Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Ontology: Privacy Hour,
Recorded: March 26, 2026 Duration: 0:55:11Space Recording
Full Transcription
And we are live on X.
We are live on YouTube.
Looks like we're good to go.
All right.
Hello, everybody, and welcome back to the Ontology Privacy Hour.
It's been a while, but also we're live.
We're with video.
So this is a new experiment in media with Ontology.
Still talking about a lot of the same things when it comes to identities, privacy,
you know, decentralization, but with a bit of a different twist. And I think something that's
more contextually relevant and personally for me, interesting AI and, you know, these large
language models that are being trained on the data that we produce, be it consciously,
or should I say, where we're sharing it on social media, or maybe through some other platforms that
we wouldn't think that our data is being, I'll put it in quotes, compromised, right? We'll have a
conversation about that. But before we get started, I do want to allow our wonderful special guests an opportunity to introduce themselves.
We have two very special guests and two, I think, distinct points of view or maybe not distinct, but definitely unique kind of areas in which they're building.
I'm going to get started with Julian, just because I know you and I mean, I've known you for quite some time. Good to have you on. Julian, give us a brief intro to you.
What are you working on and how is AI relevant to what you do?
Yeah, I mean, yeah, thanks so much for having me, Humpty. Happy to be here. I've been in
the, you know, blockchain space now for over a decade, I know, along with yourself.
So we've been here for a while, you know, and had the pleasure to work together as well.
And, you know, a lot of the background that I've taken has been, you know, typically coming in with what are the issues that blockchain as a technology can actually address?
Like, where does blockchain as a technology itself fit into the world, right?
And what problems does it solve that can actually make the world a better place that are sustainable long term?
And we've explored that in a ton of different ways.
But today, where I think it's like most relevant is in making a fair playing field for humans and AI.
And that's what it is.
And that's why I think it's important is it's actually fair.
Can we prove a piece of content is made by a human or AI versus AI?
Probably not.
Can we prove when it was done? Absolutely, for sure. Can we prove that it
was done with an account that just I have access to? Sure, like things like this can be done. And
I think that's what's so interesting about the tech is it gives us maybe the only fair field
in the space. And so the things that I'm working on today are largely based around data context especially um doing this with uh within the artists especially here in los angeles and in
hollywood we do a lot of work with guilds and studios today uh we work with c2pa or coalition
uh for content provenance um and uh yeah really excited to be here and chat more awesome thanks julian nick off
to you how are you um tell us a little bit about yourself good humpty and first of all thank you
very much for having me uh it's a pleasure to be here i've obviously followed ontology and i've
known jeff for a while so it's good to be be connecting live um slightly different background
to julian although one of the things you said about the even playing field and sort of democratizing the things for the people.
I think that's a core part of my motivation for doing what I do.
But a bit of background. I've spent 25 years sort of at the cutting edge of technology, typically at B2B startups.
The last nine years, almost to the day, have been very specifically focused
on decentralized identity.
And that sort of journey starts, gosh, it starts before Gartner had even had a name
attached to things like self-sovereign identity.
So the hype cycle hadn't even been envisaged then, right?
And we were trying to solve this problem of how do you this very real problem that still exists still has not been solved today which is how do you prove simple things about you online
right and and what's the next iteration of of data provenance and digital identity going to look like
right and we had to take some fairly wild leaps of faith including you know what's going to be
the root of trust it's probably going to be a blockchain, right? Is it going to be Ethereum? Is it going to be Bitcoin, which was still quite
nascent back then? No, they probably aren't quite good enough. So we actually went and built a
blockchain, right? So we built the first blockchain for identity called Sovereign. We signed up some
big enterprises like Cisco and T-Mobile and IBM. But here's the thing, right? We were very early.
If you think, and you'll know this intuitively,
if you think about all the things that need to be true
to get to this sort of perfect world
where everyone owns and controls
and has consent over their data,
all the things that need to be true, right?
You've got to have governance.
You've got to have usability.
You've got to have technology. You've got to have governance you've got to have usability you've got to have technology
you've got to have awareness you've got to have the right business and use cases a lot needs to
be true right so it doesn't mean that we were wrong with this model it just means that it's
taken longer to get there and the thing that has been fascinating over the last year or two
is ai has just come in and turned this problem,
cranked it to 11. All of these problems have proven who you are online and protecting databases.
These are now house on fire problems. So AI has just come in and cranked the dial massively.
But my more recent journey, the last two years, I've been running a small,
we call it ventures consultancy, helping organizations, governments, startups, big
enterprises make sense of this new world order of digital trust, of user-centric identity,
how to think about it, how it's contextual for them, what the market forces are they need to
react to, and actually how they should react to it.
Amazing.
Jeff, why don't you give yourself an introduction here as well,
and then we'll start tackling some of these big questions.
Yeah, absolutely.
