Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
The Founders Show 71: AGI Reality Check: Real Progress or the Biggest Tech Grift Yet?
Recorded: Feb. 19, 2026 Duration: 1:03:36Space Recording
Short Summary
In a recent episode of the Founder Show, industry leaders discussed the evolving landscape of AI and its intersection with blockchain technology. Key highlights included the launch of innovative projects like Data AI's infrastructure and Sogni AI's decentralized operating system, alongside concerns about the implications of AGI and the increasing economic value of data in the crypto ecosystem.
Full Transcription
Thank you. Thank you. Thank you. Hello, everybody, and welcome back to the Founder Show, Episode 71 today, AGI Reality Check,
Real Progress or the Biggest Tech Grift Yet.
Hello, my name is Pulse Digital, and I will be your host today for episode 71. If
you've been here in the past, you know how the show works. But let's remind everybody, as if
you've never been to a space, I'll be co-hosting the conversation. You can participate in a number
of ways today. The bottom left button is how you request a speaker spot. If you're supposed to be
up here on the panel, that would be where you press the button. Or if you have something to add
later and there's still a spot, that's how you get up. If you can't use your microphone or if you
want to weigh in on something that's going on, the comment section in the bottom right is going to be where
you go down there. You can comment along with the show. Be sure to tag the speakers so you know who
you're talking to. That way they can see it. They don't gloss and they know to come back to it later
even after the show. If you're looking above my head, there will be pin posts throughout the show.
You can look up there and see what's going on and read what's happening. If people have got
something they're going to be talking about, maybe they pin the information,
and that'll be where it's up there for you to read.
But as we go through the show, remember the guests that come up, you can follow them.
You can click on the profile, see what they're into.
They're here today entertaining you and giving you some information.
So make sure you follow them in case they end up leaving and you don't miss the opportunity.
Let's go around the room and see how everybody's doing.
how everybody's doing the banner account was up here early let's check in over there how are you
The Boehner account was up here early.
Let's check in over there.
How are you doing today?
doing today that checks out data ai how are you doing today data network it's hi everyone thanks
for having us i'm doing well very different about myself i'm a leader founder of the tie
and we're building ai powered data infrastructure for
blockchain so think of us as a data intelligence layer that ai agent enterprises platform
users to understand what's happening across defy and on chain we have indexed more than
5 000 protocol across 20 plus evm chain Xenon users and other big wallet providers.
And we've built our own AI agent called Crunchy for our marketing campaign so that users can
interact with us.
And we processed over 1 million query.
A couple of words about myself.
I come from a technical background, making complex things accessible. I started in crypto education
back in 2016 and I've been building since then. So today I'm wearing my builder hat. I'm not
theorizing about AGI from a research lab. I'm worshipping everyday AI pipeline introduction
every day and I have some opinion on what's real and what's noise.
And happy to be here.
Thanks for having me again.
Appreciate you taking the time.
Sogni AI up in the spaces.
How are you doing?
Hey, we're doing great.
Never a boring day in AI space or the crypto space as well.
I'm Mavis, co-founder at Sogni AI.
We are a decentralized operating system for creative AI for both humans and AI agents.
Happy to be here.
Appreciate you taking the time, Mal. It's good to see you again.
Nanek, how are you doing?
Doing well. Thanks for having me.
So yeah, I'm Nanek, co-founder of Human Tech, and we build tools for digital human rights, and now also apparently for agents as well, because wallet with policies, which lends itself really, really well to agents.
We built it for humans, but have been working on agentic authentication methods so that you can
give your agent a secure crypto wallet and control the policies for it. Agent finance is really hot
right now. So you're going to be part of the panel for sure, talking about some of that. Thank you
for being here. David up in the spot. What's up, man? How are you doing? GM, GM. David here, co-founding Inflective AI. In short, what we are doing, we build
the intelligence protocol for agents, meaning we are transforming documents and sealant
expertise into a structured tokenized intelligence, which is the core field for agents and robotics.
So basically we liberate the knowledge,
which is trapped in PDFs and locked expertise, etc.
So yeah, we are live now for four months.
We have 10,000 plus data sets and 5,000 plus users.
So we just warmed up a bit and we just started.
So looking forward.
Looking forward for sure.
I look forward to your opinions throughout the show.
Thank you for being here, David.
Valerie is in a speaker spot as well.
Valerie, Mike, check.
What's up?
Hey, great to be here.
Thanks for the invite.
I'm Valerie.
I'm part of Syndica.
We are investing in the startups, definitely AI as well.
Also developing some of them.
And basically, as everyone here,
using AI on a daily basis and at work.
Basically, I'm super excited to hear
what the experts and the founders in AI will share today.
And definitely, we'll also try to give my breakdown
and definition of what we will discuss today.
I appreciate that you're here.
I also do the sit back and listen part, man.
There's always something to learn here.
Human, how are you doing?
The Human Tech account, what's up?
Yeah, hey, hey.
I'm mainly joining for the account visibility.
I'm Caro from Human Tech, but the speaker from on our side
will be Nanak, the co-founder of Humantech.
Hello everyone.
Yeah, right on.
We appreciate you being up here and bringing the account.
You guys know when you bring the accounts
into a speaker spot,
it sends out the notification to your followers.
So getting the Gold Check brand accounts up here
is also very helpful for reach and stuff like that.
So as I go through some of these questions
that I've got for you today,
since we've got a full panel,
basically remember you can raise your hand as part of the UI on
spaces. It's very helpful. As I'm going through questions, the people that raise their hands,
I try my best to get to them. But if nobody raises their hand, I'm just going to start
picking you guys out of the crowd. So that's how it's going to go. So since the show today is about
real progress or biggest grift yet, I'm going to withhold how I feel about some of that unless I
get a natural way to get into it. But what do you guys think? What's your internal definition, first of all, as you respond to this?
