Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
AI FRONTIERS: COVALENT X NILLION
Recorded: Jan. 17, 2025 Duration: 0:28:54Space Recording
Short Summary
The transcript is a discussion from the AI Frontiers series featuring Tristan, co-founder of Nillian, and Ganesh, co-founder of Covalent. The conversation covers Nillian's focus on privacy-enhancing technologies, particularly in AI and distributed systems, and their upcoming mainnet launch in February 2025. They discuss the importance of privacy in the crypto space, the potential of personalized AI agents, and the challenges of conveying complex tech concepts to a broader audience. Nillian is also working on partnerships in the healthcare sector and exploring the use of agents in AI. The session concludes with a call to action for developers to explore Nillian's new documentation and demos.
Full Transcription
Good morning, good morning.
Good morning.
Okay, we'll just wait a few more minutes for people to start rolling in.
But hi, Tristan, it's great to chat with you.
Hey, how's it going?
Is my audio okay?
Yep, loud and clear.
Yeah, doing great.
How are you?
I'm doing well, doing it well.
It's a bit back to back these days, but exciting stuff.
Yeah, that's great to hear. Yeah, I just want to thank you for taking the time out of your probably very busy schedule to have a chat with us today.
So, yeah, I really appreciate it.
Yeah, for sure. I mean, I think there's always fun stuff to talk about with the white space and the AI agent stuff.
So I think it's fun to get a little bit more speculative, which is what AI lets to do about what the future might look like.
Yeah, exactly. And that's what we aim to do with this new series that we're kicking off is to, yeah, just start the conversation and have like really good speakers to kind of join us to talk to travel a bit more about this kind of stuff. So yeah, I guess we can get started.
Yeah.
I'll kick things off.
So yeah, welcome everyone for those who aren't familiar with my voice.
My name is Dalia and I'm part of the marketing team here at Covalent.
And I just wanted to welcome you all to our first ever AI Frontiers series.
Oh my gosh, that's a tongue twister.
But this is something brand new that we're kicking off this new year.
Where each week we're aiming to bring on special guests from different projects and
to discuss everything there is to discuss about blockchain and AI.
And I'm joined by Ganesh, Covalin's co-founder, as your co-host.
And our goal for this series is to bring our listeners basically just any sort of alpha or insight from the biggest players in the AI space.
So let's kick things off.
We're, yeah, extremely excited to start it off with our first guest, Tristan, who is the co-founder and director of crypto at Nilean.
So thanks to both Ganesh and Tristan for joining me.
So, yeah, before we dive into the questions that Ganesh will have.
Good morning, everyone.
I'm your host today, Jane Harrell.
And today,
Jane, Jane.
Jane.
On AI Frontiers, Covalent and NILIA.
Sorry, Jane.
We have, of course, Ganesh, co-founder of Covalent,
and co-founder Tristan from Nilean.
What is...
So, for those of you aren't familiar,
I'm part of the marketing team here at Covalent, and I wanted to welcome you all to our first ever AI Frontier series.
This is something brand new that we're kicking off.
What is there something off here?
This year, and we'll be bringing special guests from different projects to discuss everything there is to discuss about blockchain, AI, agents and all sorts of things.
Our goal for the series is to bring our listeners some incredible alpha from the biggest players in this space.
So let's kick this off. We're extremely excited to start the series off with these guests.
And thank you both for joining. How are you both doing today?
There's some kind of cross talk here. It's so weird. She's not even on this Twitter space.
That's so funny.
Yeah, let's continue.
Something is having some technical difficulties.
This is a classic Twitter space moment.
Mesh or Tristan.
I'm so sorry, everyone.
Just give us one second.
We're trying to reach out to her right now.
Yeah, this is exactly how I find Twitter space tech works basically every time.
So there's always a little bit of a little bit of rugging at the beginning.
Yeah, we're trying to reach out to her right now.
Give us one second.
Okay, I think that should be good. Sorry about that, everyone.
Yeah, okay, well go ahead, Tristan, sorry about that.
Can you just maybe take some time to introduce yourself and Nilely into our audience?
Sure, yeah, happy to. Hey, I'm Tristan or Crayon. I am a co-founder of Nilean and I'm more focused on the crypto stuff. I guess I'm the closest to the trenches of the co-founders.
Nillian is building what we call humanity's first blind computer.
Blind computation is really just a moniker for different kinds of privacy-enhancing technologies as applied to distributed systems, AI, things like that.
And what the blind computer is is,
basically trying to abstract all of the like three letter acronyms, all of the complicated cryptography, difficult to use stuff, and bring all of this kind of new privacy tech to people, to builders, to developers in a way that, you know, makes it easy to have privacy for everyone.
We have a lot of focus on AI, both on the research side.