So assuming a few people who are tuning in have heard me go on about this
for quite a few years now.
So I've been in the space not as long as everybody else.
So seven years, I've been into my seventh year, which is okay.
That's not a bad amount of time.
Academic background, interesting identity in terms of its use case.
I want to pick up on something Nick said.
You know, I've been with Ontology now for five of those seven years, I think.
And I think what's really interesting is
we've gone from this position of
almost a niche idea that was important
and people in tech and people in identity
knew it was important,
but trying to convince the wider world
that it was important was always really hard
to suddenly being, wow, this is urgent, right?
This is actually something we've got to figure out.
I'll give you an example.
I'm currently renovating my house.
I'm not very good at renovations.
I'm learning the amount of AI-generated content that has convinced me
that I can do things I cannot do is insane.
And this has got real-world implications.
I have a wife who looks at me
disappointedly, right? These things have to change. And so, yeah, I think what's really
exciting for me is how urgent this is becoming for solutions.
Well, thank you for that introduction. And wild to think that you took on a renovation
challenge purely on speculation, not from crypto, but from AI. We're going to make it,
maybe. All right. Well, I guess the first question I want to throw at the group here is,
before AI, and I know AI has been around for a long time, but when I say before AI, let's just
put this into terms that we can agree on. Before this recent wave of AI, where it's proliferated, it's reached a consumer level where
almost like anyone can use it really. Before this recent wave
of AI, where we've seen companies like open AI consuming
tons of data in order to train their large language models,
right? And there's like them, there's a few other handful of
companies like Google and others that are doing the same.
Were there good systems for protecting data?
Firstly, our data.
Were there good systems in place?
And I guess the reason why I asked that question,
so to think about it this way, is if we had good systems
in place for protecting our data privacy, specifically, have those
transition well into this new era of data consumption? Anybody want to take that? Feel
free to take a shot. I mean, I would just start with, I don't think that, I don't think such systems were really necessary, to be honest.
And I suppose to some extent, data privacy was something every company took on as a liability, you know.
And by privacy, of course, I'm referring to, you know, human data, like people data, you know, stuff like that.
And so I think it's a liability that a lot of those companies took on um and probably made really really good efforts at but it's just it's always been a honey
pot problem right it's just it's just been as simple as that and i i think that you know it's um
it's it's kind of tied up in this you know if you have that data in a raw format somewhere i mean
You know, if you have that data in a raw format somewhere, I mean, it's just a matter of time.
It's a matter of time.
It's interesting because it sounds like what you're saying is, and correct me if I'm wrong here,
that the data kind of was in terms of the privacy or the protection of that data was up to the institution.
But the security risk was a human security risk where someone could leak that
data or sophisticated actors like hackers or nation state attackers could then access
that data. But in terms of its proliferation and its usage, it was limited by the human
behind these attacks.
Probably, yeah. That's much of sense.
Yeah, that's good.
I like that question, Humpty.
Look, by its very nature,
the fact it's called a honeypot of data
means that it's going to be sweet
and somebody's going to go after it.
I've used this metaphor for years
of security around honeypots of data
almost being like a bucket around water, right?
And there's leaks in the bucket and you have to keep patching them.
And if you look at the security industry today, and this week it's RSEC, right?
But it's all about what layers can we put on to continue securing that data?
And look, fundamentally, that's the wrong model.
Like you have to smash that.
We need a new bucket is the point because you can't,
anytime you've got that concentration of data,
somebody will find a way.
And of course now with AI,
the ability to do cyber attacks
has been entirely democratized.
Anyone can do that.
If you talk to any CIO or CISO
and you talk about the number of breach attacks
they've had over the last six to 12 months,
it's exponentially increasing. So the last six to 12 months is exponentially increasing so
that the model needs to shift right and it's got to be this whole sort of notion of distributing
that data to different nodes on the network the different nodes on the network being either
organizations or people right or things or agents that have and can control their own elements of the data.
And it's more than that, right?
There's a second factor of sort of security and I guess practical usability here,
which is it's not going to be enough to just have the data.
You have to tie it to the person who's authorizing its use, right?
So it doesn't matter that somebody's got my social security number.
In the new world order.
I should have to prove that it's me, Nick, showing you my social security number.
So the fact is the data in the bucket becomes less important and less valuable anyway.
If you don't have the added context of proving it's you or it's your agent, for example, that is using that data.
it's you or it's your agent, for example, that is using that data.
Great. So I guess there's two questions here that I would like to pull a thread on,
because I think two different points were made. One of them is the authentication of data.
You know, in other words, am I the producer of this data versus the general data protection, like who is in charge of protecting data, right?
I think I want to start with more like authentication, authenticating data.
What does that look like today?
I mean, I can take a stab at this because I have a, I'll start with the DGEN perspective.
because I'll start with the DGEN perspective.