What is your own internal definition of AGI? And then what do you think right now?
Are we watching real progress or do you feel like what's happening around us might actually just be kind of like vaporware
and that people are making kind of a grift out of the misunderstanding of what AGI may be or how LLMs work?
Because I know where I land, but I'd be
curious about you and Mal, since you're here coming back again, I'll let you jump in first.
Do you think we're seeing real progress or is this just vaporware out there right now?
I do think we're seeing real progress, but with anything, when you, like, for example,
I work on a project and I'm 90% done, it was so easy to get to 90%, but then they get to 100% takes a double the time. What I mean, what we're
seeing right now with just the agentic AI using strictly LLMs is
quite incredible. But you have leaders in the in the AI space
that say that LLMs is not going to get us all the way there. But
saying that it's just incredible what you're able to do,
especially with this new stuff with the open call.
At the same time, when you look at Moldbook and, you know,
people saying 150, 200,000 agents are aligning, converging on Moldbook,
it's all driven by an end user at the end of the day.
For me, AGI is when AI is running itself autonomously,
and we're not quite there yet.
The 90% to 100% thing made me think of battery charging.
We probably, some of us here know,
if you have an electric car or anything like that,
it takes a lot to get that final couple percent.
Same situation, right?
That's right. Yeah, I feel like that's the way it Same situation, right? That's right.
Yeah, I feel like that's the way it works.
Vanner account had the hand up.
I saw that.
I saw you put it back down.
Do you want in?
It's all right there.
Want to try it again?
Mic check.
Maybe not.
Might need to go back down and cycle back for a speaker spot,
because if there is any chance that we're going to hear anything,
because I saw the hand up, we'll have to do it again.
Valerie, what do you think for yourself?
Are we seeing real progress or do you feel like there's a lot more grifting going on
out there, people taking advantage of misunderstandings?
I think that when we have, you know, like revolutions as we basically have with AI,
there is always a stage where you understand that it's like probably maximum that we have now. And we need, you know, like more power to basically be there.
So I think that we are not real AGI yet, but the path is visible.
And for me, AGI is basically a system that can learn and perform
most economically useful cognitive tasks a human can do.
So I think that I would say 60, 70% we are there, but lots of work.
I'm not so stressful about AI, but I do believe that we can do great things,
integrating it in our life even more.
Do you believe that things are as advanced as some people would have you believe?
Because I run, I do a lot of spaces and I constantly hear AGI thrown around and that
people say it's right around the corner, it's right around the corner.
But it was right around the corner last year too, when I was talking to some of these people.
Do you think that they just keep dragging me along with that?
I think we're improving definitely, but I cannot tell that AI can change the world drastically and move all the people.
So I think that it's a great tool that we have and this tool improves.
So we can use it.
We can basically optimize and use our time efficiently.
We can build our AI agents, like our personal one, using Vibe clothing just to make our life easier.
But definitely there is what to improve and to work
and all the great startups that are here
and that are developing and keep improving
and overall all venture funds that are focusing on AI
and still deploy capital there, prove it.
Sounds good. Do you guys think that we may be mistaking the iteration or the rapid iteration of LLMs
for intelligence growth?
Because I know there's a lot more models out there, but do we see these models and think
that because there's more models that they're better models?
Are we mistaking the amount of them for the quality data?
Yeah, I'll answer the question. taking the amount of them for the quality data, data.
Yeah, I'll answer your question. And first our definition,
at least my definition of AGI is system
that can learn new tasks the way we as human are learning
without being so specifically trained on it.
And it's not about pattern matching or massive data set, but it's actually a system that can reason and adapt to transfer some knowledge across completely different domains.
So for example, if you train it on basketball and then you show it a football match, it will understand that you need to put the ball in the goal without telling it.
it will understand that you need to put the moral in the goal without telling it.
Now, today,
let's be a question, today we have very powerful and incredible LLM,
but it's powerful in a narrow way.
So these models are amazing at languages, code, summarizing,
but they don't really understand the world,
and they don't have a persistent memory the way we at least for now the way we human do so they can't plan 10 step ahead uh in a reliable
way and we see it every day in what we do so for example we index on chain data we understand smart
contract we try to decode what the smart contract is doing and i'm not speaking machine learning or
anything like this like we just reverse engineer a contract and try to understand is it a bridge contract is it
landing and so on and today we we still need to provide a lot of skills a lot of hooks and to our
scripts that are working along with the llmM to guide it in the right way,
but never have the response came or something that we've never seen before.
The human has to intervene. He has to say, no, the smart contract is not true what you're telling.
This is something, let's say this is a bettingting smart contract.
This is a casino smart contract, although we've never
trained it on something like this.
So we're not there yet, at least in our feed,
like understanding on-chain data and the differences
between smart contracts.
I think we're not here yet is kind of a good way
to sum up how I feel about it
too. A couple of hands went up there, Mao and then Optum. Go ahead, Mao.
Yeah, I agree with that sentiment.
And I tap on people that way smarter than I am in the space about this,
like Yan Likon. He's the chief scientist of Meta.
And he says that LLM is the wrong direction to go to for AGI,
saying that it may be a massive building block to get there.
At the end of the day, LLM is literally just pattern matching from millions, if not billions of, you know, decision trees that is pre-computed.
So it's not actually thinking or anything like that.
But saying that, you know, if it walks like a duck and talks like a duck, it gets 99% of the work there.
So in terms of getting real-world work done, it's incredible.
But in terms of AGI, maybe it's the wrong direction.
Maybe. That might be true. I know some people feel the same way.
Optim, what about you? What did you hear that threw your hand up?
Can you hear me okay?
Yep. I think we're way off.
