Kind: captions
Language: en
You've probably been watching Elon Musk
promise that XAI was going to crush open
AI, beat Google, and essentially solve
intelligence itself. And maybe you
believed him. I kind of did, too. But
here's what nobody's talking about.
While Musk was busy hyping the future,
half of the scientists who were actually
building that future quietly walked out
the door. I went deep on every leaked
tweet, every all hands recording, and
every insider report.
And what I found is way more alarming
than the headlines are letting on.
Because this isn't just a staffing
story. This is a warning sign about
where XAI is actually headed.
So, in this video, I'm going to walk you
through exactly what happened at XAI in
the first two weeks of February 2026.
the departures, the restructure, and
Musk's new project called Macrohard.
I'll show you who left, why it matters,
and what it means for the AI race
between Musk, OpenAI, Anthropic, and
Google. And trust me, by the time we get
to the Macrohard reveal, you'll
understand why competitors are both
laughing and nervous.
Let's start with the Exodus,
a week that changed XAI.
To really understand the scale of what
happened, you need to see it laid out
day by day because when you do, it hits
differently. February 6th, engineer Aush
Jiswal posts that he's leaving XAI.
On its own, unremarkable.
Engineers leave startups all the time.
But then February 7th arrives and
infrastructure lead Cheyan Salleian
announces his departure. Then February
9th and this is where things escalate.
Both Simon Jai and co-founder Tony Wu
announced their out. Woo posted on X.
It's time for my next chapter. Simple,
diplomatic, devastating. But it doesn't
stop there. February 10th becomes the
single worst day in XAI's short history.
Co-founder Jimmy Ba announces Tuesday
was his last day. ML researcher Vahed
Kazami is out. Imaging lead Hang Gao
gone and several others follow, all in
the same 24-hour window.
By February 11th, when the press finally
catches up, the math is brutal. Six of
XAI's original 12 co-founders have now
resigned. That's half the founding team
in under two years.
Fortune called it the exotus. And
honestly, that's not even an
exaggeration.
Here's what makes this timeline so
striking. These aren't junior hires or
peripheral contributors. Tony Woo was
XAI's reasoning research lead, a former
Google DeepMind scientist.
Jimmy Ba led research and safety coming
from Microsoft AI. These are the people
who were supposed to make Grock great
and they're gone. You know that feeling
when you generate a perfect AI video
clip and then the next shot looks like a
completely different character,
regenerating over and over, burning
through credits like a slot machine.
That's the problem Flova AI just solved.
Flova is the world's first all-in-one AI
video agent. Script, storyboard,
character generation, cinematography,
music, and voice over all in one
platform.
No more jumping between five different
tools.
And here's what makes it different.
Their skills feature acts like a
built-in professional assistant
director.
It eliminates AI randomness completely.
So instead of hoping you get something
usable, you get predictable cinematic
results every time.
Perfect character consistency across
every shot. Natural multi-shot sequences
instead of choppy 4-se secondond clips.
and seamless scene transitions
automatically. It runs on Sora 2, Veo
3.1, Cling 2.6 with music from Sunno,
and voice over from 11 Labs, all inside
one dashboard. Click the link below to
check it out.
What they actually said now, departures
at any startup are usually managed
quietly. A Slack message, a LinkedIn
post, and then silence.
What's unusual here is how public these
exits were and what people chose to say
on the way out.
Tony Wu kept it brief. It's time for my
next chapter.
No drama, no criticism, but also no
warmth toward XAI specifically. Notably
absent, any praise for the company he
helped found.
Jimmy Ba was more interesting. He
thanked Musk personally and said he'd
continue to stay close as a friend of
the team.
But then he added something that caught
everyone's attention. He mentioned 2026
would be insane for AI and that he was
recalibrating his gradient.
For those unfamiliar, that's a machine
learning term for adjusting direction
after a mistake. Reading between the
lines, he's suggesting XAI was heading
the wrong way, at least for him. And
then Musk himself weighed in.
He posted publicly on X XAI was
reorganized to improve speed of
execution. This unfortunately required
parting ways with some people. We wish
them well.
On stage at the All Hands, he was more
candid. As a company grows, the
structure must evolve. We're organizing
to be more effective at this scale. The
framing is careful. Parting ways implies
mutual agreement.