So, you know, it's like privacy to me is like the ability to choose to say this is mine.
But if I don't say it's mine, that's how it remains private.
Now, the funny thing is 2025 is the last year.
It's coming out that 2025 was the last year that humans produced most content.
Most data was made by humans. 2025 was the last year that humans produced most content most data was made by humans that was the last year ever so that was our last that was peak
of human dominance of data creation was 2025. that was it that was the best we'll ever get um we'll always be second at best from now on what does that mean well it's actually kind of funny um
What does that mean? Well, it's actually kind of funny.
99% of data is AI generated, and it's indistinguishable from human generated.
I mean, that's maximum privacy, right? Does anything need to be secured?
No, actually, because you would assume everything is fake.
The only thing that needs to truly be secured and authenticated, your point, is who?
Is me, is when I choose to say this is mine.
This is human made.
The data doesn't even matter.
It's just when I say this is from Julian.
That's it.
Like, that's what's so powerful.
And that's where efficiencies and blockchain come in.
And I think Jeff and Nicola have far more to say here,
but that's my two cents on where we are today.
It's just a weird little point that we're in.
And I want to hear those counterpoints.
Hang on, Jeff.
I want to hear those counterpoints.
But to your point, real briefly, I want to say it's interesting that that's the potential, I guess, of how our personal data could be obf this data, right? That just became noise.
And it was just, you know, it's all false. But at the end of the day, it just kind of erased
the reality of who you were, anything that was personally identifiable from the internet,
right? Because you can't delete it technically, but if you overwhelm the system with this noise,
there could be privacy within that,
which is really interesting to consider.
Go ahead, Jeff.
I know I interrupted you.
No, that's fine.
I was just thinking this kind of,
and this goes back to the previous conversation as well,
but I think there's kind of a second order thing in here,
just in direct response to you on this, Julian,
in that, yes, privacy is the ability to say,
I don't want people to associate that data with me.
And then your control and your sovereignty, if you like,
is to be able to say, yes, this is me.
But there's another thing going on here, and that's intent.
The intent of the person who you give that access to.
And you can liken it to sending nudes, right?
If I send a nude to Humpty, he's gone, Jeff, send nudes. So I send him a nude.
That's fine. That's privacy. It's in this same thing. But if Humpty then has gone,
Julie and Nick, look at Jeff here shaking his thang, right? That is, that intent matters.
And I think that's true of privacy within data as well. So what happens once you've authenticated it also matters. And you know this can go back to things like
Cambridge Analytica problems and things like that in the past, but there is this extra thing that
goes beyond just that ability to say yes I give you permission to know this is me. Makes sense, I agree.
Nick anything you want to add here in terms of authentication or more personal identification, I guess, of data?
Only that, to riff on what Julian said, I think if you boil this down to sort of primitives,
like just the fundamental issue of proving if something is AI or human, right?
There's a lot of people chasing this proof of personhood notion from Vitalik to WorldCoin to the first person project.
There's a bunch of people trying to solve this
because if you boil things down like that,
first of all, is it a bot?
Is it a human?
And it might be fine that it's a bot,
but then who's authorized it and what's it authorized to do, right?
authorized to do, right? Those are just critical questions of our time.
Those are just critical questions of our time.
I guess the question begs, or the statement begs the question,
is privacy any more of a concern now than it ever has been in the past? It sounds like to me,
at least from what I'm hearing here, is maybe not so much because of the amount of information that's available about everything else.
Is that what I'm hearing? I mean, I want to make sure I'm not misunderstanding that statement either.
I don't know if I would say that we're there yet. That would be nice.
When there's 100,000 active Julian Brabans on Twitter,
I think we're probably there.
But then wouldn't the current Julian Braban,
please stand up, right?
Yeah.
On X be identified by your username,
which has potentially reputation of you having been active
in certain times before AI, right?
Yeah, very true. And then it comes down to what is the source of that?
How do you, who's making that determination? Is it X that's making that determination?
Is it the internet archive? Is it my own memory or relationship with him?
Is it some other source of truth?
And I think that's where that becomes important.
If we can boil everything down to something like when,
like timestamp and cryptography,
oh my God, what a great world that would be.
If everything could be simplified down to just time and proof,
everything becomes so much better.
Interesting. I wonder, and you know, there's definitely a lot more to unpack here, but this idea of like attribution is still kind of in the back of my head. What is there value to attribution
of data? I guess you're talking about the when timestamp, right, Julian? And I talked briefly about the who
from a user account point of view or user ID point of view. Is attribution valuable? And I guess
maybe I'll send this back to you, Julian, because I know you work with data a lot and you're building
products that are, to some degree, attributing some information to that data. Is there value in data, like
attribution of some sort of way in the data that is being collected?
