The Unlikely Anthropic & SpaceX Marriage, OpenAI Trial Revelations, AI Layoffs Or Cope?

Channel: Alex Kantrowitz

Published at: 2026-05-11

YouTube video id: 38Q4mhLQEv8

Source: https://www.youtube.com/watch?v=38Q4mhLQEv8

Anthropic and SpaceX partner up in an
epic collaboration that could change the
AI race where a wash in revelations from
the open AI trial and are AI layoffs
just cope? That's coming up on a Big
Technology Podcast Friday edition right
after this. Welcome to Big Technology
Podcast Friday edition where we break
down the news in our traditional
coolheaded and nuanced format. Well,
it's tough to pick what to lead with
this week because we have lots of big
news about the AI race, including news
that Anthropic and SpaceX will partner
up, where Anthropic will use SpaceX's
capacity. We also have lots of stunning
revelations from the Open AI Musk trial
and a number of big layoffs happening in
the tech industry. Are they actually AI
displacing people or just cope? We'll
talk about it all. Joining us as always
on Fridays to do it is Ranjan Roy of
Margins. Ranjan, welcome.
Man, those text messages. I cannot wait
to get to that. But I think certainly
space compute's always a good place to
start as well. I was tempted, very
tempted to begin today's show with a
dramatic reading of the text messages
between Sam Alman and former OpenAI CTO
and former OpenAI Intram CEO Mirror
Morati that came out in court this week.
Don't worry listeners and viewers,
Ranjan and I will be doing that. We will
be giving you that dramatic reading in
the middle of this show. Um, but I
resisted the temptation because I
believe this anthropic and SpaceX deal
uh is I think about some of the biggest
news you will get in the AI race this
year and we will take you right there.
So, here's the story from the Wall
Street Journal. Anthropic Inc. deal to
use all of SpaceX's Colossus 1 compute
capacity. Anthropic will use all the
computing capacity from SpaceX's classis
one data center in a new agreement with
Elon Musk's rockets and artificial
intelligence company. Uh SpaceX will
supply 300 megawatts of new computing
capacity using more than 220,000 NVIDIA
GPUs by the end of the month. Anthropic
said Wednesday. Anthropic also expressed
interest in a partnership with SpaceX to
build AI data centers in space. Long one
of ma Musk's top priorities. Uh the
agreement comes as Anthropic has
struggled to gather enough computing
power to fuel enormous demand for its AI
models. Um I don't even know what
adjective to use here. Stunning,
shocking, mind-blowing. I mean Elon Musk
has spent the better part of I don't
know a year longer um railing on Daario
talking about how uh Anthropic is
actually misanthropic
and now uh he is renting them
effectively an entire data center uh
worth of GPUs to help them pursue their
goals. What's your reaction here?
>> Yeah. What is the adjective to describe
this deal?
I don't know. I I I agree. I'm having a
hard time. I think the problem with so
many of these deals, and I'm putting
that in quotation marks, is they're over
a longer period of time, they start to
provide access, but everything with
Oracle and Open AI, everything like
these giant compute deals all feel as
much signaling as reality in terms of
kind of delivering compute. Yes,
Anthropic is in dire need of compute and
rate limiting and all this has been one
of the big issues that's actually been
uh you know hurting them recently.
The idea that Elon Musk and SpaceX are
going to kind of come to the rescue of
Anthropic is just it's both shocking and
mind-blowing, but it's also as SpaceX is
looking to IPO, having this massive new
revenue source where you basically
become like what is the fastest growing
company in the world? Who are they
completely dependent on? Us. So I think
I see that side of it but like do you
think one year from now that SpaceX will
be a major provider of compute to
Anthropic?
>> Yes. Yes. And I I you know I am stunned
to hear you downplay this news here. Let
me tell you why I think that this is
significant. First of all, this is all
but an admission by SpaceX which of
course is the company above open uh
sorry XAI which Elon Musk basically Elon
Musk merged X into XAI and merged XAI
into SpaceX. So SpaceX is the company
that holds all of them. There was a time
not that long ago where XAI was
considered one of the top players or at
least one with a great shot. remember it
had used this capacity to develop
benchmark breaking uh AI models and here
we are it's in May 2026 it this is
effectively it giving up it's saying you
know it doesn't know how to effectively
use this compute capacity to train
models or if it did it wouldn't be worth
the expense because it would still be
behind OpenAI uh and Anthropic and so
therefore it's now turned its business
to renting data centers right it's now
the new core weeave as opposed to the
new open AI. So I think that is a
fascinating development here. Um and I
also think that yes it is um if you
think about it from what where the race
stands this will have a meaningful
impact because the way that this race
has been setting up you've had open AI
which uh has had more compute um and has
sort of now following anthropic in this
you know sort of code model with codeex
uh as anthropic has cloud code and we're
going to get into the demand about cloud
code in a moment it's crazy um and then
you had you had anthropic And the big
thing holding anthropic back was it was
unable to deliver the models because of
capacity constraints. And so now you
have Elon and Daario teaming up to
effectively take on Open AI here, which
is fascinating.
I I I know I shouldn't be this not
cynical about these announcements. I
think what is interesting about these
kind of like big types of alliances and
announcements is when they're so clearly
beneficial to both parties in their own
specific way even when it's uh you know
so clearly like it makes for a great
story. Elon and Daario have been at each
other's throats. Is this actually kind
of teaming up against OpenAI in this
case? And maybe it is and that would be
make the most sense to me. But but still
Anthropic needs to come up with a story
around how it will deliver compute as it
moves towards an IPO. SpaceX needs to
deliver a greater narrative around its
revenue streams going forward as it
moves towards its IPO. And it just feels
like kind of signing on the dotted line
here. It was a very very convenient way
of pushing that. I know. I know. It's uh
I mean there's been too many of these
announcements over too long, especially
with with an Elon Musk involved that I
don't know. I'm going to I like to wait
beyond the announcement for stories like
this.
