Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Making defi accessible
Recorded: March 27, 2025 Duration: 0:50:45Space Recording
Short Summary
In a recent discussion, industry leaders explored the evolution of DeFi from its nascent stages in 2017 to the present, highlighting significant improvements in user experience, the emergence of MEV awareness, and the ongoing trends towards more sustainable token economics and governance structures. The conversation underscored the importance of prioritizing usability to drive mass adoption in the crypto space.
Full Transcription
Thank you. Hi Anna, one second we're just waiting for Cain to join.
Yeah, no worries.
Hey, how are you doing?
Yeah, good thanks. Thank you. Welcome everyone.
For those just joining, we're just waiting for Cain to connect and then we'll get started.
Nice to see some Infinix hats on in the audience. Thank you. I think for next time I need some background music to play or something like that.
Cain's here now, so I'm just going to bring him up as speaker.
Welcome.
Is your mic all good, Cain?
One, two, check, check.
Yeah, perfect.
Okay, welcome everyone.
Thanks so much for joining.
Thanks, Anna.
Thanks, Cain.
The space today is about making DeFi accessible.
For those that don't know, Anna, I'll give you a quick moment to introduce yourself.
Hi, everybody. Nice to meet you.
Yeah, I'm a co-founder of Cow Protocol.
And yeah, Cow Protocol launched 2021 in May.
And initially, we were still part of NOSUS, then we span out and
have been keeping building on it ever since.
And so much about me for now.
Everyone should know who Cain is.
Well, at least the 10 people that are in here, hopefully. Unless you've accidentally stumbled in.
But what time is it for you guys?
It's like midday or something in Australia, right?
Yeah, it's midday at the moment, yeah.
Late in Berlin, huh?
Well, thankfully I'm not in Berlin right now.
It would be like 3 a.m. in the morning.
Yeah, 3 a.m. I thought it was aggressive.
I was like, wow, okay.
No, no.
I mean, we are dedicated, but it has its limits.
Straight from Bergheim, right?
Exactly.
No, no, I'm in Miami right now, so that makes it much better.
Amazing.
So I think both of you have been around this space for a while.
Before Caldow, Anna, I know you're at Gnosis.
I think the first point I thought would be good to talk about is how the experience of
using crypto and DeFi has changed since the 2017 era to now.
It's almost completely different.
But what do you think the biggest evolutions have been?
Oh, gosh.
I think back then it was ESA Delta, right? It was like one of the
few DEXs around. It was quite... And IDEX. IDEX and ESA Delta, one of the two. Right. Definitely
very different experience to now. Definitely still facing a lot of the initial early issues of, yeah, basically trying to work around the on-chain
environment and like the slow transaction speed. And I think, yeah, in the last few years,
products have evolved to work around those hurdles. But yeah, I don't know, did you use
IsarDelta a lot? I just remember that I had some issues with like huge price spreads because market
makers essentially had to provide liquidity and then had to pay gas costs every time they
wanted to update a quote. It was quite something else.
On-chain order books were a mistake, I think was the conclusion from that. But yeah, I used it all the time actually
back in the day. I thought it was like, I kind of, I think back then everyone would
sort of use things, even if it was painful, just for the sake of using it, right? To sort
of see, you know, it was like an ideological thing.
Yeah. And then, I mean, it was also fun to experiment around research.
Yeah.
But definitely at a very different scale compared to today.
And then I think Uniswap wasn't even around at 216.
I think it launched like in 218 only.
And then obviously Uniswap, I think, was the first one that got more mass adoption.
That still obviously was like the pain points of not providing a good solution to set a large volume trades.
And there was still this pain point of when gas costs were really high and users were trying to place a trade.
Transactions didn't execute and they still had to pay $200 for a
fair transaction.
So yeah, I think your X back then was pretty wild.
Yeah, I think, I think it's interesting, um, you know, going back
into like OG, you need swap interface.