So again, maybe this is a bit of an extreme perspective, but I think attribution is the
only thing that's valuable. If Bloomberg comes out and says Tesla stock is XYZ price,
if Bloomberg comes out and says Tesla stock is XYZ price, right.
That's worth something. If I say the same thing, probably not really,
if I'm posting the same thing on, on, on some, some social, uh, you know,
place probably not worth as much at the end of the day,
the source of that is, is more important than the content.
And I think this is, that's kind of the direction that I see it going to and that's like it's something that's like a lot easier to you know manage and control
and even when you come back to kind of like the ai data set and i don't want to get ahead of
ourselves but the ai data set training issues right um deduplication is a major thing uh ai is
it's not 2015, right guys?
The whole of the internet isn't human-made.
We can't just train off of whatever we find now.
And training off of content that you yourself made is an Ouroboros effect,
and you'll destroy your own model.
So things like C2PA and these content provenance things exist.
So you can say, hey, this was actually made by AI,
trying not to train on it.
So nowadays, attributing it back to the person
is probably where all of the value is.
It's even more important than what it is,
or maybe not quite,
but the needle is moving in that direction.
And I kind of actually like that
because I would love to live in a world where it's like, I don't need to worry about protecting every byte of data I leak.
I just need to worry about it being associated with me and that attribution being the source of values.
I kind of like that, but that's just my opinion.
Can I riff on that a little bit? I think the attribution is really, really important.
I think the other big question mark here, and you kind of touched on it,
when you think about privacy, why do people want privacy? In many instances,
it's to protect themselves from bad actions that happen. But the other aspect here is like,
we talk a lot about data ownership,
and I think that's the wrong lens to look at this.
I think it's not necessarily about owning the data,
it's about consenting to its use.
It's about auditing where it's being used.
Frankly, it's about revoking permission to its use.
So I think auditability is a huge part of it.
I think that the consent
to how that data and to some extent the control around that data and how it's used and is it being
used for my best interests, that is a really important antidote to that sort of bigger
question of what is privacy anymore. I guess my follow-up question is are there any good systems
today that are being developed or already exist to attribute that data
i mean julia mentioned one right the content authorization
phenomenal example it's really good.
OpenAI does it on every single output they produce.
Adobe baked it into everything they have.
It's in Google and Meta and all of these things use it.
I think it's great.
I feel positive about the way it does this but it's quite fragile um it can be the problem
is it can be separated from the content right and so that's that's not ideal um i think it's a great
framework for attributing uh information to data but i don't think it's a great framework for like
keeping that together or blockchain i think does an exceptional job of these kinds of things.
It's kind of direct association or permanent association. I think blockchain is far has massive advantages there.
I guess for the layman and me being the layman, can you talk a little bit about that?
Because I think I found that I find that interesting and I I know you and I have talked about this already, like offline. So this is I understand it with some kind of very high level understanding of it. What is the difference, I guess, between the way that these systems are currently built and how blockchain could actually improve on that model?
blockchain could actually improve on that model.
So, I mean, I can just speak to how it's kind of broken here.
Um, and then I think these guys would be great to speak to how blockchain can go further.
But in terms of the, the, the way it works right now is, you know, you're basically taking
like a photo or some piece of content, some information.
Um, and we, we basically create some kind of identifier for it, you know, like a hash or a fingerprint
or something like this.
And then along with this, we can say, hey, here's the data it's related to or came from.
Put all this together into a really nice, you know, list.
uh uh list and we take that list and we generate a uh a ctpa uh document from that or json from that
um these are signed uh they're signed by generators they can be validated as well within that network
and everybody can say yeah this this image yes was used to generate this uh this associated uh jason and these are the companies
involved it's related to this work and they would they can all agree and say yep that is all true
and that's great and that works actually extremely well problem is you know um where is that data
stored right and it's in the manifest itself there's no guarantee that manifest stays with that content
everywhere forever right and ai's good at like cutting things off and separating and
segregating information that's that's kind of the point where uh blockchain can come in and say
hmm like you do a really simple thing like if you find this hash uh here's some information
and we can associate
that in such a way where it can't be disassociated permanent right it can be recorded embedded in the
network forever but the internet doesn't work that way and if i associate julian with an image
i mean that's great as long as that association is there but the moment that it's separated it's
like there's no way to find julian again so these are like just that's just one of the downsides of like regular web 2 or i think web 3 you guys can do a way better job
yeah i mean i i agree with that as well i wonder jeff i anything you want to add here before
we keep moving on from the only only Only really quickly, actually, and I find this positioning really interesting.
It's, there's a couple of things
that just struck a chord with me.
And that's one, I just wonder if tagging things
as AI generated is actually the wrong positioning
and actually tagging things as human generated
be a better
positioning. But actually, thinking about my workflow and people I know, their workflow,
how the hell do you start untangling that as we move forward? What is AI? What is human generated?