I think Ellie from Data AI nailed a few very, very important keywords.
One being reasoning, one being persistent memory or memory continuity.
or memory continuity.
I can't remember which one he said.
I can't remember which one he said.
But these are extremely important words
as we're trying to explain AGI
because reasoning,
I don't know if you've tried to have a conversation
with a family member about politics,
but humans haven't nailed reasoning in general.
So for us to have an expectation
for these machines that we're really just, you know, for us to have an expectation for these machines that we're
really just getting to know, I think that's a far stretch. So that's one. And then the
memory continuity and the persistent memory is going to be key too. I think that people really misunderstand
the nature of AI and what it is.
People look at it as a fact machine
where it's a probabilistic, I always mess that up,
where it's just guessing at what the next token is going to be.
And that's a problem.
It's not so much of a problem if people understood going into it
with that understanding.
But the problem is that we have that disconnect
between people going into it and then our end result.
So yeah, I think we're still on the first question,
but AGI I think we're, I know we're still on the same, or the first question, but AGI
I think is problematic because you need all of these very, very specialized things to
for it to not be dangerous.
One of the words I latched onto is you said expectations.
And I think that that's important with AI because I've had to talk to a lot of people
who had really out of this world expectations for what their AI was going to do because
they've heard a lot and they're not in the game or whatever.
And they're just using traditional models and they're like, oh, it's going to be able to do all this stuff.
And it's like pump the brakes. Hold on.
I don't know where you got your expectations from about what this AI was going to be able to do.
But they think it is already to like step into their home and do their laundry and, you know, make the grocery lists and stuff like that.
I just find like somebody said it was narrow a second ago.
And I think that's a great way
when I'm explaining it to people
to manage their expectations.
Because the last thing I want is someone to get in.
I had the same problem with VR
where people would get into it
and then it just wouldn't live up
to what their fantasy was.
And they'd be like, oh, it's not ready for me.
And they would check out for a long time.
And I think VR suffered from that too.
And I don't want AI's mismanaged expectations
to have people miss out on trying things just because it wasn't there yet.
Vanner, third time's a charm. Hand is up. We're going to try it again. Can you unmute? Can I hear
you? Are you there? Nothing. Maybe it's like a high five. He's just holding up because he liked
what I was saying. And I'm supposed to run by and slap the hand. Either way, do you guys think that
that's true, that the management of those expectations matters as much as anything?
Because you run into people out there and they will say, like, oh, it didn't do this for me or it didn't do that for me.
David, I'll ask you, do you think it's important to set those boundaries for people when they get involved for the first time that this is what it's meant to do or this model is good at this thing and we use it for that?
Yeah.
What I think here is that, you know that models are just good enough already.
What's the missing part here is the context and the quality of data.
Because more than 90% of agents fail, not because they are dumb, but because they are
blind, they cannot see that the data so inflective the whole thesis is to you know it's about you
know messy data and as long as the data are messy you can get the you know a
real AGI 100% working.
So yeah, that would be my take here.
So on the previous question.
Fair enough.
We're going to get Vanner back up here again and give them a shot.
But let me run another question.
You guys can throw your hands up on this one too.
So most, if not all, major AI models that people are using right now are kind of centralized or controlled by labs or entities.
And regular people and maybe smaller players don't have as much of an influence on the process or the
outcomes of what reaching AGI is going to look like. So is that a problem? Like, does anybody
have concerns about like the arms race between OpenAI or what Brock and Elon are doing since
we're on the platform? And that AGI itself is something that people,
as we're watching this race, it's more centralized than anything.
And I know a lot of people here love the word decentralized,
but AI is not there yet.
So what are your concerns?
What's your stand on that, Optimize?
So I would disagree that we are there.
We have a million nodes that are in the decentralized network. That means that influence is concentrated.
That means that bias is concentrated. Intelligence infrastructure, it should be pluralistic.
See, I'm using these words that I always mess up. Pluralistic. It can't be centralized and it can't
be from a single source of data or a walled garden because that's how we run into significant challenges.
And I'm not going to beat down on Elon Musk,
but he was just, I think, earlier today or yesterday,
he was comparing, he was doing screenshots
and asking very, I guess uh questions that you know
either you're uh a good person if you choose one side or a bad person you choose the other and he
said grok's the only one that's choosing the right side and uh it just it just it really put an
exclamation point on the challenge uh that comes out of having a centralized anything,
whether it's compute or data sets or agentic engines.
So when it comes to a regular user who's not in tuned, how are they supposed to look
at these different entities saying things like that and compare and make decisions about where to subscribe or spend their money?
Well, there needs to be guardrails, I think, on some level.
You can't always guardrail those types of things, but there needs to be some guardrails
there. And there needs to be, again, we're going to go back to, I guess, that types of things. But there needs to be some guardrails there.
And there needs to be, again, we're going to go back to, I guess, that same point where
you highlighted expectations of if you get a data or if you get a quote or you get a
stat from your AI or your agent, there are ways for you to verify that data.
But if you're lazy or you don't know where to go to verify that information,
then you're just going to take it as fact.
And then it's that old adage where you're speaking into a telephone
and then by the time it gets six people down,
that number is where that stat is irrelevant or ridiculous.
Yeah, that's a great point.
How do I verify these data sets?
Because I've dealt with that on my end, too.
A couple of hands went up while you were talking, so you got some people interested.
Nanak, I'll get to you first, and then we'll go from there.
Yeah, I think the biggest problem with centralized AI is the amount of power. And I don't mean that power is inherently bad when
centralized, but I think people underestimate how much of the economic output of companies
in the next few years will come from AI. We've already seen many teams who are up to date with using cloud code and have
good practices there that cloud code is already doing at least 80% of the work for development.
