Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Curation as an Infrastructure: kpk's Design Philosophy
Recorded: Feb. 25, 2026 Duration: 1:04:42Space Recording
Full Transcription
Music
Music
Music
Music
Music
Music
Music
Music
Music
Music
Music
Music
Music Music Music Music Music Music Music Music Music Music Music Thank you. Oh I Oh I'm going to go to the next video. Thank you. I'm going to go to the next video. Thank you. Music Music Music
Music
Music
Music
Music
Music
Music
Music
Music
Music
Music Music Music Music Music Music Music Music Music Music Thank you. I'm going to go to the next video. Music I'm going to go to the next video. All right.
Thanks for joining everyone. Today we are hosting our Twitter spaces titled
Curation as Infrastructure, KPK's Design Philosophy.
I would just like to start with some intros. Starting with myself, I am head of research at DL Research.
We're a part of Llama Corp, which includes DeFi Llama, DL News, Llama Nodes, and a few other Llama-branded entities.
Would like to pass it over to Seb, one of our analysts, before introing the KPK guys, and then our friends from the Edge podcast.
So, Seb, if you want to say hello.
Yeah GMGM everyone. Yeah I'm Seb research and data analyst at DR Research and I wrote the report on creation and how KPK approached it basically.
KPK approach it, basically.
Awesome.
And then I'm not sure who from the KPK side, if it's from the main account
or I see Deep Crypto here, but feel free to intro.
Yeah, I'm turning from my personal account, but yeah, I'm Deep Crypto
from my personal account, but yeah, I'm Deep Crypto Dive, Head of Risk Curation at KPK.
Dive, head of risk curation at KPK.
So we built and operate curated vaults on various DeFi platforms and the product isn't just yield,
it's predictable behavior under stress. I've been active in crypto for over 15 years now and full
time in DeFi since 2019. Yeah, and I'm mostly interested in what still works when markets or conditions
get messy, like we've seen over the past few weeks and months.
Awesome. Thanks for that. And DeFi dad in Nomadic.
Yeah. Hey guys, thank you for having us. This is a killer panel. Yeah, I love what Deep Crypto Dive shared there. So I've been in crypto almost 10 years. I've been a DeFi punk, which is a new term. I just saw the Ethereum Foundation coin today or publish on one of their posts about what's real defy so i guess i've been a defy punk since 2018 2019 and uh we are proud power users of
defy llama and we are uh also uh extremely excited for this like next curation as an
infra layer that is i think being pioneered by the the team here at KPK. So appreciate you guys having us and subscribe to the Edge podcast.
It's the best place to hear interviews with people like us on the space.
Nomadic, did you want to say anything?
I just talked over to you.
Honestly, you covered most of it.
I do the podcast with this gem of a human, DeFi Dad.
I think I made my Coinbase account in 2016.
So yeah, been here for a long haul,
wondering when I'm going to make some money off this stuff.
Just kidding.
But yeah, good to be here, guys.
Excited to explore this new curation frontier with you all.
Yeah, and last but not least, Patrick, if you want to do an intro.
Yeah, and last but not least, Patrick, if you want to do an intro.
Hey, guys, Patrick with the DeFi Llama team.
I think I know pretty much everyone on this panel already over the years.
Great group and excited to talk Volts today.
Awesome.
So we'll get right into it.
So, again, just want to thank everyone for joining our speakers, our listeners, Seb for putting the report together. It's been great working with the KPK team.
Um, you know, when most people think about infra and DeFi, they're thinking about rollups, chains, oracles, really, you know, the technical stack that kind of powers everything that we use.
Um, but over the last year, um, we've seen something else quietly becoming infrastructure and that's how capital is actually managed and that's curation.
and that's curation. A couple of years ago, curation was mostly who's getting the best yield,
who's rotating from pool to pool the fastest, who's squeezing out these extra basis points.
But the conversation has really shifted since then. The more important question now, I think,
and what we're seeing in the market is what happens when things are stressed. Can you withdraw? How fast can you?
Does someone need to jump in manually?
Is there governance friction?
Basically, what hurdles are in the place for a user to exit a vault
when shit hits the fan?
We've seen curated TVL go from around $2 billion
to almost $10 billion at the peak,
now settling at around six and vaults are becoming the main interface for deploying capital.
So the main question has shifted from, you know, what's the APY, although that's still pretty important to how does this vault behave under pressure?
pressure. So the report we put out with KPK basically argues that curation is becoming
infrastructure, not in a marketing sense, but in a structural sense. So it's less about optimizing
yields, more about discipline, liquidity, and encoded risk boundaries inside of these vaults.