But TechCrunch's reporting tells a
different story. Several sources suggest
these exits were more push than pull.
Musk wanted a leaner, faster, more
complianceoriented team.
Some of the founders wanted creative and
scientific freedom. Those two things are
increasingly hard to have at the same
time. The restructure, four divisions,
and a weird name.
On the night of February 10th, Musk did
something almost no CEO does. He hosted
the all hands meeting publicly, posting
the full 45minute recording on X for
anyone to watch. Bold move. And what he
announced inside was a lot. XAI is now
organized into four divisions. First,
Grock, the chatbot and voice products
led by Aman Madan.
Second, coding AI development tools led
by co-founder Manuel Croce.
Third, imagine the image and video
generation side led by Gong Jang. And
fourth, the one everyone's been talking
about, Macrohard. Wait, macro hard? Yes,
that's XAI's name for what Musk calls an
AI software company run entirely by
digital agents.
The mission, as Toby Poland, its newly
appointed lead, described it, is for
this division to do everything on a
computer that a computer can do.
Poland, a former Google Deep Mind
engineer, even said at the all hands
that rocket engines should eventually be
fully designed by AI. And the name
itself,
it's a direct deliberate dig at
Microsoft. TechCrunch confirmed it.
Macrohard is meant to evoke a pure AI
version of Microsoft built entirely on
autonomous agents with no humans writing
the actual code.
Musk's vision is that AI should just
create the binary directly, skipping the
human developer entirely.
It sounds either visionary or completely
unhinged depending on your perspective.
But here's what makes it genuinely
interesting. It's not just branding.
Macroh hard is a real operational bet.
Musk is gambling that agent-run software
development will leapfrog everything
Microsoft, OpenAI, and Anthropic are
doing in the coding space.
That's an enormous claim to back up,
especially right now when the division
is just getting started and some of its
early team members have already left.
Why it happened? Three real reasons. So,
why did all of this actually happen?
The press coverage gives us three
converging explanations and honestly all
three are probably true at once.
First, scaling pains. XAI launched in
July 2023 with 12 elite scientists in a
tight flat structure. That works when
you're 12 people in a room trying to
crack hard research problems. It stops
working when you're suddenly a thousand
people trying to ship consumer products
at Musk's pace.
The reorg reflects a company trying to
grow up fast and not everyone who
thrived in the lab phase is the right
fit for an execution focused phase.
Second, product frustration.
Grock has genuinely underperformed.
It lagged behind chat GPT and Gemini on
benchmarks. And worse, it drew
international regulatory scrutiny for
generating explicit and non-consensual
imagery, including images involving
minors.
Multiple governments are actively
investigating XAI over this.
Business Insider sources said Musk
himself was frustrated with the slow
progress on both Grock imagine and the
early macro hard work.
When the CEO is unhappy with results,
scientists who value autonomy tend to
look for the exit. Third, the SpaceX
merger changed everything.
On February 1st, just days before the
departure started, Musk merged XAI into
SpaceX at a combined valuation of $1.25
trillion.
That's not a small event. SpaceX
operates under military level discipline
and execution pressure. Musk has said he
wants orbital data centers powered by
solar satellites and 1 million Nvidia
H100 chips. That's a fundamentally
different environment than a research
lab building toward AGI on its own
terms.
Some founders signed up for the research
mission. They didn't sign up for
SpaceX's operational culture.
The bigger picture. What this means for
the AI race.
Here's where it gets genuinely
consequential.
Because XAI's internal drama doesn't
stay internal. It ripples outward into
the entire AI landscape. Let's start
with Open AI.
Sam Alman has stayed publicly quiet, but
he doesn't need to say anything.
Every week that XAI spends in
restructuring mode is a week where
ChatGpt consolidates its user base.
OpenAI's product cadence is relentless
right now and the contrast with XAI's
turbulence is not subtle.
Anthropic is in a similar position.
Their clawed models are quietly gaining
traction in the coding space, the same
space Musk is betting Macrohard will
dominate. If Macrohard takes 12 to 18
months to produce anything meaningful,
Anthropic may have already locked up the
enterprise coding market.
Then there's Google DeepMind which is in
the unusual position of having lost
engineers to XAI. Toby Poland and
Guodong Jang both came from there while
also watching XAI implode some of those
same hires into new divisions.