>> Look, as a reporter, I really appreciate
the the skepticism and wait till the
announcement passes. And you know, maybe
that's my one weakness is that well, not
one, one of many uh is that um no other
>> no that one of many, I'll admit it, is
that uh sometimes I'm too quick to to
believe these things, but we do have the
evidence right away that uh this is
going to help Anthropic. Um here's what
Anthropic says. First, we're doubling
Claude Claude Code's 5hour rate limits
for Pro, Max, Team, and Seekbased
Enterprise plans. Second, we're removing
the peak hours limit reduction on cloud
code for pro and max accounts. Third,
we're raising our API limits
considerably for cloud opus models uh as
shown in the table below. And of course,
they put a table. We won't we won't make
you suffer through reading that. But
this is instant impact. I just crunched
the numbers. The data center that they
are going to be able to lease from Elon
Musk is a 5 to6 billion data center
that's fully built, right? That's just
from the GPU cost alone. It's full fully
built and ready to work to help
Anthropic deliver these services. That
was the company's that was the company's
one weakness, right? Is that it could
not deliver the compute and therefore
OpenAI had this chance to pass it
because it does have that capacity. This
changes that significantly. That's what
I'm I'm saying. And we can see the proof
in the rate limits.
>> No, no, no. You're right. You're right.
the rate limits again like I still saw
plenty of complaints on Twitter around
uh like especially the weekly resets
around rate limits with cloud code and
the kind of lack of uh or the opacity
around that I think but you're right
this is direct they're at least kind of
correlating this action directly with
that kind of increase in compute but I
don't know I still think this is one of
the most interesting parts you know as
the kind of market rises again today and
the people I've been talking about like
and again working in the industry and
seeing where this technology is going to
go and believing this is going to
transform enterprises and most knowledge
work as well but it's still more
everyone from a Google to a meta to open
AI to anthropic to now SpaceX as the
core vehicle for Elon Musk outside of
Tesla like everyone has the same story
to tell there's going to be this
unlimited unsatiable insatiable appetite
for compute and anyone who can go to
fill it and then there's same customers
buying up the compute providing the
compute it's such a small circle of
companies that all benefit from these
kind of transactions and this story
having to move in this one direction
that I don't know I think I've become
definitely more skeptical in the last
two week two to three weeks around even
though I'm watching firsthand how
valuable this technology is whether this
story actually will play out in the way
everyone wants it to.
>> Wow. Interesting. I mean, I I think that
if you look at the capacity demands, how
can you really argue with that? Well, I
think the over the last six to eight
months, everyone had a free license to
do whatever they wanted, especially in
the enterprise and it was everyone's
first kind of foray into claude code and
claude code and you know autonomous
knowledge work products like writer and
we talked about this since last November
and October anyone could do what they
wanted. No one questioned anything
around tokens. So the actual like the
graph for that usage which was
essentially zero because the token
consumption pre-reasoning and especially
pre-agentic knowledge work or agentic
coding was so minimal relative to what
agent the token consumption of agentic
work. So the graph was has been insane
over the last 6 to8 months but that's
because no one cared about anything you
could spend whatever you wanted. If you
are to ex you now everyone is
extrapolating that level of growth
rather than will there be a cheaper
model? Will there be cheaper ways of
approaching this? Will companies
actually kind of like bring token
consumption instead of token maxing
people actually be like optimizing for
tokens? I think like that's the the the
story is over the last eight months
which I've watched firsthand. No one
cared about anything. We literally came
up with token maxing the industry to say
more is better and now everyone assumes
we're going to grow like that. So it it
really even though I believe strongly
that AI will kind of like have a massive
impact on the world and certainly the
enterprise I don't know all these
compute stories really push around let's
just blindly extrapolate over from the
last 6 to 8 months. Well, okay. So,
there's two sides of this here, right?
There is the demand for compute and then
there's also the model makers are going
to make their comput their models more
efficient and they have been making
their models more efficient and they
will be making it more efficient, but
you'll still have the demand here. Just
one example. Um, everyone who's used
Anthropics Opus 4.7 model uh has has
certainly experienced some frustration
where they try to get it to do some
simple thing and it burns like half a
data center's worth of tokens. uh only
to come back and be like, "Oh, shoot.
This was a simple operation and I tried
to like do advanced calculus to figure
it out." Um just one example, I asked it
to make a PDF for me yesterday. Uh it
spent like 30 minutes trying to find
this PDF and I'm like sending commands.
I'm like, you you created the PDF, just
export it. Just export it. And it's
like, uh plugging into tools, every tool
in the book. And then it finally comes
back to me and it says, I owe you an
apology. I went down a rabbit hole
worrying about a constraint that wasn't
actually blocking us. the files there.
Send me the file.
>> So, they're gonna fix this stuff,
though. That's what I'm saying.
>> This is what I'm talking about, though.
And again, we have talked about I I've
thrown this out in the past that kind of
like nudging systems towards greater
levels of consumption. Do you remember
what was OpenAI's product that would
work all night for you to give you a
news update in the morning that no one
pulse?
>> You talked about this. That's right.
No, this is this was my theory that like
nudging people towards greater levels of
consumption was actually going to be
part of things and and again like what
you're talking about either one of two
things can happen either anthropic will
fix that and actually like make it work
in a much more efficient manner which
will be good but also changes the
compute story a little bit as a direct
direct extrapolation or users will kind
of fix that themselves because they
don't want to pay for those tokens. So,
I think that mindset shift of do I watch
what I'm spending on an AI platform
because it's no longer blindly $20 a
month or even $200 a month. So, I think
that's going to be a big big shift.