So we put out, for example, a cool paper with the research team at Meta a couple of months ago on privacy preserving LLM inference using multiparty computation.
And then also on the practical side, we have a few different what we call blind modules that you can go check out in our docs or check out demos that can be used in various ways as part of the kind of AI supply chain.
Amazing. Thank you for that intro.
Tristan, this is Kanish Swami, one of the founders of Covalent.
So let's get into the specific Q&A here.
Question number one is, what is the exact gap or challenge you've noticed in the decentralized or data space
that has inspired you to build Nilean?
Yeah, I think, well, I think it's...
quite complicated when it comes to privacy because of two things. One, in crypto, privacy for a long time has really only meant transaction privacy. And transaction privacy is super important. And, you know, I have great respect for and think it's very important that that stuff gets built.
But we haven't even really broached the subject of things like data privacy in the space because you have this sort of radical transparency that Bitcoin, you know, brought with it and a lot of the projects since have brought forward.
But privacy, notions of privacy are really important for really almost everything that we do on the internet.
And I would even argue, you know, off the internet.
And so that's a lot of what pushed Nillion into existence is us seeing this problem and saying,
okay, if we even want to be at feature parity in crypto with what you can do right now with
big tech or like on the normal internet, right, you will need the notion of an account,
my version of the internet, my algorithm, my DMs are very different than yours.
I can't just have like root God access to all the whole internet, including in your version,
And that was a big piece of it.
And then the second part thing that that makes it a little bit harder to,
to, to think about privacy is that we think that that we're basically at an inflection point now because of AI where,
A lot of people, rightly so, have looked at privacy and said, wow, there is like, you know, everyone's chosen convenience over security every time on the internet.
Why is this time different?
And we think that we're in an inflection point that is different because we have this notion of three levels of privacy.
Level one is...
public stuff. Level two is stuff that probably should be private, but you've already kind of given a way to Google or the government or, you know, whoever, right? And then level three is the stuff that today is still private, or at least not centralized to any one big player, but that we think that the kind of next phases of AI, personalized AI,
the coolest stuff that people want to be able to do in their lives with AI will require putting that all in one place.
And we think, you know, this is sort of a make it, break it time for not getting relatively dystopian, I would say.
So those are some of the big things that pushed us to build a million.
Very cool. Now, if I were to draw an analogy here,
Millions technology is not constrained to blockchains and on-chain specific stuff.
There is the training piece and the inference piece.
Now, tell me...
How is this different from something like HTTP request privacy,
where you can definitely inspect HTTP headers,
and isn't that kind of the deal with the LLM inference as well?
You obviously cannot get the results because that's encrypted party to party,
but the headers and the queries themselves,
are they not visible publicly?
Are you saying in-nillions technology or kind of in general with the status quo?
With the status quo.
Yeah.
So I think with the status quo, the notions of privacy, like I mentioned, are basically
built on third-party trust assumptions.
So if you talk about LLM inference and you pick one of the big providers like OpenAI or
Anthropic or what have you.
You, if you log on to your open AI account, cannot see my responses, the, you know,
information that I put into my prompts and the results that come out.
But that notion of privacy that is very useful is built on a third party trust assumption,
a trusted third party that in this case is, you know, Sam Altman's machine, right?
All of the people that work there that have different levels of access, et cetera, et cetera.
HTTPS, as you say, is like it's a transport layer privacy where from them to me, you know,
even if you're inspecting my packets on my, like I'm using Starbucks Wi-Fi or something,
right, you won't be able to see it.
That's great.
But what about if you want to get rid of that third party centralized trust assumption,
that's where the transport layer that you're talking about breaks down.
Incredible. Yeah, I think what's exciting about what Nillion is building is that the
arc of technology, especially on the LLM side,
is trending towards open source and potentially
self-hosted LLMs.
If you look at the new models from Microsoft,
obviously LAMA's models, they're all open source.
The parameters are open source, the weights are open source.
So that's pretty exciting.
And with decentralized inference, you can also
I guess, like federated inference and these things are actually in Nillian's favor, which is super exciting.
So it's not like a technical barrier just yet.
Incredible.
Okay.
Tristan, the next question is more on the softer side.
What has been the most unexpected lesson you've learned as an entrepreneur while building Nillian?
And how has it changed the way you approach leadership or decision making?
Wow, that's a heavy one.
Let me take a moment to think.
I think the biggest challenge that we as a company in this space have faced and especially
with the kinds of things that we're building, although I think that this is for most,
let's say infrastructure layer, or maybe even most projects in crypto, if not in tech, is
is how do you tell a story and how do you, you know, just get people to care or even maybe more importantly understand what you are working on, what the problem is and what the solution is?