They are this symbiotic relationship all of a sudden. I cannot believe in two years how much my workflow has changed,
that, you know, what I produce, what I do, how much AI is in there, how much Jeff's in there,
I don't know how you even begin to pull that apart. That's really difficult. And also just
to rewind slightly, I think there's two different questions here. And I think we've been focused
very much on public privacy, you know, when we share things publicly. But I think an
equally important question is when we share things privately with AI models
and so on. I've just gone through a range of health tests, blood tests, I'm
getting old, right? They prod me with needles and all sorts of things. That's
gone through my clawed co-work. I've got a file set up. It has access to it.
It tells me that this isn't very good. This is like a superhero. This is brilliant. This is bad,
but it's all in there. I have no idea how private, I have no idea what they're doing with that data.
None. Zero. Right. And that's probably true of most people in the world using AI at the moment.
And so there is also this private privacy thing, if that makes sense, with what we give to the models that's not
just on the outside. So two things I think will prove difficult. How do you
detangle AI in human content? Because it's going to get more and more tangled. And
how do we differentiate between public privacy and private privacy, if that even
makes sense?
I think it's interesting, and Nick, I saw you nodding your head, so I'll just see if that even makes sense. What?
I think it's interesting.
And Nick, I saw you nodding your head. So I'll just say this briefly.
I find it interesting that you bring up
these two very interesting points.
And one that I was already thinking about raising.
And this is the idea of like,
is there a third data set that is,
instead of just looking at it like human data
and AI generated data,
but it's like collaborative data where it's like produced together.
And what's the value in that?
Is there value in data that is produced together with human intelligence and artificial intelligence?
The other thing that I thought about when you were talking about introducing all of this data onto Claude, which I personally wouldn't do. And more so
because I think that you are sharing this information and it
could potentially have some sort of like unforeseen
maybe medical insurance things that things are shared
without your permission. I'm in the UK, Humpty. I'm in the UK.
We don't give a damn about medical insurance. I'm in the UK, Humpty. I'm in the UK. We don't give a damn about medical insurance.
I'm showing my red, white, and blue over here.
I'm chaffling at the notion of Jeff giving open-claw access to his NHS health records
and just be like, yeah, just go and get me some appointments for blood tests and stuff.
Just YOLO.
that's how i live nick and this idea that that when you were saying that was i remember uh that
That's how I live, Nick.
when it came to try to obfuscate the crypto that you were moving around going to cryptocurrency
um you were putting them into these mixers like tornado cash so that you know your data if you
will of you know tokens being moved around from accounts to account was hidden because of the number of other accounts and tokens that were being put into the same pool and then redistributed.
So maybe in the future, there is a data mixer for AI prompts and results that kind of removes some of that provenance,
that connection between your prompt and the result from Claude.
Just thinking, because I was thinking of Tornado Cashman
when you said that.
But go ahead, Nick.
Did you want to add anything else to that too?
No, just laughing at Jeff in a friendly way.
in a friendly way.
But actually, one point, it's interesting.
it's interesting you talk about this.
I find the hyperscalers and the LLMs
almost insidious in the way
that they kind of gamify ingestion of data.
And it's, you know,
and this is coming from someone
who's spent the last decade
caring about privacy, right?
What's happening in the broader world?
It's mind-blowing.blowing the power the concentration of power
in these hyperscalers i i think it's gonna you know make web2 world look like a
walk in the park honestly can you define hyperscalers for those who are listening
and don't know what that means you know claude know, Claude, OpenAI, Gemini, Grock,
these sort of, you know, large,
that are publicly available.
You know, the guys who brought, frankly,
who brought AI into their public domain, right?
Because people have been working on this stuff
for 20 or 30 years, but it's only been in the last two years
where it's been accessible to the, you know,
to the normies like us because of the UI on top.
But I think, sorry, can I jump out?
I think this is really important, Nick, because people,
if you're not careful, me included, we don't care enough about privacy.
This is one of the problems, right?
And the power of AI, the power of these models is with the data you give them.
The more you give them, the more use they are to you.
The more you share, the more use they are.
And that's the draw, right?
That's the pulling.
And that's the trap we fall into.
And I think this is where, to come back to blockchain,
to come back to blockchain and particularly things like zero knowledge proofs and so on.
and particularly things like zero knowledge proofs and so on,
And I listened to Tim Berners-Lee talk about his ideas about these little capsules of data
and things that were quite private. And I think these are the real solutions we need to look at.
I genuinely believe AI's natural home is on chain with privacy preserving things in place,
is on chain with privacy
preserving things in place
because for it to be
at its best, we need to share
as much as we can with it. It needs
to be embedded and it's little
weaselly hands in everything
we do. And at the moment,
that's Anthropic or
OpenAI or Elon or
you know, that's where it's going. We need
better solutions to that.
I guess that Elon and I Grok. Go ahead, Julian. You know where it's going. We need better solutions to that. I guess that Elon and I Grok.