>> Well, this gets back to the whole Jebans
paradox conversation and of course, if
you've been paying attention to it, the
argument in AI industry is, you know, as
a valuable good gets cheaper, you'll
actually use more of it versus less
because it will enable you to do more
things. And I would argue, let me just
throw this out there for the sake of
argument, and you can decide to rebut it
or agree with it, is that we've seen so
much model improvement over the past,
let's say, year and a half, that those
who might have been skeptical in the
Deepseek days, right, where this Jevans
paradox thing, well, it's just going to
do the same thing and it'll be cheaper,
so no one's going to spend more on AI.
Um I I think this Jevan paradox uh idea
is Jevans paradox idea is more plausible
than ever that if they bring the cost
down you'll be able to do much more and
you might have like let's say more
agents running in parallel. You
disagree?
>> No I I I agree. I agree. But I think
what that says about the kind of is it
an exponential demand for compute or
more of a linear one. I don't think it's
necessarily exponential in the way
everyone is looking at it right now and
the way these deals are being done and
the way people are chasing it and again
when it benefits both parties involved I
do think that it can be more of a linear
path than exponential one again when
everyone has such a vested interest in
it playing out in a specific way it
moves away from and when it's a very
small insular group I do think every I
don't know it just moves it away from uh
necessarily being the way it's going to
play out.
>> Okay. Now if you're saying all right
maybe there's a percentage of this that
is gamification and inefficient models I
would I would agree with you for sure.
Obviously that's the case. The question
is if how much of that gamification and
token maxing would you say accounts for
the growth in the industry and we're
seeing a lot of growth and uh Anthropic
had its cla code with Claude event this
week uh in San Francisco and Daario
Amade the chief uh executive of
Anthropic was on stage uh and here is
what he said according to the New York
Times. Dario Amade the chief executive
of Anthropic said on Wednesday that his
artificial intelligence company had
planned to for growing about 10 times as
big this year only to reach a growth
rate that could make it 80 times as big
this year instead. He said last month uh
Anthropic oh he said let's see um
Anthropic had been overwhelmed by the
rate of grow rate of growth and has
incre that's increased the company's
need for computing power to deliver its
AI products for customers. I hope this
is from Dario. I hope that 80 times
growth doesn't continue because that's
just crazy and it's too hard to handle.
I'm hoping for some more normal numbers.
That can't all be gamification.
No, I it's not gamification. But 80
times, is it 60? Is it 40? Is it And
that's again when claude code was kind
of in a league of its own and there was
no competition. I think like when you're
saying gamification or like yeah gaming
the system basically I had two separate
instances I think I'd spoken about this
recently like where like very senior
executives I would kind of had casual
conversations with in passing bragged
about their level of spend and it was
definitely on the like more in the CTO
CIO side but people are bragging about
how much they're spending because it's
kind of become a badge of honor to say
I'm spending an ungodly amount of money.
Now, no one is talking about their cloud
bill and puffing out their chest
anymore. Like, yeah, do you know what I
just spent on Azure last month? like uh
so it's when you're having that kind of
like you know when you look back and
you're going to th those are those kind
of moments I've been experiencing more
and more where you're like that was a
clear sign that no one should know no
technology executive should be bragging
about them how much they're spending and
not talking about what they're actually
doing when it comes to something like
this
>> this is from the financial times uh
anthropics annualized revenue which
extrap calculates fullear revenue based
on recent weeks is expected to cross $45
billion imminently, a five-fold increase
from 9 billion at the end of last year.
Now, Ranjan, I know you have your
feelings about annualized revenue, but
again, let's just go back to it. If this
was lighting money on fire broadly, um
that would be one thing, but it's it's
not. I mean, you can't can't have 45
billion of annualized uh revenue and and
you know, that sort of just being smoke
and mirrors.
>> You can't have 45 billion of actual
annualized revenue, but you can
certainly have a couple hundred mil of
uh just like budgets being thrown at AI.
I mean,
>> right,
>> what is the difference? I mean, god,
come on, Financial Times. I would have I
would have looked to you to at least not
get caught up in the ARRI. But actually,
I mean, if I guess the FT uh like even
the way they said that, they didn't
actually clearly define it. They said,
wait, how did how do how did it read
again? A few weeks.
>> Yes. Based on recent weeks,
>> the recent weeks, week, month. Um but
yeah, I mean I think do you what do you
think is happening and what do you think
will happen and should we extrapolate
based on what's been happening?
>> Well, okay, I I don't want to lose sight
of the bigger picture here because I
think I brought this in to really talk
about how this added compute and this
teaming up of SpaceX and Anthropic is
going to be answer one of Anthropic's
biggest problems and has made the AI
race much spicier, right? Because now
you have anthropic and its key models
against OpenAI and they're both less
constrained by compute than they were
previously. I mean, OpenAI was less
constrained and now Anthropic is also
less constrained. So, this is going to
be a heavyweight battle. I do see them
as being the likely winners. I do think
they'll be able to get the models to act
more efficiently. I mean, goodness
gracious, after my experience with 47, I
hope I hope Anthropic does. Um and so
therefore this is this is a we've seen a
consolidation and a solidifying uh of
this race and yeah I mean are there
going to be bumps in demand? Sure.