Uh, it didn't even have slippage protection.
Um, and, or, or I think it may it didn't even have slippage protection.
Or I think it may have had a default slippage protection,
but you couldn't modify it.
And so for the first month or something like that,
everyone was getting sandwiched.
It was one of the first examples of MeV in the vile.
Yeah, it's surprising actually how long it took
for people to get higher.
Obviously, the OGs were understanding the issues,
but I think the wider user base of Uniswap
didn't really understand MEV was a thing
until it became much more known,
I think in the late 2020 that people started talking about it.
Obviously, it had been the Uniswap design
as prone to Surface MEV,
but it wasn't really discussed as deeply.
And I think users have recently become much more aware
of hurdles in the space
and have hence also demanded more protection against
I think one of the most interesting things to me about that time period in Ethereum
was again, like there was a level of sort of naivety, I guess, around how things would
play out. And I remember having discussions with people and going like, okay, you know, there's this
like theoretical concept of MEV, but, you know, it's not really a practical thing that's
going to happen.
You know, it's like, there's a lot of theorization of like, oh, this could happen or like these edge cases.
But, you know, on chain, you didn't actually see MEV happen in the wild, right?
And then all of a sudden it did.
And we're like, oh, this is a huge problem we're going to need to solve.
And I think everyone was caught a little bit flat-footed because there was this sense that like,
if MEV was such a big deal, people would be doing it.
But the reality was there wasn't much value to be extracted. And all of a sudden, the value to be extracted,
like it probably seemed to us at the time, like there was, but there actually wasn't in the grand
scheme of things, right? In an efficient market sense, for anyone to actually take the time to
really like figure out how to systematically extract it you
know people were uh were like manually sandwiching orders right it's that crazy right like they were
like reading the mempool and and you know manually sandwiching orders back then and and then all of
a sudden it became industrialized and then everyone was like oh wow okay this is actually
a practical problem we're going to need to solve and And, you know, it was, again, a little bit of a sort of reactionary response, I think,
to that situation.
Because, you know, it had been years of, like, the network was live and, you know, no one
had really seen MEV happen.
So I think, you know, today, obviously, the market's much more efficient on Ethereum.
And I think the big moment where it became really obvious that MEV was this huge thing
and a huge problem that was going to need to be addressed at the chain level was the COVID crash.
On the day of the COVID crash, the amount of ARB and MEV that was happening was just insane.
I remember the situation with the maker liquidations and one of the liquidation bots going haywire and extracting a bunch of MEV out of that.
They were sort of at the right place at the right time.
It was just such a crazy period, that one week window around the COVID crash. And I think after that, everyone
was like, okay, we really need to like address this problem.
I think that sort of follows along nicely into the idea because a lot of MEV solutions actually end up sort of centralizing things
to hide order flow.
So I guess how much is it okay to sacrifice decentralization
to make something more usable?
I think, I mean, I would answer that granularly I think so car protocol obviously took some
decisions to not be fully decentralized at least at the beginning the reason is more so that
initially before we worked on car protocol we had we were working on two previous DEX designs that were both fully on-chain,
but they also came with some hurdles.
For example, we had one that we called NOSIS Protocol,
where users had to deposit and withdraw on-chain
and then had to tap into MarketMaker liquidity as well.
And that basically obviously anything
where you have to like pay in order to be able to do a transaction later is very against like
easy user onboarding and like ensuring that the user is like expending minimum costs for maximum returns. So I would say it's like our approach then essentially was,
let's try, let's go back to the drawing board,
and let's not put decentralization first,
but let's really put user experience first,
and come up with a product that is smooth, easy to use,
and really maximizes user benefits and for that we were
willing to cut some corners on the decentralization however we do work currently on a full
decentralization roadmap so it's still our goal to fully decentralize the protocol it's just that
we felt rather than starting with fully decentralized first and then trying to get
adoption it also takes much longer obviously to build something starting with fully decentralized first and then trying to get adoption,
it also takes much longer, obviously,
to build something that's fully decentralized.
