisers.
I think So okay, so personal
intelligence, I saw some like very nice
flashy looking demos and and we've
talked about this for a long time that
Google does Google should own this.
Google should just like destroy in this
part of it in terms of knowing the most
about you and uh
um but you know, as we're speaking, I
just I I've been running this test. I
ran this test the moment Gemini showed
up in Gmail. I always ask what was my
first email with my wife.
I just asked it right now again. And for
reference, it was 2011
January. It says based on my search over
your Gmail, the earliest email thread
with my wife was on Thursday, November
13th, 2025
and it's about one of my kids' school
events. And then I said I need my first
email ever. And then it says based on a
search, the earliest email thread I can
find is Wednesday, October 2021.
Like how do they not get that right?
This is what This is This is like the
missing link in all of this that again,
I'm not
Gemini has gotten amazing. I think
Gemini's going to supplant like
standalone Gemini a lot of ChatGPT users
and prevent them from ever going there,
but I just am baffled from a product
standpoint. That is the most
straightforward question to answer. It's
like a combination of some search and
some LLM and it can't get it right. Why
do you think that is?
Well, I just I was just about to ask
Gemini what my first email was with my
Wi-Fi
and then I realized my muscle memory is
probably
unfortunately geared to typing Wi-Fi
versus wife. I need to go to therapy for
this. [laughter]
But um
my answer I That's the true The true
companion was never ChatGPT. It was my
internet connection.
It was Time Warner and Spectrum.
>> I am a simple man. That's all I need. Uh
A wife and Wi-Fi. Did it get it right?
>> I
know. It got my wife's name right, but
it says I can't definitely confirm the
single first email in your entire
history. And then it gave me a query
to run, which is interesting. Yeah,
which is Boolean search. Like Did
Yeah.
I I'm just going to answer. I Listen, I
I think this is one of the main
limitations of AI and we spoke about a
little bit last week. It's that the
context window runs out and it can run
out fairly quickly. I mean think about
how many characters you have in your
Gmail. It the technology that it cannot
put all that in the context window.
>> [snorts]
>> So, what what we see is that like
there's now some shortcuts,
but then again, if if we're talking I
have to say
I I you know
this should not be that hard because
even now I'm and you know this too, I'm
thinking through the steps, right? Okay,
five you know, first do a query, find
out who the who the person's wife or
wife I is depending on their persuasion.
Then
you know, you can if you if you are in
love with Wi-Fi,
clearly I respect you.
Um
But then you can actually go ahead and
do do the search. So, obviously from the
product standpoint here something is
off. You're right. There's a problem and
it's one of the most inexplicable things
about Gemini to be honest.
>> But so and this is going to be a good
setup to our later conversation on
Claude co-work. I think like this is a
good example of agentic versus just some
kind of like
low-grade LLM search. Agreed. All of
your email of all time needing to be in
the context window is not going to get
you the right answer here. Coming up
with a system that create the Boolean
search query, go do it, return the
result. Should how is Gemini in Gmail
not doing that, but it's not. So, this
is more like there's still a bit of this
Sundar still got some reorg McKinsey
style work to do here. I'm just telling
you, but
if if if he's listening, but overall
very good position.
Do you think it's a safety thing? Do you
think that like there it might be,
right? They might be concerned that if
they allow it to do too much in email,
you'll get creeped out. I mean, that to
me is the best explanation. I'm I'm
striking the technology limitation
thing. It's the it has to be either
competence or safety. No, no, I it's my
call. It is actually organizational
dysfunction. It's like the Gemini in
Gmail, there's some
back and forth between the team that
owns Gmail, product managers saying I
don't want to give you access to this
and like
cuz again actually standalone Gemini as
a connector to Gmail probably would get
that right.
I I what couple quick things before we
end. I just want to talk quickly about
what Siri in the iPhone's going to look
like. This is from the information.