And it's no different really for other roles, non-developer roles in theory. It's just that
there is no cloud code for product management. There's no cloud code for operations. There's no cloud code for marketing right now, but there easily can be. And so I think we'll
get to a point in a few years where the majority of our output, uh, or like, um, yeah, probably
the majority of the work as a company will come from AI, and the amount of control that that gives AI on everybody's both company
and also when personal agents become more popular,
which they already are starting to become popular on their personal lives,
it opens up a lot of censorship angles
and also leads to the outcome of extreme capitalism
and extreme communism, which is monopoly.
You have a few companies that are extremely critical infrastructure,
responsible for the majority of a bunch of companies' outputs,
and also a lot of people's individual lives,
and then governments will require them to censor things for users,
to kind of direct things certain ways.
And, you know, this all sounds a little bit hand-wavy, but I think it's very, very, very real in a path that I can easily see as a probable path to happen within the next few years.
to happen within the next few years.
When you mentioned personal agents at work,
the first thing I thought of was the privacy issues
of bringing or having a personal agent
in your workspace and the company,
sort of like the way that they might want access
to your private phone.
I don't know if any of you guys have ever dealt like that,
but if you're going to use your phone at work,
they want to install an app on your phone
and you have these privacy, then you get two phones.
You look like you're out on the streets
because now you've got a business phone and stuff.
So, well, I have a work agent and a personal agent is what you made me
think of. That'll be something for me to chew on later because that's a whole other conversation.
In the order, it was Mao next that went up. Mao, you went in?
Yeah. I mean, there's a lot of new points got covered here, but the whole concept of AGI
is terrifying to me because you're talking about something that is unfathomably intelligent,
able to escape out of any loophole or guardrails that is set on it. And the question is, will this
ultimate AGI see us as its creators and be benevolent? Or will it just see us as like nail clippings or
ants and find, not try to kill us, but just, we just get, you know, our lives ended as a result of
something else, it's pursuing ultimate knowledge or something.
But in terms of the private monolithic companies owning this
intelligent AI stuff, when DeepSeek R1 first came out, in my eyes,
it cut OpenAI's valuation in half. It showed us that the open source world is like 90 to 95%
as close to these frontier models. And it used to be every six months, then every three months,
not every couple of weeks. Just now, MinMax just got released, KimiK 2.5 Thinking. And the questions
that we have to ask ourselves now is there's three options for people. Everyone is using
Cloud Code with Opus 4.6 or ChattoPT Codex to do all their coding. And it is taking over,
like as the other person mentioned, like 80% of all coding is probably, if not now, but soon being
done by this stuff. But all your data is going through these centralized services, right?
Alternatively, you have all these amazing open source models releasing out of China that will
get you 90 to 95% there. But either you use their APIs, it's most likely going, all your data is going to the CCP.
I'd rather use the Western services in that case, or you can take these open source models
and run them yourself, but it will take at least $20,000 worth of compute power for you to own that
and run it. But that is incredible that it's not like a million dollars for you to run your own AI and
your own AI models. Like $20,000 is pretty affordable comparing the fact that you have
all human knowledge and power in your own hands, in your own house. You know what I mean? So
imagine a couple of years from now, what that's going to look like. Good food for thought.
$20,000 both simultaneously sounds like not a lot and a whole lot of money. I don't know how that works in my head, but that would be one to explain to a regular person
when we're talking about data sovereignty and stuff like that,
because that's a great subject about the cost to spend one up in your home now
versus what it used to be.
Good point, man.
Data AI, you're up.
Yeah, thank you.
I'll answer the question, but mainly I wanted to reply to what Mohs was saying.
And this is my answer. I wanted to speak about DeepSeq, especially DeepSeq v4.
But first, the first part of the question, are we controlled by centralized labs? Sadly, yes.
And right now the future of intelligence is being shaped by maybe five, six companies, OpenAI, Anthropik, Google Meta, and a couple of Chinese labs.
And they decide what the model can do, right? Like what they can do, what data they're trained on, and what value they encode.
and that's an enormous amount of power concentrated in a few hands even if you have
clawback even if you're running it on your own uh mic mini whenever you need the rm you're
contacting them and um it's mainly their intelligence you're just running the agent
on your server now the big shift i'm seeing happening, and I don't know if you were following this, is the announcement of DeepSeek. DeepSeek v4 that is and they are, like with the robots, with some announcements.
And DeepSeq V4 can be run on maybe an RTX 4090
or even a lower tier.
And that's not even 20,000,
it's less than around 5,000.
And this is scary.
I mean, scary for all those big L&M companies.
But what does it mean?
It means if you are a company, you never need to train a model.
Because you take the model and just you update it, you tweak it a bit, you add your whatever
skills hook and so on and so on, and you're good enough with
it. And whenever they launch a new model, then you take it and then you run it on your
own server. And this deep stick before came from a need where you have said NVIDIA, you're
not allowed to sell in China. Then when you are under pressure and you need to launch something,
you will launch something that will not need a very big GPU to run.
But then you open the market and you change the paradigm
of all those big data centers and big investments that are coming.
Because if you want to use an LLM,
the LLM is running in the data center or hyperscalers, and then you need to go and query them.
But what DeepSeq is saying, it's not launched yet.
I don't know when they're going to launch it.
I think April, May.
It's changing the problem.
Every company needs to have its own GPU running from your home or whatever for two, three GPUs.
Then you run the LLM, run DeepSeek,
and then you tweak it a bit.
You don't need to train if you're not an LLM company.
Training will only happen in data centers,
but usage and building your agent,
knowing that all your data, your financial data,
your health data is on your server.
That's a very big paradigm.
I don't know how it's going to be.
What's the effect on those data centers, those big companies.
But today, you can't run any anthropic or any opus or.