Let's rather find a product first,
define it, get it really fine-tuned to match the user needs.
And then afterwards, you can granularly decentralize it over time.
And that's currently what we're working on,
but it's definitely still going to take us,
I would say to full decentralization,
probably still a couple of years.
I still remember the first time that Martin pitched me on Gnosis Protocol
and the call that I had with him where he walked me through it all.
And I was like, oh, wow, this is like such a cracked scheme to like handle price discovery.
You know, this like round robin auction format, et cetera.
And I think the conversation that we had at that time was, okay, this is really cool, but how do you get people actually trading on it?
Because there was a lot of friction to get people to actually engage with a protocol like that.
that a lot of the OGs, you guys, there's a bunch of people that have kind of come to
this conclusion, including myself, that the primary thing that we need to solve for is
to kind of make these things as accessible as possible to drive adoption.
And then we can go, okay, there's actual genuine need for this.
You know, I think Gnosis has a long history of like cool experiments, you know, very, very experimental things that haven't necessarily got adoption.
And the big shift in CalSwap that I think you guys did was to kind of put adoption first, right, to build a thing that people would want to use, rather than, you know, building the kind of like theoretical construct of how it should be, which I think is a big
shift. And, you know, obviously, it's worked, right, it's been, it's been much more successful
than some of the other experiments that Gnosis ran, these very like theoretical, you know,
kind of ideas.
Yeah, I mean, the best example is that initially we were working on Dutch exchange.
We also wanted to have basically on-chain price discovery, where you would always start
an auction at twice the previous auction's outcome price.
But so you basically have to make sure if it's fully decentralized that you leave enough room for if there's like market price movements that you still start the auction high enough to ensure that you don't accidentally sell at a too low market price.
And then in order to put enough buffer for these price movements in,
the auction took six hours.
And obviously no user is willing to wait six hours
until the trade is being settled.
And I think that might even be something that we still slightly struggle with,
not so much.
Like how protocol now is also batching transactions
to allow for better
price improvements to users so it's also slightly longer than a direct uniswap trade but it's just
a few seconds longer but even for that sometimes there's like some user feedback complaining that
it's taking too long but i think there's also a problem of generally like this whole question of decentralization and UX,
like how, what, what is like a quick enough design that you need to deliver for users?
And what are like the, like what, what are like design compromises you're willing to take?
I think for us, you know, the, the question is like, okay, what are the compromises we're willing to make?
And the thing that we were willing to compromise was censorship resistance over being custodial.
So we kind of maintain the non-custodial approach, but we accept that there's a whole bunch of centralized components that are facilitating,
you know, these interactions and they could be nuked for sure. And if they were, then people
wouldn't be able to do that stuff anymore, but they wouldn't lose access to their phones.
And so it's like, you know, the existing protocols that are on chain, you can interact with them
through either scan if you want or whatever. It's a really bad experience, but they are there and your funds are still there. We're just putting
a really nice wrapper on top of it to make this easy as possible. And you can see that in even
our integration of CalSwap. It's the best way to use CalSwap in my personal opinion, obviously,
but it's by far the best way to use CalSwap. Like you're not signing things, you're not dealing with, you know, like bridging or anything
like that.
You can just very easily get, you know, get access to this very deep liquidity without
needing to understand anything about different chains or different assets.