Siri will have the ability of to answer
factual questions, tell stories, provide
emotional support, or wow, or help
people accomplish tasks such as booking
travel. Some of the features will launch
this spring. Other including Siri's
ability to remember past conversation it
had with a customer or proactive
features that could suggest they leave
home to avoid traffic ahead of an
airport drop-off that's listed on the
Apple calendar are expected to be
announced at the company's annual
developer conference in June. Um
I just wanted to circle back on this
because we had already started the
discussion.
I don't know if I I just don't trust
Apple on this one. It is we're going
back to the same like demo and release
schedule. I mean, of course like it's
just the deal was just signed, but I
don't know if they're going to be able
to they're still trying to do Apple
intelligence. I just don't know if
they're going to be able to do it. Yeah,
I it was so baffling to me that like
it's still what is it with booking
travel and AI that everyone is so
obsessed with? Like there's so much
other stuff researching travel. You
don't no one is going to book a ticket,
let AI book the ticket for them for a
long time.
And also the universal search thing,
it's already again, it's already been
solved. Like it's you can do that with
any chatbot. You should be able to do
that. You shouldn't advertise that as
this amazing capability. So, I I
honestly I would love if Apple's entire
marketing campaign on this was
it will be baseline good as everything
else and we will deliver that to you.
And that's it. And I would be I would be
ecstatic. Like I can't wait for the the
Super Bowl ad. Siri
we're going to be sufficient eventually.
>> Siri. We're we're we'll be as good we'll
be as good as the rest. There's your
tagline.
>> [laughter]
>> Think the same. I think that's the new
Apple. Not think different. Think as
good hopefully. Um yeah, that's
literally if they did that, that
transforms the entire business. It
should not be that difficult. That's all
I'm asking. I still have my iPhone. I I
might get that iPhone fold for like
$7,000 or whatever it'll be, but they
got to fix Siri.
Yeah, you'll be folding. All right,
let's end this segment. Credit words do.
This is back back to this Verge article.
Credit words do for a company not
exactly known for its ability to focus
on coherent product strategy, Google
managed to marshal its considerable
resources in a single direction. Now, if
chatbots are in fact the future and most
of the AI industry continues to bet they
are, there's simply no other company
currently set up to truly compete with
Google. Google has the models, it has
the resources to improve them, it now
has the distribution necessary to get
people to use its bots and the data
required to make them uniquely personal
and useful. At least for now, ChatGPT
has the brand power and the daily active
users, but Google has almost everything
else even the iPhone. Are you buying
this?
I kind of am. Again, like this is where
yeah, from a device perspective, Apple
is vulnerable. We know that. Um and
Google I actually I want one thing I'm
curious. Do you think does Google they
must have some kind of pin
or AI first Johnny Ive style device
under the works as well, right? They
have to.
>> I'm definitely going to talk with Demis
about this next week because I think we
talked about this Thinking Game
documentary
or I might have talked about it earlier
this week that Google has up on YouTube.
All right, so I spoke with that about it
with MG Siegler. It's a great great
documentary, kind of a look into
DeepMind. And for like half the
documentary, it seems like they're like
pointing a phone at things and asking
questions about it. Uh
it is it's never been clearer that
there's going to be a big wearable push
within that company without a doubt.
Yeah, and and actually I mean, Meta, do
they have any
do as a non-Android
ecosystem user, I know Samsung watches
are pretty good. Are Andr- are Google is
there a Pixel Watch?
Yes.
Okay. Is it
>> I don't think it's I have well, if we're
asking that question Yeah. probably not.
I think basically having just I just
bought the Garmin, having just done a
lot of research in the smartphone
market, it's basically either I'm going
to get some angry emails for this. I
could be wrong here. This is all right,
don't hold me to this, folks. I'm just
going to talk about my cursor experience
uh testing this stuff out. It seemed to
me like the market is basically on
smartwatches
the Apple Watch or Garmin. Yeah, that's
what that's the sense I've gotten as
well, but the Pixel Fold I I look at a
lot. I've been researching and now
Google based on old-school machine
learning serves me a lot of Pixel
content in the Google app in terms of
news stories and stuff like that and
push notifications. So, so maybe I'll
get maybe I will fold. Get it. If Siri
doesn't work with Gemini. You're fold
you'll fold and get the fold. All right,
I I I want to
sorry.