Right. I mean you you working in
enterprise AI I'm doing a lot of
reporting on it. The number one thing I
hear is that a lot of this is just
boards telling CEOs you need an AI
strategy and them spending money right
without really figuring out what the
problem is or diagnosing how they how
they would solve it. So yeah, there's
part part of that does exist. Um, but I
think the broader story is can't really
deny I mean neither of us will deny the
progress that these models are making
and the progress that to go to your
point the industry is making in sort of
building the appropriate uh scaffolding
or harness or orchestration layers
around it uh to make it useful. And so
um and so this goes from I mean this
goes from this goes we still have notes
of is this a bubble? There are probably
parts of it that are a bubble. We still
have notes of it are of is this going to
destroy certain companies or industries.
You know there's a good chance there.
But I think ultimately it it has now
transformed into a lot of this is
working despite people uh people's
skepticism and it is going to be this
heavyweight business story between two
companies with a tremendous history and
and Elon Musk now stepping in and saying
if I can't win I'd like the uh en the
enemy of my enemy to win. Actually, I
want to get into the Elon Musk angle
about this in a moment, but
do do you know what's going on with
Stargate right now?
>> Oh, yeah. MG Seagler and I just just
spoke about that on Monday. Um, so
Stargate has basically become this sort
of catch-all phrase for OpenAI's uh
compute efforts and it has shifted from
them effectively building their own
compute to securing it through partners
like like a corewave.
>> Wait, hold on. But the Oracle Soft Bank,
what was announced when was that?
Probably like a year and a half ago.
January, February, that they're walking
away from, right?
It's unclear. I mean, as with many
things in this infrastructure build,
>> no, but that that that's so that captur
that captures everything perfectly. big
announcement. And in that case,
obviously, like I mean, when you think
about everyone who was up there from a
Larry Ellison standpoint, from a Masa
Sun, from a I mean, you got the all-star
crew of like, you know, people who can
certainly sell things well when it
benefits their own companies. Um, and
then just it's what is it 14, 15 months
later, it's just kind of fizzling. It's
unclear what's happening. There's
certainly groundbreaking and things
being built, but there's still an
ongoing discussion around compute
shortage, which that's theoretically
should have kind of really solved and
put us in a position even within 18
months to actually be very strongly
positioned for. So, I think that one
that that's kind of what I see happening
with this announcement.
>> But this is already built. That wasn't
built. Colossus one is built.
>> Yeah. Yeah. No, no, I know. But I don't
know. I like I don't know how it will
happen.
>> Han has been burned too many times.
>> I don't I I'm going to wait. Is
>> that fair enough?
>> I until my claw rate limit is quadrupled
or quintupled then maybe. But even hold
on. I don't know. like you included this
tweet in our prep doc to me
>> on the Elon side where
>> we reserve the so just as SpaceX
launches hundreds of satellites for
competitors with fair terms and pricing
we will provide compute to AI companies
that are taking the right steps to
ensure it is good for humanity we
reserve the right to reclaim the compute
if their AI engages in actions that harm
humanity
first of all I love like clearly the
entire industry got The memo went after
what what's Sam saying now? Uh, magic at
hypers scale is what open AI is about.
Elon is talking about achieve a great
future with amazing abundance for all.
But do you think this is going to be his
out in like two weeks over anthropic?
>> I mean it's possible. But I think don't
don't aren't you missing the forest for
the trees here which is um and I I noted
this Don also. I mean,
you could have not found someone that
dislikes someone more than Elon Musk
disliking Dario. He's constantly
tweeting how it's misanthropic and way
too woke, right? And he's on that whole
whole bandwagon. And Anthropic has
performed so well and become such a
successful company going after the same
market as SpaceX. SpaceX has now become
an enabling technology for anthropic.
Doesn't matter how long it lasts. That
in in of itself is remarkable. It's the
ultimate revenge for Daario.
>> I It is. But do you know what I've been
thinking about is I really believe a lot
of anthropics power has also come from X
and like I mean even within the
industry, everyone you talk to like
everyone talks about like seeing the
power of claude and cloud code and the
use on X. It's LinkedIn. Nah, it's uh
it's X and like Elon
>> X. Well, that's the platform of that's
the AI platform.
>> That's the AI platform. Yeah. And and
Elon has his finger on the dial and can
control that in whatever direction he
wants. And like I still imagine kind of
leads Daario in, reels him in, and then
actually shuts off. You're not engaging
in actions that are good for humanity.
And that I mean honestly 47 the backlash
again. We've used it. It's a little
wonky. So we all feel it. It's not just
being made up all the complaining around
it. But like it is kind of crazy to
think about
the power that still remains now and
that's amplified by now being your
provider of compute.
>> Okay. Yes. But now, actually, I'm going
to revise a statement I made just a
couple of minutes ago, which is when I
said you couldn't find someone that
dislikes more than Elon dislikes Daario.
Yeah, you could. It's Elon disliking
Sam. He dislikes Sam more than he
dislikes Daario. And I wouldn't discount
the IPO math here because let's say
you're Elon Musk and you want to kill
OpenAI. So, you bring them to trial over
their conversion from a nonprofit to a
for-profit. Okay, so he's doing that.
Now, what else do you do? You think
about the IPO calendar.
And there have been reports that
Anthropic is trying to go uh IPO this
year. So now think about it this way.
SpaceX will be the one that IPOs first
out of this. I think we could both agree
there. Then if Anthropic goes ahead of
open AI and uses this compute to support
that ADX demand growth and can show
exceptional demand uh uh you know an
exponential curve of demand before it
hits the public markets and then leaving
OpenAI to sheepishly walk in third and
you know with its reputation damaged
after the trial and uh with its compute
advantage not erased but minimized
because of this deal uh you potentially
and and it needing to raise more capital
to continue to serve its use case. That
to me is potentially the way that you go
about trying to kill open eye and maybe
that's what Elon's doing.
>> Okay. Okay. I like this theory. I like
this theory. This is See, this makes
more sense to me than actually like the
numbers around demand and whether it's
meeting supply and shortage of compute.
