Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
MoltBook, Rogue Agents, and the Power Limits of AI Security
Recorded: Feb. 3, 2026 Duration: 0:24:18Space Recording
Full Transcription
Music Thank you. I'm going to go to the next episode. so you Thank you. Thank you. Thank you. so Thank you. Okay, sorry about that.
Once again, Twitter spaces.
Somewhat a little unreliable and freaky.
But anyway, let's keep going here with some stories that we have popping out.
We just finished off talking with Lisa, who's not here yet, about how AI is really starting to look at mental health
and really provide some interesting and very
powerful solutions and tools.
Mentioned about a wearable that allows AI powered translator, really designed to help
people with autism who have trouble, difficulty communicating verbally.
So it's really powerful how at least in the healthcare industry, we're seeing profound changes and really amazing ways AI is being applied.
I mentioned about electricity and infrastructure booming, you know, which is causing incredible shortages on RAM and hardware as that's pushing through.
An article out of TechSpot saying that Microsoft is building an additional, their own 750-watt AI chip to really challenge NVIDIA's dominance.
So NVIDIA is stopping become uh essentially a single source failure
um but really the chip war is is a power war and and unfortunately only the biggest players are
kind of getting in this area um there are a number of smaller companies that are also jumping into
the whole um supplying alternative chips um i saw broadcom jumping into that area. Of course, then AI data
centers will obviously need all those chips. The interesting thing I saw was that traditional
data centers are in many ways unsuitable to the new AI requirement because AI equipment is
incredibly heavier than traditional IT equipment, I guess, with the servers they're
running and the way NVIDIA is building it.
So it's not a simple shake and bake and push this stuff out.
There are many restrictions and infrastructure issues that are occurring around the push
out for AI Dent dentist data center expansion.
Article out of bar chart.com AI data center expansion
is turning into an asset class.
They're saying that the framing shift is important.
Cloud AI infrastructure is being treated
like industrial capacity, not just an IT spend.
And that really matters
because it changes the investment behavior.
It really also accelerates consolidation.
I mean, the human, in fact, is the critical AI capability becomes owned, financed and priced like a utility with users increasingly dependent on whoever controls that factory, which is an interesting shift that we're seeing.
The cloud becomes industrial scale capital deployment. We're
also seeing massive nation state investment. We saw about two weeks ago, I think it was
340 billion going out of the EU for that. We're seeing massive investments out of Middle East
countries, also out of Southeast Asia. Again, I mentioned earlier in the podcast,
there is a strong sense that the second
and third world countries integrating AI
not only into the government,
but into their technology development, R&D and so on,
allows them the ability to really start to step up
and have a level of equality.
So I think that's why we're seeing massive investment across
that as well. One thing kind of in relationship to the healthcare stuff, a story out of why,
AI companions are now filling emotional gaps. AI companions are being used for comfort,
affirmation and emotional regulation. I mean, this matters because it's really reframing AI
from a productivity tool to a relational presence.
I do know they're developing also AI companions
for senior citizens in nursing homes,
in retirement homes and so on,
which I think is an amazing application.
But the human impact really is double-edged.
I mean, some users are going to find
support. And I think others maybe risk withdrawal, or maybe one side attachment to systems that
really can't reciprocate accountability. It's an interesting quandary when you start to look at
the interaction of AI companions, and you start looking at the ethics of that, dependency, disclosure, emotional safety standards.
Again, we come back to this, where are the guardrails?
That continually, this issue just continually rises,
given the absolute speed at which we're seeing AI are popping in.
One article that kind of looks at that from shnaya.com, the article's titled
Kill Switches Are Now Becoming a Design Requirement. Security experts are calling for hard stops
as agents gain autonomy. I mean, this matters because recovery requires an interruption
capability. And the impact on us as humans is, you know, safety.
People need assurance that these systems can be stopped before damage spreads.
And I think, you know, that gets the human in the loop.
It really restores human authority.
But I think we have to design.
Design and architecture have to assume failure.
You know, every system, I think, really should have an off button.
