Starting 2026

A welcome back and early entry into 2026. 

Sponsored by: Knocknoc!

Become a Member: https://danielmiessler.com/upgrade

See omnystudio.com/listener for privacy information.

Episode Transcript

S1 (00:18):
All right. Welcome to unsupervised learning. This is Daniel Miessler.
It has been a long time. Sorry about that. Yeah,
I've been, uh. Yeah, it's hard to explain. Uh, so
many things going on. Just trying to figure out what
gives you guys the maximum amount of value. So I've
been trying different formats on YouTube and obviously on the

(00:42):
podcast as well, with little clips of like, uh, I
don't know, interesting things that I'm thinking about or whatever.
So I think those are a high value because, you know,
1 or 2 minutes or five minutes or whatever. And
it's very condensed. The thing I didn't like about the podcast,
which I still don't like, you know, someone just reading

(01:04):
their newsletter or kind of going through the newsletter or whatever.
What people kept responding to is like, hey, listen, you know,
it's not so much about reading the news or whatever.
I can get that from the newsletter, although some people
did want to get the news from the audio because
they don't like reading newsletters. It's it's it's hard to,
you know. Basically, uh, do the right thing for everyone

(01:29):
at the same time. Um, so that's something I'm still
thinking about and experimenting with. The part that people really,
really wanted to come back was kind of the first
part of the newsletter where I'm just kind of free forming,
you know, thoughts and just talking about what I think
matters and stuff like that. It's it's more like the

(01:52):
tangents is what people liked, which makes sense, because that's
the same stuff I like from other people. So, um,
with that in mind, that's what I'm going to do
right now is kind of go into some thoughts from
the last few months, which have been absolutely insane. And

(02:13):
before I do that, I'm actually going to read a
thing from our sponsor for this episode. Preventing exploitation is
all about removing reachability. Knock knock. Does this through allow
listing at the network layer tied to an identity? Your
servers say invisible, unreachable, and unexplainable by default. Knock knock

(02:35):
dynamically and safely updates firewall trusted sources, allowing only what's needed,
just in time. Removing access automatically. It runs on premise,
uses the controls you already have, and requires no end
user installation. If you want to see how this could
actually work in your environment, book a call at knock
knock knock knock I. Oh yeah, so that's the, uh,

(03:01):
sponsor copy there. I just want to say I actually
talked to the founder of this company and, uh, really cool.
So basically, uh, I got into security through network security.
I was like a hardcore like firewall engineer admin guy
at a company called CIS, and it was just insane. Um,
so anyway, the whole concept here is, uh, kind of

(03:25):
similar to port knocking, which is why they call themselves
knock knock. Um, anyway, it's a cool idea. It's like
everything is turned off by default. And then you do
like this strong authentication to get into a thing, and
all you've done is changed your normal like control rules,
like firewall rules, um, you know, router access, whatever it is.

(03:49):
And it's like, that's the only thing that changes, right?
And yeah, it's just a really cool concept. I love
it just reminds me of Unix actually. I was telling
him as well. Reminds me of Unix because it's like, look,
just do the basics really well. Like firewall rules already work. Anyway, um, okay.

(04:10):
So going back to what I was saying, um. Yeah,
I just feel really sort of cliche and trite saying
that things are different now. Oh, there's been so much change.
I just don't even know how to encapsulate that. It's
I guess one way of saying it is, I feel

(04:31):
like the last three months in AI have been equal
to or greater than the ChatGPT bump itself, and the
last month has been greater than the previous two months.
So the speeding up is speeding up. It's essentially what's happening.

(04:51):
And I want to talk about one particular piece of
tech which is now called open Claw. It was molt
bot before that and before that it was called clawed
bot um claw. And essentially what this technology is now
called open claw. It's like a lobster as the logo.

(05:15):
Essentially what it is, is an autonomous. Agent. I think
it is actually AGI. It is very close to AGI.
It's not quite perfect because it's not like an installable product.
But we've been talking here for a very long time
about how AGI is essentially going to be a product.