Yeah, I think this is really so important for the design space going forward that I guess users today, they know enough about different networks and there's a lot of users who actually jump between them.
wave of adoption of new users trying it out like I would I do not foresee any of
them having a network button adjusting it and like knowing what type of tokens
they hold on what network but they essentially just have this very simple
UX design I have my wallet I have all assets in it I don't care what network
they're on I don't care what network I'm trading in. All I know is I'm holding these
assets in my wallet and that's it. As the next thing we had to talk about, I know earlier on,
I was really active in sort of 2021 bull market was when I was most interested and sort of when
I got sucked in and governance was
a massive deal then like I felt like all of the alpha was in like make it our governance forums
and people were really interested in what was going on with decentralized governance all the
time and DAOs sprung up for everything it feels like more recently it's sort of taken a back leg
and in general especially the more retail type users are much less interested
in following decentralized governance. Do you think that that's going to continue or
is it just the natural progress that sort of the nerds will be interested in governance
and it's not for everyone? Or do you think we need to sort of bring it back and get everyone
involved again? I guess I can take this. Honestly, like I'm so tired of governance. I really am. I think It's one of those things where you want it to be,
you want it to be something that you don't have to worry about. Like you want to design a governance system such that,
you know, it just works and you handle the edge cases
and exceptions, but that's a really hard problem to solve.
And, you know, there's a lot of value at risk
and it's easy for people to come and and you know there's a lot of value at risk and it's easy for
people to come and try and exploit governance um you know i mean we saw arguably uh the type of
issues that you have if you don't have you know um well through both through governance structures
today even with hyperliquid right you know rightly or wrongly whatever whatever they did today, you know, it was very obviously centralized, right?
And that's fine, provided you believe that the decisions that were made are legitimate.
And this is like this idea of legitimacy is the thing that, you know, for me with governance runs
through every single decision for every single protocol is like, are the decisions being made legitimate? And so, you know, to,
to kind of, um, illustrate that, I suppose,
if everyone who was holding the hype token had, uh, you know,
going onto snapshot or whatever and voted for Gaffa as the benevolent dictator of
the protocol and he can do whatever he wants,
then his decisions are definitionally legitimate, right?
Like everyone said, hey, you know, we've elected this guy and he's going to do whatever he wants.
Even you could go so far as to say, we're going to elect him for life, right?
And we can never unelect him.
We can't do anything about it, right?
But, you know, we're locking in.
And at that point, anything that Jeff does is, again,
definitely legitimate.
You might not like what he's doing or it might, you know,
benefit one cohort of people or, you know,
be to the detriment of some other cohort.
But, like, it's a legitimate governance decision
because everyone has opted into that.
And I think that in DAO governance,
we've seen a lot of things happen where there's, you know, LARPing as decentralized.
But the reality is that there are these huge centralization vectors in terms of who can actually do things.
Right. You know, it's one thing to have a vote and say, OK, we're all going to vote to do this.
But then if there's some centralized actor who can bypass that,
or even if there's some voting power
that can bypass the will of the majority or whatever,
it's like kind of tyranny
of the majority of voting power almost, right?
Versus the minority of voting power
held by the majority of people.
You get these weird kind of edge cases and issues in terms of how governance
plays out. For me, I think like at the end of the day, what we should be optimizing for
is agreeing on what the legitimate decisions are, making sure that the people who are making
legitimate decisions have the power to enact them and that there aren't weird, you know, hidden edge cases and, you know, hidden places where people can bypass legitimate governance.
And then we should kind of get out of the way of it.
Right.
And I think part of the issue that we've had for the last four years until a few months
ago is that the optimization algorithm that people have been applying is not necessarily optimizing for the
best governance outcomes. It's optimizing for the best governance outcomes and this other constraint
of regulatory arbitrage. And that just shifts the design space into occasionally places that
are not great. And so ultimately what we should be trying to do
is rationalize our governance such that
we get legitimate decisions and a clear mechanism
for making those decisions and a lot of transparency
about how decisions are made and how they're enacted
and ensure that all of those things are lined up
very closely and very clearly.
And I hope, you know, I know certainly in Infinex, we're working on that.
We're doing a big governance overhaul to kind of, you know,
I guess, rationalize some of our governance and make it more clear and transparent
and straightforward what's happening.