And I mean, would you trust an AI that you can't shut down?
I mean, that's the biggest thing that I think, you know, we need to start looking at and
need to start thinking about, especially as we see this stuff roll out so hard and so
Especially as we see this stuff roll out so hard and so fast.
fast.
I mean, trust, another article from Fast Company,
trust is moving from institutions to interfaces.
You know, the story really is connecting declining institutional trust
with rising interface trust, okay?
So AI is feeling responsive, it's feeling personalized, it's present,
which institutions really don't.
I mean, how often have we gone and dialed up and we go through voicemail hell as we're trying to
contact a real human? And I think given interfaces are now being so responsive and so human-like,
it really shifts legitimacy and systems now are gaining the and so human-like, it really shifts legitimacy,
and systems now are gaining the influence,
not through accountability, but through usability
and oftentimes emotional resonance that they're displaying.
There's the question of, well,
are they really, quote, feeling emotion?
In many ways, it doesn't matter because it's a black box,
and you're just looking at the outputs.
So this legitimacy in many ways is really experiential.
And it's interesting that more institutions are really now pushing out AI companions.
I mentioned the story earlier about Singapore deploying it across 80 government departments as an interface for their citizens. We'll see that more and more and more, you know, just because people are sick and tired, not being able to talk to someone, even if it's an AI and just get run around in circles. Now,
maybe the AI will be designed for that. But it's, it's, we're seeing that basically a lot of
human interaction is now being diverted to AI companions across it.
Now, a brand's gonna keep doing that.
I think so.
I think we'll see across what Walmart's doing,
what Target's doing, what Etsy's doing.
And then recently Starbucks had an announcement as well
in the news and I'll see if I can find that.
Yeah, Starbucks is actually using AI not only for companions, but for sustainability,
which is nice to see, because that's one of the biggest issues that we've seen,
that the impact of adding all this massive infrastructure, rolling out data centers,
rolling out additional energy systems to support these,
it's just really causing massive strain on us environmentally.
And if you look at the environmental section of my newsletter,
you see many, many stories of people being very concerned about that,
as well as stories about AI helping to make things more sustainable in many specific areas.
Now, in the other story from Starbucks, they're testing AI technology to improve customer experience and streamline their operations.
They're including AI chatbots.
They're helping out on auto scheduling systems.
So we can just see that being applied across the board with many, many other brands as
well because it enables them to not only be more efficient, but also track that efficiency
to a level that we've never seen before.
Unfortunately, an article out of Technology review titled human oversight is becoming symbolic
and they're saying that basically oversight increasingly is now happening after execution
and it really matters because review without veto power is just cosmetic and the impact is you know
we have this force reassurance because we believe that someone
somewhere is in control. And that's not really true anymore. And I think we're going to see
this whole accountability gap widen and widen and widen. You know, I mean, real control requires
timing and free oversight, but that's just being lost on a massive scale across the board
with all these new systems coming in and everything being implemented, not only government, but
across all the brands and everything else we're seeing.
One thing, oh yeah, here we go.
This is an article.
So out of NPR, children are growing up with AI as a social actor.
And I think, you know, what could possibly go wrong?
So kids are now treating AI like a social being.
And researchers are observing children attributing agency, emotion, and authority to AI systems.
And obviously, this really matters because, you matters because early developmental framework shaped lifelong
patterns and human impact is really generational.
We start to see norms around interaction, trust and learning being rewritten because society
has agreed on little to no boundaries and AI is now interacting directly with our children. They're anthropomorphizing
these AIs as a natural consequence of just being with them, playing with them, looking,
working with them, you know, and social learning in many ways is being mediated by machines.
So I think we are going to see this new generation, developmental norms are going to shift, you
know. I mean, what do we expect next?
More parental guidance tools, hopefully educational policy debates.
But AI is really becoming a classmate.
And social cues are now being learned not only from their peers, but from machines.