>> [laughter]
>> I'm in rare rare form today. I
apologize. Um
Before you get before you get serious at
Davos, let it all out right now.
[laughter] Uh there will we will not
have seriousness there for sure. We will
we will make we will make bad jokes
in every country and continent possible.
All right, I just want to end this
segment. I already said that, but this
is just my perspective. Uh I I I feel
like all these arguments make sense on
paper,
but then you use OpenAI and you use
Anthropic's
models and their their products and
they're damn good. They really are.
And I I I I don't know.
I would
I don't think
maybe Gemini is on par in some ways, but
I would not write those companies off at
all. No, I'm not writing them off. I
think the story to as of today is that
Google is now one of the leaders. I
think and and well positioned for all
these other external factors. From a
pure
Gemini {slash} model {slash} product
standpoint, I think it they're on par. I
mean, they're certainly
on par, but
I don't think I don't think anyone's
like counting out the others.
Right.
Yeah. But I I thought with with the the
end of this Verge article was so
definitive.
Um
just that like let's see. Um there's
simply no other company set up to truly
compete with Google. I disagree with
that. But all the other factors are
correct in my opinion.
Agreed. Agreed.
Okay. Stock mirror. Yes, so we're going
to talk about Thinking Machines Lab and
the fact that there's been
the drama continues. It seems like the
Thinking Machines Lab, by the way, is an
offshoot an offshoot. It's a company
that works in AI that is founded by that
was founded by
former OpenAI executives and leaders and
there is a crisis right now going on
within that company as many people
head for the doors. We'll talk about
that and what it means along with Claude
co-work right after this.
And we're back here on B Technology
podcast Friday edition.
Uh we're having a good time today but
Thinking Machines Labs uh lab is not.
Here's the story from Wired. On
Wednesday, OpenAI's CEO of applications,
Fidji Simo, announced the company had
rehired Barrett Zoph and Luke Metz,
co-founders of Mira Murati's AI lab
Thinking Machines Lab. Zoph and Metz
left OpenAI in late 2024.
This is where it gets pretty intense. A
source with direct knowledge says that
Thinking Machines leadership believes
Zoph engaged in an incident of serious
misconduct while at the company last
year.
That incident broke Murati's trust, the
source says, and disrupted the pair's
working relationship. Here's an aside
from me. Uh it's it sounds a lot like
that source is a somebody on the PR
side. It's just like straight PR speak.
Okay.
The source also alleged Murati fired
Zoph on Wednesday before knowing he was
going to OpenAI due to what the company
claimed were issues that arose after the
alleged misconduct. All right, I'm going
to speak as a reporter now.
The to me look, I I I I don't love to
guess at sources but it is fun.
Um
and
it seems and I could be wrong here, but
it seems like the source on this one was
either one of two people uh
or one of three people.
Either Mira Murati, John Schulman, her
co-founder, or PR. That's it. This is
this is clear like
it seems like clear you know, spin to
me. So, that's the story.
>> spin, like what is this serious
misconduct or breach of trust? I mean, I
guess I know they can't say it, but like
>> That's why I think it's spin because if
it was a like somebody else that wasn't
like giving the official Thinking
Machines message, we might know.
Wait, sorry. Walk me through that. Walk
me through that. All right. You get you
you get you're going to get uh
this is how it usually works. You get a
call from typically it's a PR person.
You know, you've reached out for
comment. You might have heard that these
people leave. Then they're going to give
you their perspective.
Right? And if it was like somebody
within the company or or a couple people
that really knew what happened with this
alleged misconduct, they will tell you.
I mean, maybe so they Zoph didn't
respond to several requests from Wired.