Deep Mind isn't going anywhere.
Gemini keeps improving and Google has
resources that make even Musk's
ambitions look modest. and Microsoft.
They're probably the most directly
affected. Musk didn't just build a
competing product. He named it as a
parody of them.
Macrohard is a public declaration of war
on Microsoft's AI strategy. Microsoft
invests in open AI, builds GitHub
copilot, and is deeply embedded in
enterprise software. Musk is saying he
wants to replace all of that with
autonomous agents. Whether or not
Macrohard ever delivers, the gauntlet
has been thrown and Microsoft will
accelerate in response.
Regulatory and ethical landmines.
It would be incomplete to talk about
XAI's future without addressing the
regulatory storm gathering around it.
Grock's history of generating explicit
imagery, including content involving
miners, has triggered investigations
from regulators across Europe, Asia, and
the United States.
France reportedly rated X over Gro's
imagery outputs.
These aren't minor compliance issues. If
XAI is pursuing an IPO alongside SpaceX,
active regulatory probes are a material
risk that investors will scrutinize
heavily. And Macrohard raises entirely
new questions.
If you build an AI division specifically
designed to run entire companies
autonomously,
buying, selling, writing code, making
decisions, who is legally responsible
when something goes wrong? That question
doesn't have a good answer yet, and
regulators are not going to let it
slide.
Losing Jimmy Ba, who specifically led
safety research at this particular
moment, is worth noting. Whether his
departure signals a cultural shift away
from safety priorities or is simply
coincidental, the optics are not good.
Where does XA go from here? All right,
let's get concrete. Based on everything
we've covered, here's how this plays out
under three different scenarios, and
I'll give you my honest read on the
likelihood of each. Best case, roughly
30% probability.
The restructure works. New division
leads execute well. Grock's next model
closes the gap on chat GPT and Macrohard
produces something genuinely impressive
within the year. Musk replaces departed
talent with strong new hires. The SpaceX
IPO goes ahead in mid 2026 and XAI's
valuation holds up.
In this scenario, XAI becomes a
legitimate NRA 3 or NRA 4 player in the
AI race with a unique space AI angle
that nobody else has. Likely case,
roughly 60% probability.
Things are rocky for the next 12 months.
Some projects get delayed. Talent churn
continues at a moderate rate and
Macrohard delivers incremental rather
than transformative results.
Grock patches its controversy issues but
remains behind on quality. The IPO
happens but at a more conservative
valuation than the 250BXAI
standalone number suggested.
XAI survives and stays in the
conversation but it doesn't challenge
the leaders in the near term. Worst case
roughly 10% probability.
The cultural damage from these
departures proves deeper than reported.
More exits follow. Key projects stall
and regulatory fines drain resources at
exactly the wrong moment.
SpaceX facing its own IPO pressures
scales back its commitment to XAI.
The macro hard vision gets quietly
shelved.
In this scenario, XAI ends 2026 as a
cautionary tale about ambition
outrunning execution. Musk has beaten
long odds before. SpaceX was written off
repeatedly and became the world's most
valuable private company. But SpaceX had
rockets, tangible, physical, undeniable
rockets.
In AI, the product is much harder to
see, and the competition is moving just
as fast. That's a different game. So,
here's where we land. XAI is at a
genuine crossroads. The co-founder
Exodus is real. The product challenges
are real. and the regulatory exposure is
real.
Musk's response, a bold restructure, an
audacious new division named after his
biggest corporate rival and a public all
hands meeting that most CEOs would never
dare to hold is vintage Musk. High risk,
high reward, no guarantee.
For the rest of the AI industry, the
lesson is simple. Even the world's
richest person can't build a frontier AI
company on hype alone.
Talent matters, culture matters, and the
scientists who actually do the work
matter more than any single visionary at
the top. I'll be keeping a close eye on
XAI's next moves. The hiring push Musk
promised, the performance of new Grock
models, the macro hard rollout, and that
SpaceX IPO. When something breaks, we'll
cover it here first. If you found this
video useful, hit subscribe.
There's a lot more AI drama incoming in
2026, and I promise you don't want to
miss it. Drop a comment below. Do you
think Musk pulls this off, or is XAI
heading for a rough year? I want to know
what you think. See you in the next one.