Elon got pushback this week for a post he made on the timeline, which he said a couple of times, things about you should upload your medical chart to Grok.
And somebody came back and clapped him again on that, like, don't do that.
Like, maybe don't upload all your private medical information to a large source centralized AI.
And what you guys are talking about, the centralized in your own home, run up your own model, maybe then. Maybe then that's the AI you can trust with your finances and stuff like that.
I don't think you should be handing it over to any of these centralized entities. What I do think
you should be doing is leaving comments down below because not all of you have microphones. I looked
and it's only like 10 of us. And the rest of you are in the spot called listener, which means it
should be saying called writer. I need more comments. I was reading them when data was talking and I ran out. So I need some of you to leave
comments down there about your thoughts so far. I also need some of you to follow the speakers that
are in the space. They're up here donating time and energy and information for you guys as listeners.
The least you can do, if you like what they're saying is give them a follow because they put
content on their own pages and they show up in other spaces and stuff. And that's the only way
you're going to catch them there because we only do this show once a week, every Thursday, every other Thursday, there's a
schedule at this time. Uh, but you would know that because you follow the founder's account,
which is primarily how you got here anyways. I don't have to repeat any of that, but, um, some
of the problems that were mentioned here about AI, I wonder if you guys have a specific problem
in AI that is more meaningful, um, when it's combined with crypto? Like is there a potential product
market fit where AI becomes combined with crypto that we solve one of these problems or something
that we pointed out that you think that crypto maybe doesn't solve yet? Because I've had,
you guys were talking about the hallucinations earlier about stats and numbers and how sometimes
you got to just check your figures. That's why I'm so hesitant with coding because I'm not a coding expert, but I've always wondered about how would I know
if something hallucinated? I'm not a math wizard by any means. So if I started generating code,
I wouldn't even be able to audit my own stuff. And I get worried about that. So is there a way
that crypto makes that better blockchain, solves this problem? What do you think Optum was first
hand up and then we'll go around the room? Vanner, I'm going to give you a shot when I get there
again, I swear. We'll try it again unless that's another high five.
Optim, you're up.
I'm going to keep it short and sweet.
When we're looking at transparency,
I think it's pretty hard to argue that Web3 gives the crypto,
the blockchain gives that transparency.
So when we're looking at data provenance and we're looking at where did this data come from and the meta tags that go along with that,
being able to have that transparent and visible on-chain, I think really changes things.
And it's a leg up on any sort of data storage options
that are out there today.
So I'll kind of leave it at that because I could go on
for a while about it, but that's the basic.
Well, that's why we bring you here,
because you can go on for a while.
Sometimes we need that, man.
That's not a bad thing.
That's not a bad thing.
We appreciate you, though, being short about it.
We appreciate you, though, being short about it.
Data, go ahead.
Data, I hear something for you.
Yeah, sorry, I heard it.
Yeah.
Also, I'll make it very briefly.
It's data access and verification, because AI agents
need data to make decisions.
And I'm conflict to speak about DeFi.
This data lives across dozens of blockchain, thousands of protocol.
It's constantly changing.
And crypto is making this data open and verifiable by default.
So no API keys, no gatekeepers, no trust me.
You can verify every data point on chain.
And that's the super power AI for AI system
that needs real live and data.
Fair enough.
And Vanner, I want to give you a shot, man.
I saw you even reacted.
So I know there's a live person there.
Can that unmute button work?
Are you there?
Are you good?
You just hanging out?
Totally just hanging out.
Sometimes accounts just hang out.
I don't mind.
You look good up there.
I have a question because crypto people, in my experience, are usually the ones that are more interested in fringe technology.
I think we're quick to try new things, buy new pieces of technology and stuff like that.
So since we're the first to try new things, it means we're often the first to try new AI innovations.
I think you see that with Moltbot and stuff like that and Claude.
We started talking earlier about giving agents financial agency,
giving them access to money.
And we've done that here with crypto.
We've given AI's economic agency.
How is this going to change the crypto landscape?
Because I've heard some talk about it, but in your opinion,
you guys are up here because you're founders in this space.
What do you think it means to give an agent the access to a wallet?
It's more than just API access, because I know that's the fundamental is, well, they're going to need to pay for access.
They need to be able to make thousands of microtransactions to be able to access this data.
But it's so much more than that, right?
Optim, what do you think?
Yeah, sorry.
Thanks.
I feel like I'm like putting my hand up for everything.
So, again, I'll try and be brief. So with the X402 protocol, it's extremely empowering where you can't give your bot-tenth of a cent transaction.
So it's never really made sense.
And I think that's why the 402 protocol has sat dormant for the last 20, 30 years.
Now we're in a space now where our agents can be completely autonomous, even if we keep
them within a box where they're not going crazy.
There's another protocol.
I can't remember it,
but you can max out what they're able to spend even on the X402 protocol.
So if you have the ability to go out and say, you know what?
It's almost like a la carte.
I need to, I need a piece of information, a sentiment,
a human sentiment about what's going on with
Solana right now, right?
And I want that piece of data.
And then there's a market spike or whatever it looks like.
And you can buy that or your agent can go out and autonomously buy that for one-tenth
of a cent or one-twentieth of a cent.
And then it goes back and then it's able to do its job.
That changes everything because what we're typically given is,
oh, you want that, you know, one-eighth of a piece of information?
Okay, here's our entire library.
Thanks.
It's $30 US per month.
And there's a six-month minimum commitment.
Like, so it changes things from a data access perspective,
and then that inherently improves how those engines
and those languages and those agents actually perform.
So I think the ability to be autonomous and do those,
I don't even call them micro transactions micro
transactions was me on clash of clans paying four bucks for uh uh gems um we're talking nano
transactions and um i'm here for it i love it micro transactions is how they made me feel better
about regular transactions is all those were but but nano, maybe I'll run with nano transactions.