What you just explained actually makes
way more sense to me of how this would
have come about and how this is playing
out. Actually, one question. I'm gonna
I'm curious your thoughts and because
we've been talking about this. I now
view everything these companies do in
the lens of everyone racing to IPO as
fast as possible.
Why do these companies need to IPO?
Oh, that's a great question. Um, can I
my suggestion here is that they were
actually not planning to IPO this fast.
In fact, the news of the SpaceX IPO, uh,
I guess that might have been planned,
but the news of the Anthropic IPO and
the OpenAI IPO is rather new. My theory
is that they were actually going to try
to raise one more round of private
capital and they might still do it from
the final boss of private capital and
that is the Gulf States. And this Iran
war made the Gulf States more hesitant
to put all this money into US companies
and hence they have to go IPO.
>> Okay, that I agree with 100% in our it
took us a half hour but finally a
harmony moment here.
>> That makes all the sense. No, I mean cuz
I really was wondering like the money in
private markets remains it feeling
infinite like and again is is mass
public market investor exposure really
going to be that different like it's
just felt like it's made me wonder why
everything feels so urgent all of a
sudden when they've just been again what
was OpenAI's last round
22 billion
122 billion That is bigger than vast
vast majority of most IPOs in terms of
>> No, sorry. That's bigger than any IPO.
No IPO has raised that amount of money.
>> Wait, not even the largest I think was
Saudi Aramco and that was 30 29 or 39
billion. So this is at least 3x to 4x
larger than the biggest IPO ever.
>> Okay. Okay. See, so that's why it it has
been odd to me how intense this pressure
of everyone chasing towards IPO has felt
very recently. You made it all make
sense though, so I'm going to give you
credit on that one.
>> I will note Anthropic is likely going to
raise at least one more round. Looks
like they're going to raise $50 billion
at a 900 billion free money valuation.
Here's an investor. They told the
Financial Times, "People are ready to
throw any dollar amount at anthropic.
It's just about when they want to pop
their heads up and say we're ready.
>> I wait I still can't I do you know what
I think like
>> it didn't you're on a roll right now. It
did not hit me until that the largest
fund raise in history at the Ramco was
only 30 billion in terms of actual funds
raised. Like
>> oh yeah,
>> this is making even less sense to me
right now. This is just like you're
getting a 122 bill in the private
markets and you're like this energy and
intensity to go public. But well, I
think yeah, they will go public. Um, you
know, if they they will go public, they
will try to raise as much money as they
can on the private markets, but it's a
balance there because you don't really
want to be this third AI company to go
public.
>> No, you don't.
>> You're going to have there's going to be
limited money out there to fund you.
>> You don't That's why No, no. I think
that's the intensity of who's getting
out and when. I think that's why it's so
intense right now.
>> All right, let me see if I can just
build on my winning streak here for for
one more one more little fun thing here.
Uh, and that is uh is was Mythos
Marketing or was Mythos real? Um, this
is from TechCrunch. Um, Anthropics
Mythos has written Firefrox's approach
to cyber security. In a post published
on Thursday, Misilla said, "Mythos has
unearthed a wealth of highsecurity bugs,
including some that had lain dormant in
the code for more than a decade. That is
a significant improvement from what AI
security tools were capable of even 6
months ago. Until now, AI bug finding
tools have come with severe drawbacks.
Uh, but Misilla's researchers say the
latest generation of tools have turned a
corner, particularly now that Agentic
systems can assess their own work and
filter out bad results. The results are
striking. In April 2026, Firefox shipped
423 bug fixes compared to just 31 a year
earlier. Um, this is from a
distinguished engineer at Misilla.
Things are actually suddenly very good.
We see that our own internal scanning.
We see that on external bug reports and
we see that in all sorts of signals
across the industry. They jumped from
teens of bug finds to 423.
Still not coming around. Not. You had a
streak. But I'm still not coming around
on this one. I don't Okay, now I'm going
to go back to
one. A 15-year-old error in how the
browser parses an HTML element was one
of those 12 bugs uh that they published
details on. Is that a vulnerability that
is kind of like like were what what of
these bugs were just true
vulnerabilities that could kind of like
cause catastrophe and take down a
system? I think that like finding bugs
and bringing more attention to software
bugs and are these things that like it's
cool. I mean that is the promise of AI
and I think is it mythos specifically or
did anthropic invest and use the
previous models to do the same kind of
work or did it just bring attention to
it now? I don't know sandwich in a park
mythos until we hear more about it see
something concrete
>> and again you we can say this is
concrete but or do you believe this
firmly confirms
mythos is not marketing I'm going to
just go with Ethan Mllik the AI expert
in Wharton professor he says I realize
that mythos as hype means two different
things to different groups for insiders
it means mythos was not a magical step
change in AI ability. For outsiders, it
means Mythos couldn't really find zero
day exploits. The latter was wrong. Uh
the former was likely right.
>> Wait,
>> so the second one, so Mythos can find
zero day.
>> So that's magical step.
>> Now I agree with you. Now I agree with
you. That's what I'm saying. It's not a
magical step change in AI ability,
>> but it can't find zero day exploits.
Okay, let's just agree there.
>> Yeah. Yeah. All right. We agree.
>> Cool. All right. Uh so before we go to
break, we have to go to break because
we're going to need to spend at least a
good chunk of our time now reading from
the Open AI uh an Elon trial and some of
the evidence there. U but before we go
to break, um I want to talk a little bit
about our AI summit that we have in San
Francisco on June 18th. Look folks, I
attempted to post a short episode
Wednesday morning talking about what was
going on with this event. Somehow for
some of you, there were some ads that
were playing in front of that. It was
supposed to be an ad-ree episode. I
tried to take those ads down. It didn't
work. Let me tell you about the AI
Summit right now. Uh we're going to have
this thing called the Big Technology AI
Summit. It is in San Francisco at the
Commonwealth Club on June 18th. It is a
day of conversations. Starts, doors open
at like 12:00 or 12:30. Conversations
start at 1:00. They run till 5:00. We'll
have a wine reception on the roof. Very
small. It's going to be 200 250 people.