And I think that will happen in a lot of DAOs over the next like two, three years,
because, you know, there's a lot of inefficiency that exists and a lot of weird edge cases and bad outcomes that
can come from the way that governance has been designed for the last three or four years. Maybe also going back to the question of has it gotten worse.
I definitely feel like in terms of vote participation, it's sometimes also struggling to get people to vote and to some extent I hope it's also part of the reason that
it's potentially silent agreement when people don't vote and I hope if they really saw something
that they didn't agree with that they would become more vocal about it but yeah I mean for
Cauda for example we have been trying to put more structures in place, like vote delegation,
that gets just more people to become active, to ensure that the variety of stakeholders that in the end decides whether vote passes or not is larger. And to be frank, I think I'm also
a bit disappointed generally with VCs. Maybe we just got unlucky, but my impression is that VCs,
most of them are also very inactive when it comes to actually leveraging
their tokens to participate in voting.
But yeah, in the end, what we are trying to achieve is like an ecosystem
where like really anybody who has like some sort of stake in the system itself
via integrations or building on top whatever it is also has a significant share of the token to
hopefully down the line become more vocal when it comes to like protocol design changes but
yeah I think in general there seems to be some fatigue in participating in governance
and I think that's not an easy one to change
and it takes a lot of effort
to set the right incentives for people to partake.
I think sort of wrapping into incentives and governance
because normally governance I feel is the utility given to a token,
as Cain was saying before,
as a bit of regulatory arbitrage
because people are scared to assign value otherwise.
For both Cain and Anna,
how do you think tokens are going to be valuable in the future?
Is it all going to be about speculation, governance,
or can we expect more
sort of revenue generating tokens or things of the sort
now that the regulatory environment isn't so spooky?
Yeah, I mean, right now definitely feels
very speculative still.
Like if you think about the safe hack
that happened a few weeks ago where
the largest crypto hack in history
1.5 million
billion dollars and it essentially
had no impact on the safe
token price
where you definitely feel that the news
are being traded and it's not so much
looking at the actual
achieve
what is going on,
what is on the roadmap of the product.
Like often token prices today are related to like whatever news
comes to the market, more so of course in over market sentiment,
over market shift, and much less so to the specific roadmap of a product
unless you for some reason manage to market one feature launch
significantly better than others.
But there's no direct correlation.
So yeah, today definitely a lot of speculation.
In the future, I do hope that token economics will play a clearer role.
But I think it's also up to projects to communicate better around it.
For example, I think for Kowdao, Kowdao is already profitable today and is already using some of the profits to buy Kowtokens back, for example.
But I think this was like very little talked about.
Very few people know about it.
And unless there's more awareness around these topics,
it also seems like, yeah, token economics are not necessarily priced in
if you compare different projects.
But I do hope down the line that there's going to be a stronger correlation
hope down the line that there's going to be a stronger correlation between project success,
the project roadmap and the token economics.
And I do think that it's very healthy for the ecosystem that most DAOs are currently
not looking into distribution of rewards yet, but are rather working on increasing the DAO's treasury and reinvesting the profits
into building and growing the ecosystem launching more products making the overall ecosystem better
before we think about actually revenue spread to other to to token holders I think it's
there's we're so still at the beginning of making d5 really
accessible to people that there's so many more improvements and um yeah every investment needed
from from the profits that are currently being made that i think it's too early to think of
profit distribution to token holders
Yeah, I think that's fair.
I've seen some interesting takes over the last few months of, you know, you wouldn't
expect an early stage startup to be redistributing profits, right?
You'd expect them to be reinvesting them into growth, et cetera.
I think maybe where this got a bit distorted
is from the ICO era.
NOSUS still has a couple hundred thousand ETH
or something like that.
You don't need the profits,
if they even exist, to fund growth.
You've got an infinite treasury to fund growth.