Um so, maybe they just didn't have it
concrete enough and sort of might have
been exposed to legal liability.
Uh but typically when it's not specified
like that and it's a reporter over what
happened,
um
then you know, it's typically coming
from the company itself. You would
imagine otherwise, you would actually
know what happened. It's a very it's a
very That's why there's so much drama
here. It's a very weird situation. Okay.
Which
raising $2 billion at a $12 billion
valuation with no product or even
business plan or any mention of what
your business does was always kind of a
weird situation if we think about it.
So,
it maybe this is just kind of the
logical
conclusion of that. Like do do you think
do you think Thinking Machines is done?
Do you think this is
just
a a hiccup in the road and
Mira and team come back stronger than
ever?
I I think they're done. I do. Uh
the other the the one argument I'll make
against me is that they have billions of
dollars. They raised $2 billion or
something like that.
>> Yeah. That's a lot of money.
>> Pre-seed. So, pre-pre-seed or seed,
whatever we called that.
>> Yeah, you would skip a few rounds,
right? Like seed seed used to be
$500,000. So, clearly they're beyond
that. But I think they're I just want to
say why I think they're done. So, you
have these two that have left and you
have more. There's another one, Sam
Schoenholtz, who's going back to OpenAI.
And this is the from the Wired story at
which is very well reported. At least
two more Thinking Machine employees are
expected to join OpenAI in coming weeks.
Like not only are they leaving, they're
going back to the place they broke off
from.
Yeah, that part is the oddest part to
me. I not odd, but and and Fidji Simo
kind of like uh just pouring salt into
the wound by saying like tweeting about
reporting structure and welcoming these
folks back as well was just it's all
just drama. I think to me the bigger
question is
is this the end of this absurd
$2 billion seed round safe super
intelligence Thinking Machines if you
were an ex-OpenAI person, you will get
billions of dollars without even saying
a pitch deck. Is this the end of all
that?
Definitely not. I mean, if you think
about Just think about where AI is today
compared to where it was then. Oh, no.
>> could both we could both agree that the
AI the promise of AI is more real now
than it was in 2023 and 2024.
>> When we get to our final segment, you're
right. Yeah, I mean, it is it is. It's
going to So, basically if someone has an
idea, then they're going to get funded.
Yeah, but the promise of like real AI
doing real things is greater than ever.
The promise of vague notions about like
end of society and humanity and safety
and thinking whatever. Like I feel maybe
in a way
maybe this is the like nail in the
coffin of that era of like in which I
would love and this is the best thing
ever where we stop talking about just
really vague things and more theoretical
things and actually are able to just
focus on work.
Don't you think that as it gets
as the valuations and the economic
promise gets more concrete,
that some of the silliness on the
margins is just going to get even worse?
No, because then you actually
it's like the famous Silicon Valley HBO
episode where it's like you never want
to have revenue cuz then people will
actually value on that revenue.
Yeah, when you actually are selling a
product, then you are measured on the
actual money you're making and the
multiples of your revenue as opposed to
when you got no product. When you First
of all, when you have no revenue, they
can't do that. And when you don't have a
product, it's even better.
Yeah, I mean, maybe now we have more
realistic views about what AGI is and
what the path there is. But anyway,
uh it's going to be it looks like bad
news for for Thinking Machines. Uh go
ahead and answer. I want I wanted to end
this one with my favorite thing I saw on
this was that this it introduced is this
vibe founding.
Just kind of getting out there. You
don't really do anything. You just raise
a bunch of money and
you that's it.
Vibe founding. But I think what these
VCs were probably betting on was that
they were going to get the next OpenAI.
But doesn't seem like anyone outside of
OpenAI is going to be the next OpenAI.
This is from Alex Heath. As I
um
he says the more Okay. More Thinking
Machines employees are in talks to join
the three founding members who just
rejoined OpenAI. Which okay, Wired had
that. But this is the new detail.