(05:38):
It is a piece of technology that is general enough
to replace an average knowledge worker. Right. Because just as
a review. Right. The reason nothing has been able to
replace a knowledge worker is because knowledge work is too varied.
It's too general, it's too strange. Right. The boss comes

(05:59):
back and they're like, hey, yeah, we're not doing that anymore.
We're doing this other thing. Oh, by the way, can
you get some tickets for this? Oh, by the way,
take the salesperson out for lunch or whatever. That, of course,
will require a robot. But, you know, put together a
perfect invite. Um. Hey, rewrite that report. Um, I kind
of want it to be like this report instead. Or, hey,

(06:20):
make sure you aren't offending these people or whatever. It's
like just random tasks that people are given during the
course of any given day. Even for developers who are
supposedly doing very specific work, half of their job is
still meetings, right? Or more, you still have to do
JIRA tickets. You still have to do like meetings and

(06:41):
like all sorts of other, you know, you got training
you have to do, interacting with a team, changing your
goals like all of that is extremely general. It's extremely different.
You can't have hard coded rules for that. And this
is why no tech has ever been able to, you know,
come close to that. Um, and do nearly as good

(07:03):
of a job as even an average human doing this
because humans are generally intelligent. We have natural nigh natural
general intelligence, and this is what is changing with this
open cloud project. Essentially what it is, is it's a

(07:25):
general system. They came up with their own AI loop. Um,
it's different than cloud code. It's somewhat similar, of course,
because these are all pretty similar, but they've got their
own sort of way that they're solving problems. And what
you do is you just go and install this thing
on a VPN, you install it on a mac mini. Ideally,

(07:48):
you put it on a system that has some, uh,
natural separation in it because it basically digs deep into
the system that you put it on. Right? And the
idea is it should be able to dig deep because
it gets full control of the system, right? It could
spawn things. It could look at screenshots. It can open

(08:10):
a browser and click on things like solving CAPTCHAs and
stuff like that. But the craziest part? Okay, the craziest
part about this thing is that it has this giant
fleet of tools for doing work. And it just keeps
trying to do things until it can do it. And
this is the reason I'm saying that this thing is

(08:31):
close to AGI. It is close to AGI because there's
an example that the creator of this thing actually gives.
I think he asked to book an appointment somewhere. And, um,
it was like, yeah, I tried to use their website.
It didn't work. Um, and he came back later. Okay.

(08:54):
The appointment was made. And then when he came back
and looked at the log, he's like, hey, how did
you do this? It was like, well, I tried the website.
That didn't work. Like something was wrong with their API.
So what I did was I learned how to speak
by myself. I saw we had an 11 labs key.
So I basically just learned how to speak. And I

(09:17):
called the company and I told them I wanted a
reservation for this time or whatever it was, and accepted.
And then I put it on your calendar. This is
the difference. This is the difference. The difference is the
thing hit an obstacle just like a human, and it
went around the obstacle just like a human. Now imagine

(09:40):
this inside of your company. Okay, now imagine this thing
has access to the wiki. It has access to confluence.
It has access to Google Drive. It has access to
JIRA and everything else where it could learn about the
company and how it works. Right? Plus, it has access
to slack. It could reach out to other team members.

(10:01):
It could reach out to the boss and say, hey,
this is what I'm thinking about for this project or whatever.
The difference in this systems like interactivity, first of all,
it's proactive as well, right? So it's learning about you.
It's learning about your tasks. It updates itself, but it
also sets reminders. It's like, hey, I need to send

(10:22):
out this report at this time. Hey, I need to
check in with my boss at this time. Hey, I
need to check in with whatever at this time. It's
designed to be a personal assistant. That's kind of like
how it's framed. It's also framed as an employee, which
is kind of what I'm talking about. But what I'm
specifically talking about is the ability to be an actual
worker inside of a company. For a company to actually

(10:44):
say instead of hiring Sarah or Chris or Raj or whoever,
I'm going to just install a bunch of these instead.
That to me is the AGI moment. Now, I've been
saying I expect this to happen in the form of
a product or a project or whatever before 2028, right?

(11:08):
I said 25 to 28 is when when I think
this is going to happen. Now some people are arguing, well,
that happened last year, right? Because you could technically stitch
all this together before. But I don't that wasn't my standard.
My standard was it's an actual thing that you could
go and get and easily install. And most importantly, I
have a much higher standard, which is our companies actually

(11:31):
using it. Are there real companies, startups, medium size, large
companies where they're like, hey, we're literally going to use
X product called replace humans, AI or whatever it is.
I just made that up. It's not real, but you
get what I'm saying, right? If they're actually installing it
and then they're like, hey, we were able to cut