So today we're going to unpack all this from a few different angles. We have the data side with DeFi Llama,
we have real power users with the Edge podcast,
our team who did some of the research and data analysis,
and KPK who've built these vaults.
So before getting into kind of philosophy or architecture,
just wanna zoom out and talk about, you know, the curation market as a whole.
You know, as I said before, the growth has been pretty significant. Fees are up. TVL is up. Seems like more players are entering the space every day.
my first question to patrick from a data perspective and just understand if you think
this is kind of a cyclical thing of you know the new hot trend in crypto or are we actually seeing
something where it's kind of a structural shift in how capital is getting deployed
i'm not sure i fully understand the question um so a structural shift towards money going into vaults yeah money going into more curated vaults
you know the sector as a whole i mean i as i said before the you know entire sector has grown from
two billion to six billion right now you know this is a uc continuing yeah no i I think that's going to continue. I think that vaults are, I mean, if you have been in DeFi for a while, like I think pretty much everyone on this panel, then time in the space, it's not realistic to manage a lot of those complex strategies yourself.
If you actually like just want it, want to treat it like a, like a normal investment portfolio that you check like once a week or once a month or even less frequently.
So, yeah, I think it's naturally going to move towards, towards relying on others to curate their strategies. I think there's
still a ton of infrastructure that needs to be
built, especially as far as risk
scoring, professional performance
tracking. We're
trying to
implement some things related to that
at DeFi Llama soon, hopefully.
yeah, I think that's absolutely the way that things are
going. And I think that you combine that with all the rwas that are now coming on chain and like other sorts other sources
of um more diverse sources of uh yield hopefully in the future uh that can then be looped can be
combined um can be can be built into these into these vault products then i think that like that we're actually very, very early stages still.
I mean, $6 billion, even though it's a lot for a person,
is nothing in the scheme of asset management, right?
Definitely, yeah.
Hey, guys, I might have something else to add there.
And forgive me, I'm really just parroting all of the awesome data that you dug up in this latest research report.
So these were a few data points that just blew my mind.
As someone who was early to using Yearn and Beefy and all the other different vaults systems,
I always treated those as vaults that I could get exposure to the highest yields and hopefully not get blown up and would size my position appropriately thinking about, again, size this position to a point that I could lose 100% of it
and walk away from it upset, but my life would go on, my portfolio could continue to grow.
Now, of course, we're looking at these as potentially the main entry point into the
space that this could be the ultimate vehicle for bringing
trillions of dollars on chain and bringing it on chain from traditional finance. So some of the
numbers that, again, this is all from the report here. So thanks for finding these.
DeFi has an estimated, this is, so today, DeFi has an estimated 43 curators, including 12 who manage more than $100 million in TVL.
The aggregate TVL across curated strategies has expanded from roughly $2 billion to $10 billion
as of November 2025. I think that's stabilized more recently around $6 billion.
Assets under management on Morpho increased from just
under $2.2 billion at the beginning of 2025, so just 12 months ago, to a peak of approximately
$5 billion. So this is at the height of the bull run here in 2025. The number's dropped a bit here
going into 2026. And that number has settled around $3.8 billion.
Over that same time period, the number of active curators on Morpho doubled from seven in early
2025 to 14 as of today. And curation fees rose from $10,000 per week at the end of last year
to more than $200,000 per week in the second half of 2025.
These are mind-blowing numbers, and they're tiny compared to traditional finance.
So it's kind of like when we all celebrated $50 million of TVL running to $100 million or $500
million in 2019, 2020. But when we hit a billion dollars, I remember thinking,
this is just the start. We're going to be at hundreds of billions and eventually trillions
in a matter of years. So as someone who underestimated curation as an infralayer,
again, I think these numbers are indicative of just how freaking enormous this this industry is going to be.
Anyways, I'm sorry to inject there.
No, I love it. Yeah. And I guess going back to like one of the first things you mentioned, how
you know, vaults could be kind of the first interaction, you know, new crypto users have with
you know, D5 moving forward um so i guess from your perspective would
love to know how you kind of evaluate vaults and i guess like what would be your top three
you know filters when looking at these and i guess you could answer or nomadic as well
yeah nomadic you take that or kick it off and i can add in wherever's needed yeah is this so this
kind of top three if i'm looking at a vault what what i'm kind of eyeing up most importantly i would say yeah i mean okay obviously
this has evolved over time like if we're going back to probably when some of us got into the
space d5 summary era it was just big numbers right you were aping yeah um today though
it's definitely like redemption kind of stands out
how quick can I redeem
I think Midas for example
a lot of them are around three days
right now I think even Maple
it's like you should be able to redeem
most of your funds in 24
hours but it can be up to 30 days
in their documentation
so I think these are things I look at, especially if I'm like,
hey, I might need this money in some soon period of time. I don't want to screw myself. So
definitely redemption times. And then just the underlying tech, like, hey, are these new contracts?