Sources say Thinking Machine lacks a
clear product and business strategy and
has been struggling to raise a new round
of financing. All right, Ranjan, maybe
that last detail
is uh some evidence that
you're right, that we might be past the
silly era. But do do you know what is
interesting is
what if let's say
the the guy who left was leaking
information that was confidential to
OpenAI. I just want to know what that
information was. If the Is there a
product on the horizon [laughter]
that they're like going to let them
steal or I would like to I would love to
know.
>> Well, I mean, Fidji apparently in her
post said she doesn't have concerns
about the ethics. Meanwhile, we know
nothing about Thinking Machines. Right?
We we still don't really know what
they're doing.
OpenAI. Maybe they don't either.
>> [laughter]
>> Okay. I think we should talk about
Speaking of getting real work done.
Yeah.
>> Yeah, let's talk about uh Claude
Co-work. Last week you and I had a
discussion, Ranjan, about Claude Code,
about how this was the type of
technology that could get work done for
knowledge workers, whether they know
code or not. Um this week uh basically
ever you know, almost immediately after
we had that conversation, this dropped.
This is from from Wired. They they
released something called Claude
Co-work, which is an AI agent that's not
for technical people. Um and or that you
don't need to be technical to use it.
And Wired says it actually works. Uh
they say uh Co-work takes the abilities
available in the company's coding
focused tool and makes the user
experience more approachable. The tool
is designed for the wider group of
non-technical users who may want to
experiment with a new way of controlling
their computers but get freaked out by a
command line. It can do things like
organize files into folders, generate
reports, and even take over the browser
to search the web or tidy up a Gmail
inbox. When it comes to file management
and computer interfaces, this tool feels
like the start of a pleasant user
experience evolution.
What do you think about this? I mean,
this is basically is the thing that we
discussed on Friday and now it's come to
life.
What we have discussed and and for for
newer listeners, just for context, I
work at Writer, writer.com, and I've
been speaking for months now about one
of our products, Writer Agent. We're
only enterprise focused, but basically
it is this. It's using natural language
to take multi to launch multi-step
workflows. You give it the harness,
which is the word that we have not used
prior to last week, but I had said you
give it the tools and the data sets and
the connectors, and the AI goes and does
the work. And to me, I actually was
trying to find I think it was like 4
months ago I said it was the first time
I felt AGI, what I think could be AGI,
that it actually is going out and doing
stuff. I'm turning off my computer and
in the background it's working for me
and doing stuff. And dude, I've been
trying to tell you this is for for weeks
now or months. Like I have felt this. So
it it was actually kind of exciting for
me to see Claude get into this space.
Because Claude Code has done this for
software development. And I'm not saying
that just only for software developers
cuz plenty of non-developers have been
building on it. But that feeling of
actual autonomous work being done has
been kind of checked off for code build
coding. But I've already seen that with
knowledge work and and now what is
actually I've already seen the last week
it's so much easier to explain what I'm
trying to cuz co-work is kind of
defining this category. So so again, the
idea of taking documents and words and
web search and spreadsheets and Python
scripts and putting it all together to
actually get to do something for you for
work. I think like this was my
prediction end of December. I think this
is going to be one of the biggest trends
of the year. It's self-serving for my
own work, but I really think it's we're
going to see this inflection point and
and I love that Claude literally after
we talked about it Alex texted me. He's
like, "Did you see this?"
>> They did it.
>> They did it. They listened to us. They
said it was only built in a week and a
half. They listened to us.
Right. Well, we you initially I said
this was this was inspired by us and you
initially mocked me by saying it was
done in a day and a half. But then I
found that detail that it was built in a
week and a half. Yeah. Which is okay. It
wasn't from last week's show, but still.
Uh but I want to ask you about this.
Okay. So Claude Code, we know what
Claude Code does, right? It's like five
coding. You tell it to do things it will
go build it for you etc. etc. Co-work.
Uh all right, here was the demo of the
debut. The AI put a uh meeting meeting
transcripts um took meeting transcripts
in and it gave advice about how to be
more effective in meetings.