It's properly priced. I believe it's
under $100 right now, although the uh
tickets will rise as demand goes up. Um,
and we have a great lineup. We have Greg
Brockman from OpenAI who's coming. Uh,
the president of OpenAI, Arvin Vincent,
the CEO of Perplexity, Aaron Levy, the
CEO of Box, Lauren Good, the senior
correspondent uh, from Wired is going to
be there. And, um, we just confirmed
that Dylan Patel from Semi analysis is
going to be there. It's all-star lineup.
It's going to be a really great day. Um,
so if you check out
summit.bigtechnology.com,
uh, you can grab a ticket while they
still uh are out there. We're on our way
to to being sold out, so you should
definitely do that while you still can.
San Francisco, June 18th, Commonwealth
Club, be part of the first one of these
things. Summit.bigtchnology.com.
Now we go to break. We'll see you on the
other side with a dramatic reading
reading of Sam Alman and Mera Morat's
text. Back right after this. And we're
back here on Big Technology Podcast
Friday edition. Uh we're here with
Ranjan Roy of Margins. Uh Ranjan, u when
these text messages between Sam Alman
and Mera Maratti, the uh former CTO of
OpenAI came out, these are the text
messages that they were sending to each
other uh during what's called the blip
or when Sam Alman was um was fired over
a weekend and then eventually
reinstated. I texted you that we must
read them uh as a dialogue here. Uh, and
I think that will give people uh the way
give people a feel for uh what it was
like at the time.
>> Where should we start from the beginning
on this?
>> Let's start from the beginning. It won't
take too long.
>> Let's start from the beginning.
>> Okay. All right. November 19th, 2023 at
9:43 a.m. So, I will read Sam Alman's
parts and Rajan will read Meras. Can you
please officially invite me to the
office for a meeting?
Yes, I will. Do you have an update you
can share?
>> Adam is trying to get the board to agree
to configuration. He is now saying they
need till the end of the day. Sai and I
said that doesn't work and that we need
to start prepping prepping for plan B.
>> Okay. Okay. Please give me a second. I'm
about to speak to them.
>> Okay, great. Have an update. Let me know
when you can talk. And then
that was at 10:15 a.m. 5:37 p.m. Are you
on with them? unrelated problem if
you're still waiting.
>> Not yet. Just in a quiet room because I
didn't want all the outside theater.
>> 6:23 p.m. Sam Alman. Can you indicate
directionally good or bad? Satya and
others anxious.
>> Directionally very bad.
>> Okay. Can you wrap up soon? Lots of
pressure from Microsoft for an update.
>> Sam, this is very bad.
>> Can I come in?
>> They don't want you to.
What do you want to make it better? I'm
still willing to walk away if that
helps. If they are ramped up for crazy
lawsuits against me, then I'm not sure
what. Can you please tell them I just
want to resolve this however and would
like to join?
>> They are convinced about their decision.
>> For me to be fired or some new thing?
>> Yes. For you to be gone.
>> Okay. Then can I come in and talk about
a path forward with them?
>> They're saying no and need more time.
>> More time for what? They've walked me
through all the reasons and the issues
with you and why you can't be CEO.
>> Can you ask why they've been saying all
weekend they wanted me back?
>> They want a new CEO in place.
>> Can you say you will call back in 10
minutes?
>> They want to have a new CEO in place
tonight. Not me.
>> Do they know who can I tell SA? Is this
final or should you add SA in?
>> Trying to add Satia now.
>> Still don't want me. New guy is rando
Twitch guy. They don't want you.
>> End scene.
>> Oh my god. It really gives you a feeling
of like what it was like to be there at
that time. How stressful.
>> I also I mean the fact that Meera became
interim CEO is adds to the whole thing.
But Rando Twitch guys is definitely gone
from the entire thing. I think he's got
to be worth over a billion dollars. He's
like a founder of Justin TV, CEO of
Twitch. I mean, this is an incredibly
successful entrepreneur and all he gets
is rando Twitch guy.
He wrote he wrote on his Twitter, "It's
an honor just to be nominated." And I
think he made this background rando
Twitch guy. Yeah, but it's funny because
I mean I'll just say this, like when
this whole thing um continued when this
whole thing sort of seemed to wrap up,
you and I had a discussion and we both
agreed in no uh uncertain terms that the
drama within OpenAI was just not
anywhere close to being over. Um that
the structure still lent itself to uh
exploitation and uh and not exploitation
to instability. And clearly we're we're
like seeing that now.
Oh yeah. I mean I think we've said this.
Yeah. It is crazy that that was two and
a half years ago maybe. I mean that uh
that certainly the internal drama has
gotten you know has not gotten any less.
I mean remember all those departures
were just two to three weeks ago, right?
But it is amazing to know that Sam
knowing how intense and stressful that
was appears to be firmly in control
right now.
>> Oh, absolutely. I mean, they did they
did eventually, you know, they brought
him back. Um, and yeah, he's definitely
in control, but the question is like,
can Elon still use this sort of I mean,
it's interesting that this drama is
coming out now. I I guess my initial
reaction was like, well, what does that
have to do with Elon's, you know,
control of the or argument against the
nonprofit side of things. Um, but I
guess it's all it's all related because
of the it's all about the structure
basically and the structure is still not
settled.