Like if there was a new version or a fork of Morpho spun up tomorrow, I wouldn't be stoked to ape into it.
An example of this, like, you know, DeFi Dad and I love Fluid, for example.
I have a lot of my on-chain funds in Fluid now, but I didn't jump into that right away.
It's like I needed like a period of time before I put a material amount of my personal funds into like a new protocol
um even though i know that the founding team intimately we love the guys it's a great product
you know i wasn't a day one jump in so it's yeah the underlying substrate of what this is built on
is there a lindiness to it is it battle tested that's important um and then, yeah, I mean, what are the underlying assets in here? Like,
am I exposed to risk in the strategy itself, right? I want to know that stuff as well. Like,
what are they actually doing in this vault under the hood? Am I going to get wrecked
by the curation itself? So those are like just some kind of top of mind things. But yeah,
itself. So those are like just some kind of top of mind things. But yeah, DeFi dad, you might have
something to add. No, I'll just add the so the main difference today in 2026 versus years ago is
we used to be across the very small set of DeFi protocols that existed. I mean, you used to be
able to just name them off the top of your head. Now, thankfully, the industry has grown to hundreds, if not potentially
thousands of names. And so we used to understand those protocols up and down. And again, there was
lots of unforeseen bugs that could still show up. But you basically understood where you would put
your money and you would be the one who would directly allocate that money into whichever protocol. So we were 100%
in charge of the strategies within our own portfolio. Then we, again, saw vaults grow.
And so we started to examine the reputation and track record of those different vaults-related
protocols. I thought that Yearn clearly was like the pioneer in this space.
But when I think about where VEDA took us with EtherFi-related vaults
and showing that basically someone could totally focus on vaults as a service
and then grow the demand to billions in terms of liquidity.
And then, of course, going into 2025, we kind of saw this next evolution with protocols like Midas.
And that, to me, looked more like traditional finance, where we're entrusting our assets with active managers. Like we are basically using vaults as almost like an unsecured
loan and allowing those couriers to manage it. And so anyways, my point in that is that's a
criticism. I think that that was a step backwards. I have all the respect in the world for what
Midas has done in terms of growing interest in the space. I'm very bullish on them long-term.
interest in the space. I'm very bullish on them long-term. But I think the risk management that
needs to be automated, that is the next step. And that's part of the reason we were excited to join
here with KPK is my understanding is KPK is basically using AI to reinforce strict risk management protocols to ensure that the next time there's a blowup
or there's red flags that there's something wrong, maybe liquidity has shrunk within some
protocol and it's better to withdraw, or there's some indication of a DPEG.
What we used to rely on was some active manager on a multi-sig who would act
really quickly, and then they would write up a postmortem and tell us how they saved all of our
money. And again, it was still interesting and an advancement within the space, but I am so excited
to see Vaults have these sort of automated risk management protocols. So number one thing
that we're leaving out here, and to be honest, again, it's still somewhat of a black box to us,
I think KPK will be able to talk more about this, is how they are using and integrating AI into
reinforcing the strictest, most capital efficient risk management. Like this is getting us to a
point where you should be able to more
confidently then deposit into vaults. Maybe it's not today, but it is coming. It's coming to a
place where you no longer have to assume whatever I put into a vault could ultimately be lost.
We want to get to a place where the losses are so limited that traditional finance is looking at this as an equally appetizing opportunity.
Yeah, I want to jump in real quick here. And I'm sure KPK has a lot of things to add to that. But
one thing I think that is so interesting about vaults is, to me, it's kind of like it is to crypto as social media was to online news and in the sense that it
it drastically increases the space of products that can be created so like if you remember when
the when the internet first formed it was like okay like the new york times is putting some
articles online like that's kind of interesting but it wasn't really until you had the ability for new media companies to come about and launch
podcasts, newsletters, you know, social media businesses, that suddenly it was like, oh, wow,
like you can have 100x the possibilities of the types of media that you can create using this new
technology. And to me, vaults are kind of like that for financial products, where you can already start to see it where you have like DeFi was mentioning like AI incorporated in terms of risk management or rebalancing.