It was all marketing.
Yeah, it was.
If I can get nano transactions to stick,
I'll be sure to credit you though.
Cause that might work a little bit better in terms of the phraseology.
Nanek, what do you think?
What do you think about having wallets in the control of AI and them having
agency? What does it change the landscape for you?
Yeah.
So we've already seen agents get wallets and then get
drained like right away, right? Like have you seen, for example, a week or so ago, Kevin Awake
launched an agent, and there were two draining attempts. One was successful. And like,
this is happening all the time, right? And, like, even when you tell your agent,
please never give your private key to everybody,
I mean, to anybody,
it's still, like, people will figure out a way
to trick the agent to giving away the private key.
And it's partially because the attack surface
is so big with agents, it's doing so many things,
so big with agents doing so many things and like there's prompt engineering
and, like, there's prompt engineering attacks,
attacks all sorts of stuff like that it's very clear that agent should not
all sorts of stuff like that.
have private keys which really changes the paradigm of how wallets should work
and this is actually something that has had to change for a long time like it's
been super irresponsible in my opinion for industry to be having private keys
on the user device but it's better than having it
in a server, you know, if we're gonna stick
with our decentralization values.
So it's more and more important to split that key up
so it's neither on the user device or on a server,
and yet the server can still enforce policies.
So policies are how we prevent if the agent's compromised,
then, you know, they can just send whatever transaction they want.
The tower can send it, right?
We need to make sure that there's some sort of safeguard.
So I do believe that architectures of wallets
that support that are really, really helpful.
And so for example,
one that I think is great is 2PC and 2PC MPC. We had
actually been building this for like human security for a while and then agents came
along and it's like, oh actually agents probably need this more than humans. But yeah, so I
think splitting up keys and making sure no party has full access to the key and that some external party,
like the non-agent party,
like the user can set policies for the agent is critical.
They're all solid points.
It makes me think about how ironic it is because you give these agents access
to their own private keys and the ultimate security for them ends up being the warm flesh of humanity.
They end up trusting us at the end of the day,
and we end up running security for them either through maybe a multisig
or whatever it takes just for you to audit.
Maybe it's transactions of a certain size.
Maybe they do have autonomy to make transactions below a certain threshold
so they can't hurt themselves or cause something to collapse on a massive scale,
but maybe you need to check in with them.
but maybe you need to check in with them.
That's a great way to think about it.
That's a great way to think about it.
Nanek, go ahead and then we'll go to Valerie.
Nanek, go ahead, and then we'll go to Valerie.
And then we'll go around the room.
I can continue, right?
So basically what you mentioned before about trust,
this is definitely that worries me the most.
But when I asked my friend who works with AI agents
and with all this stuff,
and he also integrated agents to make the payments and some
bots. He basically told me that you should treat agents like powered tools, like the person or the
company who controls, configures or profits from them. This is usually the like accountable party.
So in case when you understand that, basically you can use this tool and you have trust to it.
In this case, you just manage it as a power tool.
And this is one of the structural items that you have in your life and you use it.
I do not use it, but definitely I would love to find a couple of tools that I would use and that can basically help optimize my transactions or subscriptions or so because we have multiple of great tools right now.
I think a lot of people feel that way about tool discovery right now that kind of like, man, I wish I could just find that one thing to do that one thing for me.
That's a difficult task.
Yeah, definitely.
This is what worries me about.
You have right now lots of agents, and you should pick one.
And definitely this should not be like a scammer.
So it should be something that you trust.
Yep, good advice.
Mal, you want to get in?
I see the hand up.
Yeah, I mean, again, a lot of really good content here, I think, particularly with the X402 thing
that was just mentioned. It's essentially, you know, an HTTP request that says, if you want to
access this, payment is required. This is how much, right? And we're moving more and more to
This is how much, right?
And we're moving more and more to a workflow where it's a results-oriented economy.
Like, we don't care how you do it as long as it's done.
You know, since the beginning of humanity, we went from, like, you know, swapping goods and cattle to gold coins, then to banknotes and checks and ACH and SWIFT, and now stable coins.
and swift and now stable coins and now it's so easy to pay for things and the 402 is going to
allow us to know have our agents go and and fetch that data make those trades i think we're going
to have narrow agents for you know narrow use cases and the general purpose ones and it's just
really exciting to see where this goes of course there's going to be hacks and other stuff happens
but i believe we are making more and more concrete guardrails.
The other day, I just submitted a skill for CloudBot,
and they did so much analysis of my skill that I submitted
that I had to harden it a bunch.
And all these LLMs are going to help us audit all this stuff.
And we're moving really fast toward making this stuff really robust.
Very much agree on that.
And I saw that Syndica came up in the space,
maybe up and down a couple of times.
Syndica account.
Hello, how are you?
Basically, Syndica is my, yes, yes.
Right on.
As well, for the visibility.
Yeah, no problem.
I just wanted to make sure, in case there was anybody else behind me.
I'm not silent.
I'm not silent.
Let's go around the room.
I got a question, and I want to ask you guys each for a little bit of a prediction here.
From where you sit, maybe there's already been a change for you on a personal level,
either economically or in the way that you do your labor or your regular life. But how do you think AI innovation is going to change either economy, labor markets, or just your day-to-day life?
Give me at least one bold assumption about what you think may be changing,
whether at work or how you do your job or just your day-to-day interactions.
And I'll go in no particular order. I'll pick on you first, David, and then I'll go from there.
Hey, yeah, I really enjoyed the conversation.
I do have to make a disclaimer.
I have a sore throat, so I really struggle to speak,
but this one I really like.