Uh it had a tool that searched
calendars. It prepared custom slide
decks. Um some of the reviewers that I
read also said it was nice for file
organization. Uh
Why do you need a separate app for this?
>> So so okay. Why why can't you just do it
all in the Claude window? I got bashed
by the way. Not bashed. We we got some
polite criticism of uh my I I was just
trying to tease out last week why you
would want to do something with Claude
Code uh versus in the chatbot window.
But for this one I legitimately don't
understand why you wouldn't just do it
in the chatbot window. I mean I I I can
tell you like from a competitive
standpoint for Writer Agent, the product
of my company, like we've been testing
it, looking at it. I am genuinely like
the file local file access and
management side of it is a bit
interesting and confusing to me cuz like
who uses files? I don't know. Like I
software developers do. Um which is why
I think it was kind of built like that.
Because otherwise like I don't know. Do
do I don't know anyone who organizes
local files in folders and everyone just
has a downloads folder and a documents
folder that are unwieldy and maybe if
Claude Code Claude Co-work helps them
with that. I do think if it's like kind
of their MVP maybe it was just something
differentiated and interesting to start
with, but but again, I think the idea of
non-technical people being able to use
natural language to build stuff that
does stuff for them
that's 2026 and and that that people are
going to start seeing the promise of
that this year and and co-work is
definitely going to be a starting point
for it. But if it was built in a week
and a half, you know, I mean
how good is it really going to be at
launch?
>> So I'll say I'm not fully bought in on
the co-work
uh use case, but I could be wrong. I
mean this it starts most of these things
start as seeming trivial and then right?
Like if even if you think about text it
starts as auto complete and then it just
starts to do it on its own. So maybe we
see a similar evolution. But you know,
before we go, I want to talk with you
about this idea. I just wrote about it
in Big Technology uh
about the way that this technology
adoption is going. And it seems to me
like there's two different sort of
diverging trajectories of AI adoption.
And the first is at the organization
level. So even if companies are eager to
implement AI um they've run into all
these problems. They have entrenched
organizational habits like we talked
about security considerations and then
the technology's limits when being
applied at a wide scale. So that's one
direction this is going. And because of
that we've seen slow adoption among
enterprises from a top-down perspective.
Then on the other side you have
individuals. Now individuals they aren't
encumbered by many of these same issues
that the enterprises are. They have a
unique set of data permissions. They can
change their own habits and they don't
really need to ask permission to do so
um and they also like when they come up
against the technology's limitations
they're able to adapt. And so we have
instead of like this technology be being
adopted by enterprises and pushed down,
it's being adopted by individuals in
some way and pushed up. Um
and I I'm calling it I called it in my
post the we're in this age of individual
empowerment where you know, you're going
to start to see a real divergence in
terms of individual performance probably
more than we've ever seen it happen
before and organizations begun to be
pressed by the people that are using
these tools.
Um and and and it's going to be maybe
it'll be a little bit weird on or
uneasy, you know, in organizations that
have basically failed to roll out their
pilots, but they have people coming from
the bottom up who have used this
technology very effectively and are sort
of changing the game. What's your
perspective on that thesis?
Um I and I I've this is like my life on
a day-to-day basis. So I think
you are correct in that
it's going to be people who understand
the it's people who are going to be
playing with Claude Co-work will make my
life even easier cuz when they open
Writer Agent and we're talking about
their company adopting it, they already
know what to do. In fact, like Claude
Code already kind of taught that way of
working. I think like we've already seen
it for a long time. I
everyone is using chat GPT now or Gemini
that I know. That was actually another
maybe 2025 the two big trends for coding
autonomously got solved and everyone
using LLM chat got solved. Again, this
year people like everyone curious and
forward-looking is going to start doing
this kind of autonomous workflow
building. It What I always think about
is like I think a year and a half ago, 2
years ago, you know the saying, "It's
not AI that will take your job, it's
someone who uses AI will take your job."