>> No, not only and it's also about the IPO
of everyone's respective uh company
invested interest in this as we were
just talking about. But yeah, I think do
you think Elon is doing this for the
good of humanity and making sure that
nonprofit structures forever uh
remain
intact and uh respectable?
>> No, we know that he's doing it because
he invested 30 plus billion dollars and
feels betrayed.
>> Wait, 30
>> in that nonprofit.
>> 30 million. Million.
>> Did I say billion? Million.
>> Yeah. And nowadays, what's the
difference? But uh
>> I think it's quite significant, but I
won't debate on that
>> for us normies. But when we're talking
about fundraising,
um well, I mean, yeah, that's why I
don't think 30 million is significant to
him like at all. I think it's strategic.
I think it's
related to I mean, I don't even think
it's X AI. I think also I I agree
actually. Yeah. Going back to your
earlier point, I think he dislikes Sam
significantly more than he dislikes
Dario.
>> Yeah, exactly. Who who who is
>> the dark horse that kind of somehow
appears and comes is it Masa's son or
who would be the most unlikely character
that somehow becomes directly involved
in the Elon Sam Daario
triangle. Sadia Nadella I mean Sadia is
going to testify on Monday and Elon is
blaming Microsoft for giving OpenAI the
money and uh sort of pushing this
for-profit move because sort of it had
to pay back Microsoft.
>> What could what do you think Satcha is
going to say or do?
I think SA is going to fully throw open
AAI under the bus
>> 100%. Thousand%. Yeah.
>> And we already have this like very
interesting email uh chain between
Microsoft executives in 2018 um trying
to like decide whether they should
invest in OpenAI and it I mean they
eventually did make $13 billion of
investments in Open AI. Um and um and it
not not after like some real hesitancy
among the executives. Um so there's this
great the email chain is actually out
there. It was a great wired article out
there talking about it. This is what
Satya Nadella's email was to like a
whole group of senior leaders about
whether they should help funding them.
He goes overall I can't tell what
research they are doing and how if
shared with us could help us get ahead.
From what Elon is telling everyone he
feels open AI is at the verge of some
big AGI breakthroughs. They clearly are
pushing AI at a level none of our first
party or third parties are. Uh Microsoft
CTO uh Kevin Scott said there could be
some PR downsided uh downsides
associated with not funding them and
having them storm off to Amazon in a
huff and shick talk us and Azure on the
way out. He goes, "I'm highly skeptical
of imminent breakthroughs in AGI. In my
opinion, they're treating us like a
bucket of undifferentiated GPUs, which
isn't interesting for us at all. They
are not out there saying there's a
critical piece of research uh that we
can do only on Azure because of his
technical differentiation. If they were
then it could be interesting marketing.
Another executive my worst case scenario
is that having them ditch Azure for AWS
and badmouth us on the way over. I mean
it's amazing that this was what they
were thinking in Microsoft. Uh but they
eventually did it and maybe credit to
them for doing it, but they eventually
uh got so sort of exasperated with
OpenAI that we've seen that they've just
been pushing uh further and further away
from the company.
>> Do you know what I love about this?
>> What do you love about this?
>> Everything is comms in the end.
Everything.
>> Yes.
This this validates when I I get nervous
that I look at everything in the guise
of like that it's communications, but
like this is a multibillion dollar
investment. They're talking about
Azure's technical differentiation from
AWS and all like in the end they're just
worried about how they look and what
it's going to look like. I love it. I
love this is the secret we are giving to
our listeners.
We did get we did get a listener write
in and say you guys are talking about
comms too much and like fair enough you
know uh we take that to heart. Uh but
also maybe we're not.
>> Look at look behind the curtain. Look
behind the curtain.
>> Okay. So look I think that that uh we
will continue to review the evidence and
this case as it comes out. Obviously
it's a jury trial. So, u as much as like
what we see in terms of the facts of
this case, you know, matters and they
and it does and I'm not going to short
change the jury, the emotional appeals,
the jury matter a lot. And I think
that's what you see both of these um
camps trying to, you know, sort of push
out in public. So, we will continue to
talk about the case um which should
resolve relatively soon. Um but we'll
have another week of explosive
testimony, I'm sure, um this week. So,
Ranjan, I want to finish this week
talking a little bit about the layoffs
uh that we're seeing uh out there. I
mean, of course, you have the block
cutting, I think what was it, 40 or 50%
of the company. Uh this week, you have
Coinbase uh saying that 14% of its
workforce is going to be gone. Uh and of
course, what what is the rationale? The
rationale is one, and I think Brian
Armstrong, the CEO of Coinbase, gets
credit for this. He says first we're
currently in a down market and need to
adjust our cost structure now so that we
emerge from this period leaner faster
and more efficient for our next phase of
growth and then ring the bell. What's
the other reason? It's AI. AI is
changing how we work. Over the past
year, I've watched engineers use AI to
ship in days. What used to take teams
weeks, non-technical teams are now
shipping production code and many of our
workflows are being automated. The pace
of what's possible with a small focused
team has changed dramatically and it's
accelerating every day. So we have to
basically he's saying we have to
restructure Coinbase to do it. Um the
question I asked at the top of the show
is the question I bring to you which
kind of maybe will bring us full circle
um from our whole discussion today. You
know is this AI like actually being able
to do the work of many people or is it
just cope?
Both. This one I have to go both. This
one I uh like again all of these
companies just looking Coinbase's uh
compound annual growth rate of employees
over the last 5 years actually it's only
it's like 10% which isn't as bad as some
of these others. Um these companies got
very big and especially in big tech and
the metas and the Googles of the world.