So you may as well
reinvest whatever, you know, minimal profits there are into something else. But, but yeah,
I think, I think like broadly, and you were talking about making things
accessible and how we need to invest in growth, what areas, Anna, do you see need the most investment
in terms of improving UX?
We just launched Swidge, which is like swapping and bridging
without gas, no signing.
Is there any other big leaps that you see that need to be built out?
Yeah, I think essentially what I mentioned at the beginning,
I think we're making significant progress. If you compare the UX today compared to 2017, I've seen huge improvements, so many more tools available, a lot more protection schemes for users.
attract like the average non-crypto user today like if you like just observe onboarding a new
team member who previously didn't work in crypto um like the many steps that are involved um
from and then also i guess even like understand like understanding what are the different
um solutions out there like do like a new user on branding like first you need to advise
them on what type of wallet they should use what are the different advantages and disadvantages
like different security levels depending on the wallet um and then there's um basically still
like so many um hurdles for them to understand.
Like if you just, you sign a transaction, for example,
it's like most people don't understand
what they're signing, right?
But at that point, I think we should,
you could also abstract away
what the user should actually see.
You should build a product that is secure to the user.
But if you want the mass adoption, you also need to find a way that is secure to the user but if you want the
mass adoption you also need to find a way that the user doesn't have to deal with these daily
complexities anymore um like a very easy example is the one that kane gave earlier of like um
slippage tolerance and that in the early days in uniswap you couldn't adjust it
and back then obviously there was only more sophisticated players in the game so for them
it makes um it would have made sense to be able to adapt the slippage um but in the end i think
we want to go probably full circle and go back to a reality where users still in advanced settings
are able to adjust the slippage but like the default should be that the slippage is set in a way for the user that they are protected, but they don't have to really know
what does the slippage mean?
What's the impact of the slippage?
And like become like more sophisticated and trying to understand like how they should
set the different parameters.
So we want to get into, I think we should always obviously have these advanced modes available for all types
of products, but there should always be also this easy version for newbies who want to be safe,
who want to be protected, but at the same time have a very limited understanding and don't need
to read through a bunch of FAQs before they're actually able to use a product.
And so part of that then would also be this entire chain abstraction.
It's like right now it's a nightmare.
It's like I open my wallet and I see, okay, there's 10 different networks that I can choose from.
And then you have the bridging across the different networks that can take 20
the different networks that can take 20 minutes or longer.
minutes or longer.
And like in the end, the experience that we need to get to for this to be more mass adoption
ready is like really the simple experience of I have my bank account or whatever it's
going to be called and I have my assets in it.
And again, I don't care what underlying chain I'm on.
I just want everything to look very simple, smooth,
any complexities obstructed away from me
and have this very simple version
that at the same time is putting the user first
and is making sure the user is protected.
Ken, I can see your mic's open.
Did you have stuff to add there?
No, no, I don't have anything to add.
Sorry.
Is there any examples?
Anna, have you used the switch interface?
Have you switched through CalSwap on Infinex yet?
I actually wanted to.
We are also launching our own swap in Bridge now in Q2,
but I haven't gotten to try our switch yet.
You'll have to just try it out.
Friendship over.
That's it.
Cancel the space.
It was not
mentioned as a requirement to join.
We've
actually got someone requesting to speak,
so I'm just going to let Danny
Brownwolf up if he's got
a question. And if anyone else
has questions, feel free to
jump up.
Talking on the sort of governance thing, a really interesting thing that I saw a couple of weeks ago was Crono, the crypto.com token. They had burnt like 70% of their supply and they had
a vote to unburn it recently. And pretty much everyone voted no, except the last minute crypto.com and all the
validators came in and voted yes because they were going to control the supply so it was a pretty
insane example of the entire community obviously wanted one thing except the validators that hold
held a large majority of votes just decided to do the other and managed to massively dilute all of
the all of the normal holders see i, that's a good example of like,
that is a legitimate decision.
It's just a retarded decision, right?