Go ahead, Julian.
You know, it's like this is just,
this might just be what I'm working on now,
starting to bleed into the conversation.
But I think that your points on context
and what you provide is super important.
And it's actually,
I don't think a lot of people think about it this way.
And I think this would be, people would debate me.
But, you know, I don't think most people
actually interact with LLMs.
I'd say it's actually quite rare these days.
I'd say most people are interacting with agents.
They just don't know it.
You know, and if you go to ChatGPT and you say,
hello, it says, hi, Julian.
And why is that?
So there's context, there's context being injected into that prompt.
And there's now when you interact with an LLM, you have the user prompt, you have, uh, the context.
So this could be a database that we've used in like vector search and re and re-ranked.
You would have a system
prompt, like how, uh, how to respond or a way in which to respond.
And then vendors your conversation history.
So all of this is sent in every single interaction.
And the funny thing is the model at the end of the day probably doesn't, doesn't
very, very, very likely doesn't know anything about you but your prompt or your prompt cache
contains all of this user information and i think this is a really important part that most people
don't think about is that's where the secret sauce is and if you can you can can properly engineer your prompt in that way, the context, your vector search,
your embeddings, your re-rankings, your graphs, your everything.
Like, I got to tell you guys, like, you can get a terrible model to perform better than
the best models that are out there.
You can make your conversations more efficient.
And not only that, like managing that context well,
that's where privacy becomes a big deal, right?
Like where that's where you can control things
and that's where all the value is.
So kind of think the balance of power is going to shift here
and models getting better.
Yeah, it's good.
But I think context is key. I think that's going to be really big.
I think that's the next wave.
I love that point, Julian, by the way, on these context things.
So I've got .md files that has got enough information for you to clone me, pretty much, right?
And that's a really interesting
point right because imagine that as privacy preserving.md files or whatever the equivalent
is for that context i think that could be a really nice future that we could envision
yeah i mean i run a lance database with a local embedding model and a local re-ranker for every single prompt
I send. Just so I can perfectly manipulate it, so I can pay the lowest price possible and get the
best price I possibly can get with providing the absolute bare minimum context about me.
And it's like, that's's like that's like how we're
doing it now but i i think that's where everything's gonna go because these privacy issues you
guys are talking about you're exactly right it has to i think my my my crypto bro is showing because when you say that also, what comes to mind is when you are using a wallet
and you want to maybe reduce
the risk platform
for getting undercut on a trade,
you change your RPC endpoint
to maybe something that you run
versus a public one like Infura right where all of these
other uh you know players are expecting these trades to happen and they're trying to front run
you right um so i i've been in crypto way too long uh because that's literally what i was thinking
when you're saying you're running your own endpoints yeah uh for determining kind of like this information from your agents kind of random hot take here but solving for
you know i'd say one of the biggest challenges in in crypto right is key management obviously
it's key management right and we have to solve that for digital identity we have to solve that for privacy in an LLM world. So I actually think AI, digital identity,
very important from a crypto point of view
because it rises the tide on the importance
of solving that personal key management issue.
Oh, yeah.
Yeah, it concentrates the risks there for sure.
Well, I want to go back to something.
Well, okay, before I i get there another piece of
context because you were talking about this idea julian of you know they're currently you're not
talking to an llm you're talking to an agent i re i was literally at a coffee shop yesterday working
and this very friendly person was asking everybody what they did which i thought was very interesting
and one of the things that we just started talking about was like AI and AI agents and
the different AI tools. And he was surprised at how much I've used it over the last two years,
starting with like GPT 2.5, I think, if not GPT 3. And so there's a lot of history there.
I've been very cautious about how much i share but
definitely enough for it to like learn my voice and be able to like write in my voice when i'm
producing content for instance uh and it's assisting me to do that but recently i moved to
cloud code and all of this context is just gone so in terms terms of like how I've used ChatGPT for two years.
So now it's this idea of like, how does context translate when you're using two different
systems that don't communicate with one another?
Is there a way, potentially, if there was a blockchain based system where they can both
communicate with one another and be able to draw the context?
Probably wouldn't happen because, you know, that is a very, very big piece of intellectual property draw the context probably wouldn't happen because you
know that is a very very big piece of intellectual property that these companies probably don't want
to share but it would be interesting if there's a way to like train based on information that's open
and protect it through some sort of like zk proof potentially but still allows for the flexibility of using any front end like those two products.
You can do that. You can definitely do it. It's definitely doable. You can change from Gemini to
Anthropic mid conversation and it not lose a beat with good context management. And where a blockchain fits in there is,
if the context that you're providing is very complicated,
which it often is a lot of,
to Jeff's point,
could be a lot of .md files, right?
You don't know if that's been manipulated
or changed somewhere in a blockchain
to really ensure the integrity
of that conversation is maintained.