While also, like, they're not going to give you that much of their time.
And a lot of them are not, you know, they don't have the prior knowledge to really understand it at the level that,
you want to need to explain and differentiate.
So what I mean by this is at the beginning,
when we were kind of first launching Nilean, right?
We were talking a lot about MPC.
We were talking a lot about, you know,
pets at a high level privacy enhancing technologies
and YRs are faster or better or whatever.
And over time, what we learned is one,
we were trying to differentiate in a category that didn't really exist in people's minds,
like in the masses minds, right?
It was a very niche community.
And two, even when you started getting broader and maybe more left curve,
like dumber with the way that you're speaking, and I say this lovingly, right,
making it more understandable so that I can actually explain it to my parents,
then it forces you to really pay attention to what matters about
the product or the offering, right?
I think that there is in the space a lot of obfuscation that can happen when you have a lot of words that sound cool,
a lot of three-letter acronyms and like kind of tech terms that you coin.
And trying to convey the problem space and the benefits that this could have, you know,
if it's built out, the kind of world that you're trying to usher in,
requires you to really focus on no, but like why, right?
Like if you take away all of the scaling words and all of the cryptography words and the whatever,
it forces you to really position the why and make sure that the why is always interesting to people.
And I think that doing that has been a continuous and continues to be a continuous battle.
And I think one that we've learned a lot about as we've built.
Incredible. Okay, I think that's a lesson for all entrepreneurs and founders, including myself. Thank you, Tristan, for that. The next question is about builders. So Nillian enables...
features and unlocks and capabilities that were previously not possible.
So what kind of collaborations or use cases are you most excited about seeing on the Nillian platform?
Yeah, I mean, you know, we get this question a lot and it's always a harder one to answer from the perspective of million rather than as an individual because if you ask different members of the team, there's really different stuff that I think drives them.
I think one of the most common ones internally that people are excited about is healthcare.
I think that there's huge unlocks in the healthcare space that can be done with, you know,
privacy enhancing technologies and especially the intersection of privacy enhancing technologies, you know, data ownership and AI even.
And we, you know, we're doing, we're setting up some really cool,
For example, like a large-scale sleep study is in the works and some really cool partnerships with teams that are working with, you know, consumer health care data in interesting ways that I think are really exciting.
And then the other big one that comes up a lot is, you know, agents.
I mean, it's all about agents everywhere, right?
But we are the people who head up our product team are very AI-pilled and agent-pilled, I would say.
And it's...
not just thinking about like the single personalized AI that can have access to all your data without it being like this dystopian world and do all the stuff that you want it to do, but also having like teams of agents, like armies of agents that can do your bidding, right, but still have policies and still have like keep your data private and also keep them honest to make sure that you're not causing havoc with, you know, your data or just out in the ecosystems that you want to be touching with these things.
I think those are the two biggest spaces that really were excited for people to explore and that we're talking to lots of teams and working with lots of teams to build into.
Incredible. Yeah. These two points are something that, you know, I think it's going to be super important, not just for Nilead, but the space as a whole.
All right, the next question I have is about the LLMs and the AI and the data that is feeding into these systems.
So the LLM is only as good as the data that it has access to.
So how is millions privacy first infrastructure
allowing these agents specifically to interact with the datasets on a,
just, you know, richer data sets in a privacy preserving fashion.
Yeah. I mean...
You asked that question with the agent lens, but I actually think that the most interesting part starts earlier.
The most interesting part is how do you get the data?
Like these agents to be useful to move on from, I wrote a little article about this the other day,
but to move on from just like Twitter bots, right?
And like what I would classify as, um,
you know, entertainment style agents to utility will require them to have, you know,
access to more sensitive data, better data, because it's all kind of a garbage in, garbage out
scenario. And getting better data to be able to, one, train, you know, better models on top of it,
but also two, to have access to your own data,
will require privacy, but also will require incentives.
And I think that's another big piece of why
crypto is an important part of the puzzle here
is that a lot of the companies or protocols building on us
are kind of trying to figure out ways to bootstrap
data sets that are valuable while allowing the,
the users who are bringing this data to the table
to retain kind of sovereign ownership of it, right?
Moving from this default
transactional, I have to give you my data just to be part of the internet and like
participate. It's sort of table stakes. Instead, you have this consent-based system where users
have a lot more sovereignty. And with that, you can create richer and richer data sets because
you're giving more security and more agency to the user to say, like, well,
Well, no, maybe I wouldn't sign up for this service because I look at the terms of service or, you know, unlikely they do that.
But maybe more likely I try to download the app or the Chrome extension or something and I see like, hey, it can see everything that I see it's put on every site.
Maybe I don't want to do that for this, right?