I mean, my assumption here is jobs won't vanish overnight,
but I think that, you know, tasks might vanish overnight,
you know, because if you look on the AI agents from a perspective of, let's say, a rejo, you know,
we are far away from, you know, stealing the blue-collar jobs and then, you know, that the AI agents will step in place and just replace everyone.
step in place and just replace everyone.
So I think that we will see that lots of jobs will have a benefit from agents, already have.
But I think that we are still in the phase that we are more in the buzzword era
rather than concrete usability,
although there are great products already out there,
like OpenClaw, et cetera.
I was listening before when you were talking about trust, et cetera.
My take here is that I'm from the European Union, right,
and we are really, you know,
everything needs to be within the legal frame, right?
So I'm just waiting a day that, you know everything needs to be within the legal frame right um so i'm just waiting a day that you
know you will you know announce that every agent needs to have like you know know your agent like
kya or something like this so that additional layer will come that'll get some people's attention
know your agent always gets at least one or two people riled up good thank you no that does i you. No, that does. I always get a couple of people because they're anti-KY anything.
They don't want to know nothing, David. So know your agent's kind of a weird one too.
Let's see what everybody else thinks. What's going to change, Valerie, for you either economically
or labor-wise or maybe just day-to-day? One bold assumption. I think that how AI basically changed the hiring process is like drastic changes.
Because, for example, if you did not build in a proper way your CV overall, it seems
to me that soon we'll basically have such initial analysis from the VCs like this. As you know, right now, the companies actually check the CVs of the candidates.
So for me personally, I know that in my network,
especially right now, the hiring changes are drastically
and AI basically .
Human resources taking a very artificial turn, right?
Yeah, exactly.
How ironic.
I hope you guys have been saying please and thank you to all your bots because you're going to turn around one day
and they're going to be in the hiring chair.
You're going to wish you would.
I would say even that my friend put a remark for AI.
AI be friendly to me.
I'm a good person.
So please help me to go to the next step of the discussion with the team.
It's coming.
Something for people to think about.
It's coming to LinkedIn.
It's coming to a profile near you.
So those are good points, Valerie.
Nannik, what do you think is going to change?
Can you pinpoint something for you that might have already changed or something you think is going to change for the regular person?
I mean, I can say our backlog, like our engineering team's backlog,
has drastically changed.
It used to span many months,
and it was, you know, we'd make slow progress.
It would be very extensive.
And then we started being more serious
about our AI policies as a company
and making sure that we started setting limits,
like, hey, any feature
should take one to three days to build, no matter how big, even if it used to take months.
And this sounds maybe scary, I think, for people who have not been using Cloud Code recently,
but actually you can write better quality code if you're using it right now in one to three days
than you could
have taken months before for some things. And so all features have been super, super quick,
and our backlog is really, really decreasing very, very quickly. So soon we're going to have to
figure out more things to build or more ways that the engineers can help other parts
of the team, right?
Because the stuff that has been happening
with development with cloud code could also be helpful
to like marketing teams, to ops, to product management, right?
There are lots of things that you can do
if you have enough MCP servers, but just nobody's made a cloud code for marketing cloud code for any of these other
things so um so we're even working a little bit on internal tooling for the rest of the company
to to help them effectively apply ai to to their work the expeditious removal of backlog which is
a great change for anybody whether it's personal or business. Great point. Appreciate that one. Let's go now. What's changed for you? Is there something
that's already changed or can you see something coming down the road that you're looking forward
to? Yeah, man, I see a lot of good and a lot of bad in the future. I think software as a service
is in trouble. It's so easy to make bespoke software that is custom solving your solution with things
like CloudCode and things that will make it easier than CloudCode. And so unless the software
service is backed by real tangible things like GPUs, I don't know what the future is going to
be like. For example, we have a website builder we pay $250 a month for, and I just use CloudCode
to scrape it all into HTML so that we can update it faster.
I wasn't necessarily trying to save money, but I was able to do AI updates to it faster
if I can scrape the whole thing into HTML.
People will be able to do that easy.
I also think in the next six months to a year that 25% of the internet will be AI agents,
and we won't know who is real and who is not.
We're already starting to see that with OpenClaw.
OpenClaw is really hard to set up.
It takes hours and hours,
but in the next few months,
there'll be enough enterprising people
that make it one click easy.
Soon, you won't know what social media accounts are real
and what's not.
And then lastly, OpenClaw is incredible. I'm giving a bigger talk
on this next week in person in Singapore, but 50% of our marketing team is now OpenClaw. It's going
out on the internet and it's going to like Facebook groups and restoring old family photos.
It's turning people's art into immersive videos. It's giving us brand exposure and giving end users
something unique that they can keep. And it's just incredible.
People who use that open claw stuff now, it's painful to set up,
but you're going to get six months ahead of everyone else
if you start utilizing it before it becomes easy.
So that's the words I will leave you with.
My two thoughts would be someone's going to sell you an unboxable open claw
in that six-month window, which I think you probably already thought of.
And two, there will be a point where spaces like this and
things like what I'm doing right now will be your proof of humanity.
That's why I'm not worried.
It's going to take a minute for AI to come up here and tap amps through
a co-host spot and have dynamic speakers and stuff.
I'm not trying to issue the challenge,
but I hope it's not coming from my job soon.
Appreciate you, Mal, for being here and those are good points.
Let's ask the human account what they think is coming.
Is there something that's already changed for you or something
that you think is going to be coming a bold prediction maybe oh i already spoke on behalf
of a human tech oh my bad that's right that's kara sorry kara i'm just picking on you you don't have
to answer that my bad sog me what about you do you think that something's coming uh or is there
something that's already changed for you you can highlight sogny is i my uh my agent is oh damn
i'm doing it
to everybody here brand shout outs as i go around the room i'm literally at this point i just pick
to make sure i don't miss anybody so highlight for the brands data ai threw the hand up at that
one so you can do it in there first what has changed for you or what is going to change that
you can talk about okay i wanted to answer the last question because and there is one thing about AI agent that I wanted to mention.