(11:53):
our headcount 30% because we installed so and so product.
And then you come back and look. And it's not
like they did this trial because the board forced them to.
But over the course of months and years, I would
say even months or weeks, like if it just works,
if it actually works, they keep it going, they actually

(12:13):
increase their usage of it. And then you start to
see virally. Yeah, more and more companies are adopting it. Boom.
That is product market fit for an employee replacing technology.
That to me is AGI. I thought that was going
to be sometime in 2027. I thought 2026 was going

(12:36):
to lay this out in like the agents were going
to get way better and the scaffolding was going to
get way better or whatever. This guy released this thing
a couple of weeks ago, like a month ago, something
like that. Very recently it has gone Absolutely, completely viral.
Absolutely viral. Everyone is messing with this thing. It is

(13:00):
unbelievably fun and useful to use, and the more tools
that you give it, you got to be careful here.
The more tools that you give it, the more scope
that you give it, the more powerful it gets, the
more fun it is to use. And you got to
watch out for this, especially around prompt injection. That's a
whole separate talk show. But yeah, prompt injection is pretty

(13:23):
nasty because someone can find out that you're using this
and then they start sending in a whole bunch of
prompt injection to all your email addresses and whatever, uh,
chat locations, everything. Right. So you got to watch out
for that. But the point here is this thing is
open source. Completely open source, absolutely going viral. There's already

(13:46):
multiple twin projects. There's so many companies spinning up off
of this. and they're just going to take and rebuild
the whole stack in their own way, right? Because he's
got some pretty cool architecture there that makes it all work.
But every single part of it is open, right? So
all these companies are like, oh, worker replacement. We're talking

(14:09):
about what is it, something like $40 trillion somewhere between
like I think it's if I can remember the stat,
it's something like 30 to $70 trillion per year companies
spend on doing, uh, compensation for employees, you know, salaries,

(14:31):
you know, uh, insurance, all that kind of stuff. Let's
just call it 40 trillion, $40 trillion. That's the Tam
for this. And somebody just released a formula completely open
source for how to build it, But not theoretical. This
thing actually acts like an employee when you talk to it.

(14:55):
It feels real. It responds real. It like, hey, you know,
I'm not sure you've eaten yet, blah blah blah. I'm
looking at your calendar. You got to get some food. Hey,
don't forget to take your supplements. Um. Oh. Also, you know,
Julie is going to be picking up the kids later.
Do you want me to text her and have her
meet you? Blah blah blah. Like it is seriously that good?

(15:19):
And keep in mind, this is just a YOLO project
from a single person. Like there's more contributors now, but
this is a single person with a project he didn't
expect to blow up. What's going to happen when people
take that? And also using AI, also using this AI
go and actually start building better and better versions of

(15:40):
this thing and for the purpose of actually replacing human employees.
So all this to say. It is a really hard
time to be just a regular. Keep your head down,
just do your work. Knowledge worker if you are that person.

(16:06):
Or maybe you're not because you're listening to this podcast.
So you're you're kind of like, you read a lot.
You you, uh, yeah, you're probably at one of the
higher edges if you're listening to this. But all of
we have lots of friends. We have lots of coworkers. Right?
Not everyone is exceptional by definition. Um, all these people,

(16:32):
they're they're screwed. They are actually screwed. I seriously, seriously
believe that. And this is why so much of the
work that I'm trying to do is to to wake
people up. I call it activating. We've got to activate
them into thinking. No, no, no, you've got to be reading.
You've got to be looking at your frames, your mindsets,

(16:52):
the way you're thinking about approaching life, the way you're
thinking about approaching work. You've got to come to the
table with ideas. You've got to get your AI skill
set so far advanced, um, that that you're able to
end to end, deliver solutions. This is kind of the
the answer to. Okay. Yeah, but so what now what

(17:15):
what do I do? How can I make this better?
And the answer is I did a thing yesterday, uh,
basically saying that I think the solution is, um, three
things that I think are the filters. Okay. The three
filters in my mind are drive, creativity and AI tooling drive.

(17:40):
That's a harder one to get. It's kind of like
an attribute, but I do think that you can cultivate it. Um,
I think it's the most important because if you don't
have it, then it doesn't matter how creative you are
or how good you are with AI, the people who
are going to crush it in this world that we're
basically bifurcating into with like this top percentage and everyone else. Um,

(18:04):
and I don't mind that characterization because it's possible to
become the top percent if you do these things right. Um,
it's like in order to get into that group, it
is basically having this drive. It is having the creativity.
And then, you know, having the AI skills. So the

(18:26):
first one is the drive. You just the people who
are going to win. Are these really driven? Just they
they can't sit idle. They just they're always wanting to
do do do create create create build build build. Um,
that's the first step. Now, if you have that, then
it's about creativity. It's about okay, what ideas do you have?