And instead, you're moving to this place where, you know, they feel secure because they know what the, they know the blind computation is there, is there protecting them.
And with that, you can build richer and richer data sets.
And those then help to inform, you know, better agents, more utility.
Incredible.
Okay.
So let's start down the controversy path.
Tristan, why did you share something that you think is going to, how do I phrase this,
dramatically change or what the industry is getting completely wrong?
in the next 12 months and in the next five years.
Oh, man.
That's another heavy hitter there.
Let me think about that one, too.
Twelve months and five years.
I think the 12-month time horizon, there's a lot of focus still on...
I think there's a lot of focus on like individual agents and, you know, monetizing these individual agents.
And even, you know, agent frameworks, I think that's already a little bit more in the right direction.
But I really do think that a lot of this stuff is going to be a lot more personal and personal.
built off of people's personal preferences and their own, you know, whatever it is, right?
You have a financial agent.
It's somebody's own personal view on portfolio allocation, risk tolerance, et cetera.
If you have a Twitter agent or like a, let's say, general social persona agent,
they want access to all of your different social medias, all of your different,
messaging things, telegrams, whatever.
But what they, what you actually are trying to do is very different depending on if you are
I don't know, an alpha researcher or you're a VC and you're just trying to stay on top of things or your whatever it is.
And right now, I think the game is still more built around, like I said, these entertainment agents.
But even moving a little bit beyond that, just like these sort of singular entities as agents, you know, a lot of people talking about sovereign agents.
But I really do think that the most useful stuff, like the things that will have the most tangible impact on, you know, people's lives will be these more personalized agents, whether that's run locally or run using, you know, centralized infrastructure or decentralized infrastructure.
And that's like part of what we're building into because we think to run a lot of that stuff on decentralized infrastructure, well, does require, you know, blind compute to work.
And then the five-year question, that one, I mean, man, if I had a strong five-year thesis rolling in crypto, I would, you know, I'd be richer than Warren Buffett, right?
I think it's too hard to know at this point, both because of the uncertainty of AI and the uncertainty of crypto and, frankly, the uncertainty of the world these days to project out that far.
That's a fair point. Chat GPT, it seems like it's been around forever. It's only two years old.
So yeah, yeah. I mean, even, you know, we, we've been, we were looking into, I remember when we first launched our white paper, we had something in there about privacy preserving machine learning, we called it, right? And we were like, this could be cool, right, for things like running, you know, proprietary data set linear aggressions or whatever, right? And then since then and now chat GPT came out and, you know, the everyone.
At Nillion, but across the industry and maybe across the world has updated on what the future might look like pretty aggressively.
Incredible. Okay, so we're coming up on time here, and I'm not going to let you go without leaking some alpha for our listeners.
FEP 2025 is a major milestone on Nillian's roadmap.
Can you please share a little more about this and what this moment means for you and your team and the industry?
Okay.
Yeah, I mean, we're working towards main net.
It's kind of a crazy rush.
We have a big announcement coming out on Monday that you guys can watch out for if you've
been following the project for a while.
And otherwise, you know, it's a lot of making sure everything's in the right state for
us to deliver this as the network, but also working with a lot of the
founding, you know, entrepreneurs, people building on top of us that are going to be using real user data and building real applications to make sure that they're in a great state as well.
And so that that is, I think, something to keep an eye on as well is, you know, who is actually building on Nillian and is ready to go out alongside us.
Exciting. So all listeners should set up alerts on both covalent and the million corporate accounts.
And looking forward to the announcement on Monday.
And good luck on the main launch in February, just a few weeks out.
Any closing parts, Tristan, before we wrap up for the day.
Yeah, I mean, I think a call to action would be if you are a developer, or even if you're not and you're just, you have ideas.
Go check out our new docs.
Go check out our demos and try and build something.
We put out something on our build on Nillian account recently that shows one of our engineers using Replit to build a Nillian app with just prompts like super quickly.
And it was the, you know, the very first app that, that, you know, we ever or I ever thought of.
It's like a very simple password manager app, but like no code, right?
Just asking the computer saying, please, and it delivered it to you.
So even if you think you can't, you know, go out there and give it a try.
You might be surprised at the state of the tools today.
Incredible.
So just from concept to prototype to production code, just using prompts.
Well, I wouldn't call it production code, but certainly something that you can work off of and polish into that.
But a fully working kind of POC from just using prompts, you can go check out at Build on Nillian, and it's one of the most recent videos there.
Tristan, you have more faith in soft-aging than I do.
So I'm pretty sure it's going to be pushed to production right away.
All right, folks, that's it for today.
Tristan, really appreciate you sharing your wisdom and your insights with our community
and excited for what's on the roadmap.
Thanks, everyone.
Take care.
Happy Friday.