And people say that and try to make it
sound like it's original and insightful,
but I think this year we're going to
start to really really see divergence in
that like it's just so much more
powerful if you figure out how to get
AI to do work for you that you will just
be leaps and bounds ahead of your
co-workers. It's like I see it
firsthand. I feel it firsthand. So like
I think this year is going to I don't
know problems what what is going to
happen exactly in that, but it's going
to happen this year.
Let me ask you this. So in the beginning
of this like workplace rollout, you had
a lot of companies who like saw chat GPT
and they said,
"We need chat GPT for what we do." And
they threw it at a lot of different
problems.
Uh they threw it at, you know, uh
adding efficiency in processes, drawing
insights from data, automating some
tests and they thought they could do
that I basically I'm curious, is that
going to just disappear and is it just
going to dissolve into like individuals
using these these tools? Or what or is
what you're saying that basically when
when you become a personal user, you
could sort of become a champion within
the organization and encourage other
people
to use the enterprise tool?
>> Yeah, even we see it like I mean that's
like our tool is built around
collaboration on this. It's like but
it's still individuals building things
that then get shared out and are used by
others. So it's still kind of like those
champions or leaders within a group. And
it's always like an interesting who you
see kind of rise up and do that. And and
it's typically not pure tech people,
pure busi it's some hybrid of the two.
Um it's just kind of curious people I
guess. Like I I mean I even saw like
like
it's funny like who I've seen vibe
coding Claude Code apps. Even like Joe
Weisenthal and Kevin Like you see like
people who just you
have seen are curious and that's the
kind of person I think that is going to
get it this year and uh and start kind
of and and it'll force people cuz also
people if you're at an organization that
doesn't adopt it you don't want to work
there and it's going to be the most
talented people. So I think that all
that kind of cultural clashing will
really come to a head this year.
You know, I'm calling these these early
adopters I'm calling them the harness
hive. Harness.
Harness harness
Let's work on that one. Let's Can't
improve on that. That's good.
Maybe
They build the harnesses for the teams
and the organizations. They They sew
together them. They craft them in with
care. They're the harness builders more
No, it has to have alliteration. Harness
Hive. All right. That's what I'm going
to call our listeners from now on given
the dramatic and effusive pro hot pro
harness feedback I got after last week.
>> we're I'm thank you to our listeners for
for being team harness over here cuz
God knows Alex isn't but But we will we
will prove him to say I can say Harness
Hive. You can't even say it. I can't say
Harness Hive though. That's too far.
That's too [laughter] far.
I'll make a folding phone joke before
pun before I I say Harness Hive here.
Maybe this might be the universe
encouraging me to log off for the
weekend. So,
Ranjan, great to have you on as always.
Really appreciate it.
>> Bring bring all of this to Davos. That's
all I ask. It's a
How many times can you say harness with
Demis?
next week
>> I'm not saying [music] it but I'll say
it one more time as we sign off. All
right, Harness Hive. Thanks for joining
us as always. Great having you. We
appreciate your time and companionship
with us every week. So, thank you for
being here.
Uh next week, very intense week of Big
Technology Podcast shows. If you're at
Davos, please do come say hi at the
Qualcomm House and [music] at the Google
House on Tuesday night with
with Demis.
Uh but here's here's the rundown. We're
going to have Cristiano Amon
on Monday, the CEO of Qualcomm. That
will run probably on Tuesday.
Then I'll be interviewing
Brett Taylor, [music] CEO of Sierra,
Demis Hassabis, CEO of Google DeepMind,
Joel Pinyao, the Chief AI Officer of
Cohere, and then again the COO of Google
DeepMind on Thursday, Lila Ibrahim.
[music] We're going to put at least two
of those on the feed on
uh next week and then Ranjan and I will
be back on Friday and then the [music]
rest will come out in the weeks
following. Thank you, Ranjan. Thank you,
everybody for listening and watching and
we'll see you next time [music]
on Big Technology Podcast. Harness Hive
out.