Like you bought into a token where 99% of the governance power
is held by like two dudes.
And that means that you're accepting them
as your benevolent dictators, right?
And like they can do whatever
they want to you so you may as well not vote so having votes is is kind of pointless in that right
if there's ever any uh dispute between the other one percent of people the you know the other
million people that hold tokens and the two people that hold all of the voting power, they're going to win. So, you know, don't hold that token,
I guess, is the lesson that everyone learned the hard way.
Danny, did you have a question for Kane or Anna?
Yeah, just because of the topic and for Anna speaking about the need for abstraction in
order to onboard the masses and the normies. I was wondering if you guys have been using Bankerbot because that is kind of like the
mother of all abstractions and I haven't seen anything like it in crypto yet.
I haven't.
Have you, Cain?
Yeah, I've used a bunch of these bots. I guess the thesis is that the biggest abstraction is just having some kind of LLM wrapper on top of everything. Is that sort of the question?
That's not, that's one of the problems that it solves.
I mean, Anna, for you, BankerBot is, it's an agent, but it really has an entire DeFi
backend that it created a wallet for every user on X and it started in Farcaster.
So every user on Farcaster.
So I can tweet money at you. I can tweet tokens
at you. So I can now, I can do that in the, I can do that in the space, but I can just say,
hey, BankerBot, send whatever, $10 worth of ETH to Anna.
Anna needs to know your wallet address because your X account is connected to your wallet address that it created for you.
So in terms of like normie onboarding, when you're talking about we don't want to deal
with any of the crypto stuff, that is like the holy grail of products that i've seen one because of the simplicity and
two because it does use the largest onboarding channel which is social social media and your
account that's already in there correct me if i'm wrong like at the architectural level it's similar to a lot of these other
uh telegram stop bots where like the funds are custodied uh by by the system like you don't have
self-custody of those assets right you don't but you you can obviously like send it to a different wallet. So it's basically like, you know, like you're either your first step into crypto and just showing normies like the concept of even, you know, being able to tweet money and the other advantages of more like the micropayments and the rewards mechanisms, etc.
payments and the rewards mechanisms, et cetera. And then from that, you know, you can have a
self-custody wallet and then transfer the funds to it. And it has like limit orders. So you can say,
you know, you can give it like, obviously it's an agent, so you can give it different orders. Like
if my balance is more than X, like send it to this wallet, or if this token rises 50%, sell it for ETH
and then transfer it to the wallet.
So the communication is done, I mean, either in natural language as opposed to, you know,
setting up transactions.
But yes, it's initially non-custodial.
But I think, I don't know the ton architecture, but I'm
pretty sure in Telegram, Telegram itself created a wallet for you, right?
I mean, it depends, but like all of, you know, but on a ball, all of those things are custodial,
right?
Like you're handing over custody of your assets and handing over custody of your assets gives
And handing over custody of your assets gives you better UX, right?
you better UX, right?
That's Binance.
So I think it's like, there's a lot of cool things you can do
if you give up custody of your assets, right?
There's a lot of cool things.
And you can say, okay, well, it's constrained by the size of the wallet.
Hopefully, it's only five or 10 bucks or something like that, right?
So it's not risky and it's just a proof of concept or shows what you can do with the wallet. Hopefully it's only five or 10 bucks or something like that. Right. So it's, it's not risky and, you know, it's just a proof of concept or it shows what you can do with the
tech. Um, I do think that the trade-off space is not ideal. Like that's my personal take, um,
you know, and it'd be better, like all of that stuff is possible to do, but also make it non
custodial. Um, so I, I'm always a little skeptical when I see an
implementation like that, that just decides to like cut the
corner of making it fully custodial. It's certainly easier,
and it's cool, and you can do some cool stuff. But I think
you lose something of the kind of value of crypto once you do
that. But like, you know, it's a market, anyone can do anything.