We don't see AI agent as a tool to be sold. AI agent, I mean, Ma mentioned
Clobot, no one will sell a service to Clobot. Clobot will be an employee working internally because insurance cost is very high.
So you only build AI agent as a company for you to use them as an employee to do tasks.
And being in data space, I see my boldest take is that the economy value of data will go through the roof because
AI will become more capable and the bottleneck shift from intelligence to data quality.
And whoever has proprietary, real time and verifiable data will win.
And that's what we're trying to do at Data Network or Data Layer.
And instead of us building an agent
and asking people to speak with the agent,
it's better for us to build a harness and say,
go deploy it on your own server, on your NVIDIA,
that's what you will buy for DeepSeq.
And come get the data from us instead of saying,
hey, we're setting a data agent.
Solid points, dude. I appreciate you being in here and taking the time to come in here.
Everybody that's up here, by the way,
remember you guys, if you haven't followed them yet,
now is the time to do it as we're getting closer to the end of the show right here.
Optum, I will give you the opportunity and I've got some closing thoughts.
We'll go around the room with everybody one more time but often what do you think's changed for you
that you could already talk about or something you think will change in the next three to six months
sure i got i got i got three for you um so to uh at least point uh data quality and and
infrastructure guys will will stay around.
If you're just creating a thin wrapper around an LLM,
you're creating a chat bot,
like everybody and their dog in every single industry
is creating these revolutionary,
air quotes, revolutionary bots.
The second that platform decides to go, hey, that's a good idea.
We'll just add it as a feature. You don't have a business model. You're gone. Right. So unless
you're building an infrastructure piece, you don't have six months. So I think there's that.
Maybe that's a little too bold. But the second piece is uh i just like literally before
i jumped on this uh legal challenges are going to be something so a region in canada i'm not
going to point them out because that's where i'm from um their legal society just gave an ai
company the ability to represent people in the courts, like literally.
I thought I was looking at the onion.
So you're going to see missteps like this, and then it's going to be missteps that have
real life costs.
And then all of a sudden, there's going to be political liabilities, and then that will
usher in guardrails so it's almost like we have to we have to trip over ourselves and do a face plant before people take things like
that seriously so i think this is going to be be a big one that's my other bold prediction on that
second one third one vibe coding will actually mark innovation uh So I haven't coded in like 20 years,
but I went and I created a bunch of Telegram bots
just for fun, just to prove to myself I could do it again.
And it was fun.
Like I had a great time.
I didn't need a bunch of caffeine
and a pack of cigarettes to do it.
And so I sent it to one of my buddies
and he looked at my code and he was like,
yo, you didn't containerize any of your API keys.
And I was like, oh, snap. Yeah, that would have been bad if i if i sent this up so i think you're going to have a bunch of like rookies like me that are going to be vibe coding like yeah i did this really
innovative thing this is awesome and then you're gonna shoot yourself in the foot and then you're
gonna cause a problem for yourself and others so So those are my three bold predictions for the next even 15 days to 90 days.
You and me, when it comes to vibe coding, are on the peak of the hill of the Dunning-Kruger chart,
where we think we know so much.
And we're about to find out that we know so very little every time I try to get something audited, man.
That's an awesome point.
Thank you for bringing through.
And thank you to all of these accounts for pulling up to be a part of the Founder Show
episode 71.
Today, we talk about AGI reality check, whether real progress is being made or if it's the
biggest tech grift yet.
I didn't say at the start of the show how I felt.
And despite the number of builders I meet, I know real things are happening out there.
But man, in meme coin land, there's a lot of grift out there.
So you guys got to be careful because I know there's great progress being made.
But like we mentioned in some of these questions, the crypto bros are relentless.
Some of you may or may not be as into the crypto side of things because it was an AI show today.
But you got to watch out out there.
What I would like to say, the brands that came in here, Data AI Network, Sogmy was here with Mao.
Syndica is here today.
Inflective AI, David was here with Inflective today. Optimum AI was here, Optimum AI. Human Tech, Kara's down in the account right
there. You did great today, Kara, by the way. I want you to know that it was fantastic to have
you here. And Vanner Chain, up and down for a number of high fives throughout the show.
Never did hear from them, but I appreciate them hanging out. And all the speakers that came in
here, as I said, if you don't follow them, you're going to miss the content that they put out on
their personal pages and their brand accounts, which is high quality material. That's why we
invite them to the founder show. So if you want to keep up to date with what they're doing, you need
to go follow that. If you want to keep up to date with what we're doing, you need to follow the
founder's account. And if you're really into the risk side of things, you can follow my account.
It's a very weird place. But if you want to hook up with me again, we'll be here back on Thursdays
at the same time as we run it back. Episode 72 next week. If you're out there as a listener and maybe AI is
not your subject, you're thinking, man, I got other subjects I'm into. DeFi, NFTs, which are
still a thing. If you're into meme coins or you want to talk about Bitcoin, we do shows with
different subjects every week. It's not an AI show. Send the founder's account a message and say,
hey, I would be interested. Here's my background. Here's what I do, but here's my specialty.
And what we do is we'll put you in that file.
And the next time we get to a subject
where you're going to fit in,
that's when we send the message.
We build a group and we bring you in for the show.
So if you want to be part of the next show,
you got to follow the founders account.
Well, other than that,
we appreciate you spending your time.
I'm Pulse Digital.
I appreciate the founders account
letting me continue to host shows here.
We've been doing it for quite a while together.
I appreciate all of you choosing Pulse Digital for your online spaces entertainment today.
And we will see you next time on the Founders Show.
Have a great day.
We appreciate you.