I mean, it's a no-brainer.
I think it's a great idea.
That's really important because actually version management,
integrity management of files is getting harder, not easier,
because people hand over so much control to AIs
that actually knowing if something's changed,
anybody who's toyed with vibe coding
or toyed with content production knows that things can change without you even realizing it.
So that integrity of the file and the integrity of the data that you're even working with
is called into question. But you don't call it into question yourself if you're not careful.
So having something that says, actually, you know, thinking about Julian's point about timestamping earlier
and those sorts of things and that, who did this come from?
Who was the last editor even?
Was I the last editor?
Was Claude the last editor?
That matters really, that matters so much.
It's almost like the LLMs, but the thing is that they're almost,
this is why I say insidious,
right? They're almost encouraging or weaponizing this false sense of, false sense of, of sort of
belief that we're doing the right thing and we've got it right. Meanwhile, it's like,
it's incompetence in many instances, but it's like, no, you're doing a great job. Am I? I think I am. You told me I am.
You're programmed to do that.
Anyway.
I love to be told I'm doing a great job, Nick.
I know they could have that just for Jeff.
Send a few nudes.
To close out, I wonder if we could address the question of monetization.
To close out, I wonder if we could address the question of monetization.
I think one of the things that we've seen come up a lot with emergent technology is like,
how can you monetize either the data or the content that's being produced?
Is there a world in which human actors are participating in the data economy?
And if so, what does that look like?
I know Julian's ready to take a stab,
so he's waiting for one of you all to do that if you'd like. I'll give a quick answer so Julian feels more comfortable
giving a longer answer.
My quick answer is that I think there are two strands to this, right?
Genuinely, I think there's this private, we're giving some data,
and I prefer to be private to AI to do things for us.
They're going to charge us for that, and rightly so in some respects.
I've got no problem for it.
Like, if we give our data over and we're using it to get them to help us do things,
check our health and buy a house
whatever it is i have no problem with paying for that that's that's part of it and i almost feel
like there should be a toggle switch where a really easy one that says actually you pay me
if you want to do something else with this data and i'll tell you where it can go i'll tell you
what you can use it for but this is the the flip side of it. So it's perfectly private when I give it to you.
And then I can choose where you can share it and you can pay me a separate thing.
So I don't mind.
It's a weird symbiotic relationship here.
We are getting a really good service.
It's like Facebook or Twitter or whatever.
They're giving us something, right?
The value of that can be questionable at times.
But they are giving us something.
But what are we giving that and i think that relationship needs to change
on monetization it needs to be a two-way street i think it's going to flow both directions in
the future because they don't exist without our data but actually they're not useful to us without
our data so it's it's it's yeah, it's bidirectional for me.
Cool.
So Nick, any thoughts?
I'll do a quick answer.
Look, I think if you see the world through this new lens,
it's all a big matrix.
Every person, every organization, everything is a data point
or a node, right?
And if you want to, like any other network,
you need the right incentive models and more importantly the right disincentive models to avoid abusing that network and i tend to think
that payment gateways are probably the most simple way to avoid bad behavior in a network so i tend
to think of payment gateways and you the kind of stuff that Coinbase is doing
as their first utility will be monetizing signaling costs
and disincentivizing people to attack one node
or sort of do the wrong behaviors
rather than content providers monetizing.
I think monetization
is more at an infrastructure level um to level the playing field than it is an individual um
you know function of of monetizing data and i'm sure julian will now correct that
i i i think it's like i think it's a really interesting problem. And I kind of think we're at, you know, we're just at a weird, we're in a transition.
It's just a weird transition time from the way things were before, where there was, everything
was on floppy disk and security didn't matter because there's no way to steal it.
Today, when it's like, it's too easy to steal everything and it's actually almost
gotten to a point where it's so too easy it's like actually kind of impossible to steal anything now
we're getting to a point where it's very difficult too because it's coming back it's coming full
circle and that's because you know for an ai like that's trying to train off of data
they they can't produce random garbage that they can't train off of their
own content. That's not going to work. They have to train off of like real content. And so it's
becoming more and more difficult to figure out like what's from a person, you know, like you
could build an agent on Anthropic to talk to chat GPT and provide a bunch of health records that are synthetic, it would not know.
It would have no idea.
It's very good at it.
It's like it would have no clue.
And this is real.
And I guarantee they're thinking about these things.
These are real problems.
And they're like, okay, we actually ate the world.
There's nothing left but us.
Where's the new, where's fresh meat gonna come from?
And the funny thing is it actually shifts,
the power shift actually comes back to the person.
But like I was saying before about this Bloomberg concept,
there could be a world, a future
where Bloomberg doesn't protect their data.
Everything is just given away.