I guess it's a broader question for Kane and Anna. Do you expect agents to be a big part
of the sort of future of making DeFi easier?
Overall, yes, but I would also say, well, if you talk about agents,
can you specify what you mean by agents?
Because for me, I automatically think about intent agents.
Yeah, sorry.
I'm thinking agents in my less technical sense of AI assistants that maybe go out and look for transactions for you
or you describe tasks to in natural language.
I want yield or I want to trade rather than the sort of clicking UX that we're used to.
On that, I'm more skeptical.
I mean, we're definitely still at the very beginning.
We did, as part of a hackathon, experiment around with it a bit,
just trying to train AI. Obviously, there's still a lot of room for what can go wrong.
I don't know.
Like, it's, I think, I don't think this is, like, the immediate 2025 AI will roll over crypto and it's going to make it so much easier.
I don't think that's the next onboarding step for users.
Down the line, I don't know.
Of course, it sounds neat, right?
It sounds really neat that if you can just say,
hey, AI, I want to convert.
I want to create a collateral position with XYZ.
Set this up for me.
But in practice, I'm a bit skeptical
on how this can in a secure way be implemented.
And I definitely don't think we are there yet.
Yeah, I mean, the way that I sort of reason about this is like, at some point, you have AGI.
And then everything you do is, you know, some super intelligent computer, or maybe everything
it does to you, but something like that, right.
And then like some period of time before that,
you've got a slightly dumber version of that
that's doing all your crypto stuff.
And I just don't think that the gap in time
between those two things is gonna be that large
given how crypto works.
So maybe we get six months of all our crypto happens
through agents and then AGI turns us into paperclips,
but I guess it'll be a convenient six months.
Anna, you mentioned that you've done hackathons in training AI to do transactions and things
like that.
What were the, I guess, things that you bumped into that were issues that make you, you sound pretty bearish on getting AI to do DeFi for you?
What issues came up?
I mean, this was really just like an internal hackathon we ran.
And like one, I think our UX researcher together with a couple of others started training this
model.
And so, I mean, it was a hackathon. So it was just really a couple of days put into it and training this model and and so i mean it was a hackathon so it was just really a
couple days put into it and training the model and so would still get a lot of things wrong
um so i i mean things that can certainly be overcome if you like invest more time into it
um but yeah we didn't proceed this further i didn't feel like this is the next big thing
that's gonna that's gonna boost us forward.
It's like really stupid things, right?
Like, oh, I want to convert my Veth into Dye
or whatever it is.
And then instead of converting it to Dye,
like it converted into S-Dye or something.
It's like really very at the infancy of this.
Yeah, it seems like that.
Right. Like the, you know, um,
crypto is such a large solution space to like have to reason about but unless you create like very tight guardrails
which you know obviously constrains the utility right if you just like hey go out and do it like
interact with any contract and do anything like you're going to get giga wrecked at the moment
in the current state of uh llm so you got to put some guardrails around it at which point
now you've got a different you know set of trade-offs so I just think like by the time we get
to a point where it's like yeah you can just trust the machine to do all your stuff better than you
like 20 minutes later you're a paperclip so enjoy it
I think that covers any of most of the stuff we wanted to talk through
uh Anna did you have any final thoughts you wanted to give to wrap things up? No, nothing specific.
Cain, any final thoughts? No, I'm good. I'm good. I'll leave
you all with that AI doomerism.
Yeah, it's all over. We're all going to be paperclips soon.
That's a great final message.
Thanks so much, guys, and thanks for everyone.
Enjoy your future life as a paperclip.
Hopefully we're sentient paperclips, though.
That would be cool.
Doesn't sound horrifying at all.
There's a...
How much percent chance?
One percent.
But there's a chance. You're saying there's a chance.
Yeah.
Alright, we'll wrap it up here. See you guys.
Thanks so much, guys.
Good night. Oh, good day.
Yeah, see ya.
I'm going to bed. Bye.