(18:49):
What do you understand about the world and wish was different, right.
What do you want to see created that doesn't exist? Right.
So it's like you have ideas bubbling inside of you
that want to get out. And there's lots of people
who are like that, but they've never been able to execute. Right.
So they kind of like quieted their minds. This is

(19:12):
the other thing I'm trying to do is like, remind them, no,
you did used to have ideas. You did used to
want to write a book. You did used to want
to start a business. But over time, more and more
people around you basically like, oh, no, that's for, you know,
really smart people. That's for really business savvy people. That's
not for you. And we've been convinced over time to

(19:34):
be mediocre. Well, mediocre will not work anymore. It is
done like it was done in 24. It was definitely
went out completely in 25 and 26. Forget about it. Mediocre.
You will get crushed. And the next piece. Okay, I'm

(19:55):
just going to go to the next one. So the
next one after drive and then creativity and having the
ideas because those are two separate things. Then you need
the AI skills. That is this is the meta skill.
Because what AI skills allow you to do is basically
go end to end, to go vertical with everything you're

(20:20):
trying to do instead of being one piece of it.
And you need the finance person, oh, I need my
technical founder with me. I need a coder. I need,
you know, marketing. I need all this. Nope. You don't
need that stuff as much nearly anymore, right? You can
actually do most of this yourself, especially the coding. Um,

(20:41):
it's funny, it used to be ideas don't matter. Execution
is everything. Now it's kind of like execution doesn't matter
so much. Anyone can execute. The question is, what are
you executing on? What is your vision? What are you
actually trying to bring into the world? Right. So it's
a major inversion there. But bottom line here is the

(21:04):
filtering system is in my mind these three things. It
is the drive. It is the creativity and it is
the AI tooling. And I think when we're thinking about, okay,
what do we teach young people? What do we teach ourselves?
How do we continue our education? What should people be learning?

(21:27):
How do people get ready for what's coming? It is
that drive creativity, AI, tooling. And I do have sort
of a hack there, which I'm going to be talking
about in the human 3.0 stuff, which is. In order
to have kind of like opinions. And this goes to
the creativity one, in order to have opinions and ideas

(21:49):
and thoughts, I feel like you need to be bombarded
with input, high quality input. So I think there's a
cheat here, which of course we recommend a lot here
at UL, which is read 100 to 1000 great books.
And the more of these books that you're reading, the

(22:09):
more input you have, the more you know, ideas you're
going to have, and you're going to really cover the sea,
and the sea might actually get you so excited. It
helps you with the drive as well. Right? So number one.
And from there it's just about, you know, getting your
AI skills massively up to up to date, keeping them

(22:30):
up to date. And that allows you to execute. So
basically drive is like wanting to do anything at all.
Creativity is what are you actually doing? And then AI
is your ability to execute on it. So that is
primarily what I wanted to talk about in this one. Um,
there's lots of news going on. Let's see, what else

(22:52):
do we have for news? Uh, see, I went on
the Cognitive Revolution podcast with Nathan. That was a really
fun conversation. You can get that on YouTube. Um, let's
see here, just skimming through this. Uh, AI coding extensions. Yeah.
A whole bunch of vulnerabilities in AI coding extensions, Microsoft patches,

(23:17):
zero day in office flaw. Fortinet has had a whole
bunch of issues. If you're running that stuff, hopefully you're
up to date. Let's see here national security. I'm not
even going to go there. I mean, how much change
is happening with the AI situation. And on top of that,
we've got national security, international politics, like, oh my goodness.

(23:41):
Like I there's just so much going on. I'm actually
going to stop it here. I think this is a good, uh,
sort of re um, immersing into the, uh, the audio
format here and, um, look forward to, uh, keeping this
going in one format or another. Uh, let me know

(24:03):
or reach out, uh, via whatever method and let me
know if you like this Freeform content. If you still
want the news, maybe you want that news in a
different format, but still via audio. And also let me
know how how much you like the little short clips
that I was doing, uh, a couple of months ago
where I would just give, like a little clean idea, just, uh,

(24:25):
let me know and we'll see you in the next one.