They post everything on Reddit,
everything open. The only thing they protect is when they say, buy Bloomberg. And so they don't even actually need to protect, to produce the data. The data could come from Yahoo Finance or
something. The moment they put buy Bloomberg, that's when it's worth something. And that's
pretty magical because that data is so much smaller.
It's like the attack surface is nearing zero, right?
And that's something defensible.
And the other thing is there are signals today.
There are clear signals today that AI companies will pay for that.
Like we were talking to an artist recently that was working at what was it meta
and uh they're paying five hundred thousand dollars a day to artists to generate like 3d
models to train off of because you can't get it like how do you find real like it's like it's all
i don't know if this is
real or not. I don't know if it's good or like, might as well just pay people to make
it in and that's that power is going to shift even more and think about that, that puts
that puts the power in like, the person, the person and the most valuable thing, your person
hood, which is very easy, which is much easier to protect than my actual data.
And I think that's really a cool future where I could see that being like needed because AI can't
detect what's AI or human. It's not possible. It's pretty much very difficult for us to do
impossible for them to do. So Julian's just made my day i've just paid
to put my elders through university as a 3d modeler so that that has made my day thank you julian
i feel like i've not just wasted a hundred thousand pounds on education i i think it's a bad i think
we're in a transition time and now it's like you can make all this stuff so easily, which means it's like, you know, I don't wear a digital watch.
Digital watches probably keep time.
I have a digital watch.
But the digital watch would keep time much more accurately than like this.
But this is made by...
The analog shift is real.
And I love the optimistic note on which we are leaving.
Because I think one of the things that I talk about personally to people is I but also certain things that are overplayed in terms of what AI can do.
But also in terms of this conversation, how we closed it, there's still an opportunity for humanity to thrive by simply being human, which I think is a wonderful thing to think about.
to think about. Well, I mean, we've reached the top of the hour here and I want to give an
opportunity to all of our wonderful speakers, an opportunity to say, where can we follow you?
You know, what is something people should look out for that you are, you know, excited to
share in the near future? We'll kick it off with Julian. Oh, sure. Yeah. Thank you so much for
having me. I always love just the conversation is great. We always have a good one. You know, Twitter's fine. I guess I'm on LinkedIn more often these days years. However, I'm really pleased to announce
we're actually gonna be launching
our first ever public facing product
within the next coming weeks called Monolith,
which is only this, which actually does exactly this.
It's if you're an artist,
you can take credit for your work and AI will be able to see that and recognize that and has to pass to.
We use cryptography, blockchain, everything I've mentioned here.
So keep an eye on that and you'll probably see us posting that soon.
We do a lot of work with artists today, like mostly guilds, think of it like that, unions, some studios.
And we're really excited to get this thing out there and make it accessible to everybody.
Awesome. What are those links, Julian?
Sure.
The first one is joinmonolith.com. Probably a good one. There you go.
Cool. Thanks, Julian.
By the way, having worked with Julian in the past, having him announce a release is like pulling teeth.
So I'm glad you were able to share that comfortably.
Yes, thank you.
Nick, where can we find you and anything exciting that you're working on now that you'd like to share?
So, yeah, find me on LinkedIn, similarly.
Google Nick Riss, LinkedIn, and it'll come up.
Yeah, what am I excited about?
This tipping point.
If I mention this sort of, you know, the Gartner lifecycle growth curve
at the beginning of the conversation,
the market forces driving the need for digital trust are phenomenal,
whether it's governments digitizing, whether it's the advent of AI.
It's causing this pressure to build up.
I spoke to a telco last week, global telco, 200 billion cyber attacks a day.
This pressure is just mounting. And there's a lot going on
in the decentralized identity space, in the digital trust space. A lot of projects we're
working on that are not necessarily public, but they're game-changing programs. And the reason
they're so useful is they become the proof point that this new approach to digital identity, this new approach to how you do trusted
data online works, and it's no longer like innovation theater. Like this is real stuff,
like MasterCard and Google announcing their verifiable intent uses verifiable credentials,
right? These are all really good news for folks who have been working in this space for so long. So I'm just, this first quarter of the year has been outrageous
in terms of the progress that's happening, largely under the waterline, right?
We're in our pre-chat GBT moment for Digital ID.
But there's just so much going on.
So, you know, keep watching the space.
There's cool stuff happening.
Awesome.
Jeff, any last words yeah
you know keep up with us on ontology and onto wallet um thinking about consumer facing things
on data and data ownership uh this can be really exciting once that starts moving forward so and
you know where i am always drop me a message. I always say hello. Awesome. And as for me, you can find me usually on X, Humpty0X.
And I guess we're dropping our LinkedIn.
It's Humpty Calder on there.
Happy to chat with you all.
I definitely am an ambassador and an advocate, I should say, for creators.
So lovely to put my creator hat on again today and host all of you lovely people and have
this wonderful conversation.
Excited to see you all again next time.
Thank you, everybody.