So, I do believe just having kind of
like and also Brian Armstrong is an
entrepreneur and I mean I get this
firsthand like if your company starts to
feel more bureaucratic in general,
you would love to be able to kind of
flatten it, clean it up, which is what
they said they were doing. I do like the
ideas around I think everyone has to be
an individual contributor and can't be
only a manager. I think that there can't
be more than five levels of hierarchy
from the CEO. I think these are really
interesting ideas and I do think like
they also like uh and it it's not just
AI too as a lot of these companies just
got fat and got bloated and it's that as
well and using AI is kind of the excuse
there. But of course, like I mean I do
see this more and more and I do think
it's right that like the whole if you're
not getting on board the AI train. I do
think it's there's a reality to it that
like if people are just being completely
resistant to changing the way they're
working, companies are going to get more
and more and more aggressive.
>> Very interesting. So there was a post
from a meta engineer named Arnav Gupta
um that is sort of the counter here and
and maybe maybe it is the same side of
the of of the coin or the other side of
the same coin. Um he wrote this the
layoffs will continue till we learn to
use AI. Um from the pragmatic engineer
frame this really well. He writes that
his point is that a lot of these layoffs
could be backwards. They are probably
happening because more AI spend doesn't
correlate with better business results
at least yet. So here's this is directly
from the post. These layoffs, even if
they are not because AI is replacing
you, and even if they are some form of
AI washing, these layoffs are still
because of AI. And these layoffs will
continue till we learn to use AI. Till
we figure out how the GDP of the world
actually grows because of AI, we have to
offset the $70 billion of combined open
AI and anthropic enterprise revenue of
annual token spend by cutting some
salaries. Until we figure out how to
unblock each other faster, we can always
be removed from the org chart itself.
Basically, the argument here is um if AI
was going to be, you know, if AI was
going to help the company be more
productive, then, you know, they
wouldn't need to cut these people. Um
they're cutting the people because they
haven't yet found a way to make
themselves more productive uh with AI.
Uh and therefore, they must cut until
they do. Your thoughts? I mean I do
think I don't know like I see this more
and more that uh like the difference
between AI makes good people great like
like the ones who are kind of taking it
on and your level of productivity but it
doesn't necessarily shift the entire
organization. So what I do think I and I
think this is the right thing to be
thinking about. Everyone is thinking
about what does an organization 3 to 5
years down the road look like and
there's going to be a lot of fundamental
shifts like I don't know I've been
thinking about this a lot and like the
way a corporation is structured as of
today it's not that old I mean it's the
last 40 30 to 50 years but even in the
last 15 to 20 years like IT departments
were more kind of like a small part of
organizations and it was the internet
and then cloud and these things that
really like blew up how big IT
organizations were. So the idea that
companies aren't going to have to
completely change the way they're
structured in the next few years I think
is nonsensical. So people are going to
start figuring it out. And also I mean
again going back to incentives and comms
not to bring everything back there when
you're at a moment that the market will
reward you for just saying layoffs and
AI. I mean, why not
>> about to happen, right?
>> Yeah. I mean, I like what this guy Rohit
Krishna had to say also. One bare sign
of all the AI layoffs is that companies
couldn't figure out how to produce even
more by keeping the people and adding
AI. I'm not entirely sure how to think
about this. I mean, ultimately it is. I
think this is sort of what you're
getting at. It's a change management and
a company type of problem, right? Um, as
opposed to a technology problem, maybe.
I don't know. Or maybe these maybe
there's a point here. Maybe AI hasn't
really increased productivity at all and
we're just diluting ourselves.
>> And uh Elon Musk and and Daario can
partner all day long, but it's a bunch
of fooy. See, I think in aggregate it
has not yet because it's still again and
at the enterprise level especially it's
still other than coding it's still not
widely adopted at any kind of scale
especially on the agentic side and
that's something like it's picking up
but it's not just massively adopted but
I don't know I think like to me it's
kind of if 15 I remember like in the
early days of Microsoft soft word and
word processing if some imagine
someone's like uh no I'm just going to
write this on a notepad for you and I
don't want to use a computer I don't
know do like do you think this is
fundamentally different than past
technological shifts or I'm not going to
do this on email I'm going to just send
a letter is it is it comparable
>> I do think that it's different than past
techn technological shifts I would argue
you uh I just think that this the tools
are more powerful this time around.
>> I guess you can say that in the past
>> but that probably means the divergence
will be bigger.
>> I don't know like uh to me I guess yeah
to me it's still not significantly
different than manufacturing automation
in the 2000s and what that did to
bluecollar work in the US and also
outsourcing within China. And this is a
whole other discussion. I know we we
only got two minutes left, but but to me
it's still it's just happening in
knowledge work right now. So it's
>> right
>> a much louder story.
>> Okay. Bottom line then before we go. Um
so you say both. What is it more? Is it
more sort of posturing um more AI
washing or is it actually more uh
productivity with AI?
>> It's more preparing for an AI future.
It's not that it's already happened yet,
but everyone gets
>> they got to figure it out now.
Otherwise, they will have issues later
on.
>> I I will I will 50 I will mostly agree
with that. Let's just say that.
>> I just said it wasn't coms. I just said
it wasn't coms.
>> But I will say this is what I'll I'll
nuance I'll caveat with. Um two of the
biggest companies that have done this
crypto companies and we know crypto's on
the outs. So, it leads me to believe it
might be business business realities
versus AI improvement. Funny that Brian
Armstrong and Jack Dorsey have both seen
the light about AI as crypto hits its
downturn.
>> Let's end on that just so we get some
angry comments on the way out.
>> God. All right, there goes the show's
ratings. Ranjan, thank you so much.
Great to see you.
>> All right, see you next week.
>> See you next week. Thanks everybody for
listening and watching and we'll see you
next time on Big Technology Podcast.