And I think we'll see this is the first AI native generation is really coming about at
this time. and I think
that's going to be an interesting and fundamental shift you know what does that actually mean in
many ways you know at this moment like I've said I don't think we can predict out further than maybe
a month two months maybe even three there are some major trends that I think we can on robotics and
other areas but there's a lot of this stuff that's happening so fast in real time.
I think it's almost impossible to kind of predict.
Article again out of Technology Review, why AI ethics is losing public attention.
After years of warning, people are tuning out ethical debates that feel abstract or unresolved.
ethical debates that feel abstract or unresolved.
And this really matters because this fatigue creates a vacuum,
which really defaults into basically accepting
what the current norms are and the current things are.
And if the human impact is the disengagement,
then society stops asking hard questions.
Again, we come back to this convenience
and this tendency towards frictionless.
Just in this
moment we live where AI systems and others are really gaining more and more and more
influence. You know, it's hard and difficult to see, but engagement really requires concrete
stakes and really kind of getting out of our comfort zone.
What else am I seeing across the board?
Oh, here's an interesting one.
So out of the Guardian,
spiritual language is entering AI discourse.
The language around AI is increasingly mirrors religious framing, omniscience, prophecy,
and salvation through technology.
Now this matters because belief systems shape behavior far more than features I mean the human impact
again is you know transfer you know existential questions you know once
held by faith traditions are being redirected towards systems built by
corporations and we're seeing lots of different faiths starting to use AI, whether it's to write the Sunday, Saturday, Friday
speech, or just to get ideas. And it's interesting, this blend over between AI as a tool for faith,
I think is also another dramatic shift. Last story that I'll mention before we close.
The New York Times, deep fakes are undermining shared reality.
This came out on January 27th.
As synthetic media improves the burden of proof shifts to the viewer.
And this obviously really matters because shared reality is our social glue.
Human impact is winding up being fatigued by all this deepfakes
and this AI slop. And obviously, then people just disengage when truth just feels negotiable and it
could be anything. In turn, I believe that will start to weaken civic trust and collective
decision-making. So it's a powerful shift that we're all going through as we go through this.
I'm going to close out now and just mention that this week in AI, you know, was not about,
it wasn't about a breakthrough model, or really in some ways about AI agents and, you know, multiple.
What I saw was a rebalancing of power between human and ai systems that we're
building you know across governments grids workplaces you know even our inner lives ai
is moving from something we consult into something that acts decides and is starting to shake the
outcomes before we've fully noticed the handoff i mean we saw authority migrate into rule sets and
frameworks we've seen nation states rush to define responsibility
for a genetic system.
We're seeing the physical reality of AI collide
with everyday life when data centers meet storms
and communities and power bills.
You know, we've seen beliefs shifting
as people treat machine output, not just as information,
but as judgment, as comfort, even moral guidance.
And we're seeing control slip into defaults,
workflows, and automation, where execution is now happening without a clear moment of human choice.
I mean, this thread that's tying it all together, it isn't fear, it's really our awareness. None of
this, I believe, is inevitable. Every system reflects the values, incentives, and boundaries
we choose to encode.
And convenience does not have to mean surrender.
Automation does not erase accountability.
And scale doesn't have to come at the cost of our dignity or our agency.
And frictionless, at the end of the day, does not mean the end of human agency and control.
You know, if there is really good hope this week, it's this moment we can name what is happening we can redesign it the moment we become aware of a problem or issue we can change it and we can
start to fix it i mean i think we need to start to demand more visibility instead of opacity
more consent instead of defaults and start to use a human judgment where it matters most
the future of ai i believe is not something that happens to us. It's something we can now actively participate in,
shaping one design choice, one policy,
and one conscious decision at a time.
As I mentioned, it's hard predicting the future.
At this time, we're really debugging it together.
I want to thank everyone today.
I want to thank Lisa.
I want to thank Achilles, our other speakers today.
Apologies for the crash with space.
Technology, what can you do?
But come again next week.
Hopefully, we'll have a smoother ride next week.
We'll cover the previous seven days of headlines of what AI is doing,
how it's changing, and how we're interacting with it.
And thanks for all listening today.
And we'll talk
to you again next week. Thank you.