I think you're going to see a lot of more creative looping strategies, a lot of other
types of strategies like that. And basically, the total space of financial products you can create is like several orders of magnitude bigger than
it would have been in traditional finance. And like, I, like I'm, I'm, I'm bullish on bolts to
the point where like, you know, I can imagine a future where they are like comparable to ETFs
in terms of the total assets under management and the total impact on, on a financial system.
management and the total impact on the financial system.
Yeah, awesome stuff.
I guess taking a step back to kind of what DeFi Dad was talking about and want to pose
a question to Deep Crypto is just understanding kind of how you distinguish delegation where,
you know, in years past we had kind of the multi-sig thing and there's post-mortems versus today
where the curation model is becoming set boundaries in code.
So just want to understand how you think about that difference today.
Yeah, good question.
I see delegation more like saying,
I trust this person or team to make a good decision, either now or in the future.
Whereas with curation, it's more about, like, trusting a policy that is enforced consistently and is also well documented.
Because if you, like, delegate your assets to, let's say, a traditional fund manager or, like, a liquid fund, you kind of give them a blank check to do whatever they want with your funds.
You have the same in some of the vaults that are out there.
So whereas in curation, and in curation, I mean the protocols like Morpho, Euler, and
others where you have a clear set of whitelisted functions that a certain party can call.
So like in curation, you have this very limited scope of actions you can take.
So you can rebalance across whitelisted markets or you can like use the allocation to certain assets.
And this leads to like very different risk results. So in delegation, the fund manager could say like, oh, yeah,
I think it's okay.
I will stay exposed during times of stress.
Let's say there's a super high utilization on the market.
Maybe they just think, yeah, it will mean revert in like a couple of hours
or weeks, we're going to stay tight.
Whereas in curation, at least in our approach,
we have systems in place that programmatically exit positions
because we, for example, want to protect withdrawals first
rather than the baseline yield.
So to me, it's more about this, yeah, like the framework around it.
So with delegation, you outsource the judgment to another team.
With curation, you are encoding that judgment, like in clear rules,
clear playbooks on what happens when situation X takes place.
Really great stuff.
Not sure if DeFi Dad or Numatic, you guys
wanted to throw anything in there as well.
I'm good. Deep D dive kind of nailed that.
Yeah, same here. I'll have more to say here in a bit, but yeah, appreciate it.
Awesome. And I guess, you know, a question here for Seb, you know, I was curious why,
you know, it was important throughout the research report to, you know, frame curation as more of a life cycle instead of a selection process.
And I think, you know, Deep Crypto touched on it a little bit, but would love to hear, you know, from your perspective as well.
I think we lost Seb there. I'm not sure if Deep Crypto, you want to talk about
how this is more of a life cycle, like how you have to stay up to date with changes protocols
might be making and talking about what Nomadic mentioned earlier, where if you look into the
docs of syrup, for example, there's a 30-day withdrawal. So yeah, I would love to hear kind of your perspective
on things there.
Yeah, sure.
Because what we do, we're constantly evaluating
both new strategies we might want
to add to the curated vaults as well as the existing ones.
And yeah, you're right.
It's a very fast-moving space.
So protocols will shift into like different underlying assets themselves.
We see now a lot of protocols moving into RWA assets because the yields on chain are compressing.
So suddenly we also have to do like due diligence on certain private credit funds, for example.
So what we do, we constantly monitor the protocols
we are allocating to in the second layer,
and we will remove them, like if some of the assumptions
we may change.
So for instance, like a very obvious one
is like if exit liquidity deteriorates,
so if more photo liquidity gets thinner
or you get like chronically tight borrowing conditions
with high utilization, where the utilization is in a zone
we're not so comfortable with for too long time,
then like our exit agent will automatically delist
that market and pull funds from it.
But a lot of it is more in the, let's say, the soft skills.
So we will try to price the risk.
So we will look at protocol, see,
are there any new dependencies, any governance proposals
that are upcoming that might impact our decision to allocate
to this yes or no?
Like, are they making updates to their oracles?
Are there incentives that are going to expire?
And if this risk becomes unpricable for us, if we cannot, like, have a good
understanding of it, for instance, there's suddenly like a 90% exposure to, I
don't know, like a private credit fund, or there's a protocol that might be
lending out against uh
and we don't feel like we have the expertise to price this then we
reduce our strategy so i think it comes down to um if anything diverges from the initial thesis, why we onboarded the asset, then we will start offboarding it.
We need to keep things predictable and to withdraw funds or make these risk assessments in a predictable way.
We rather remove it than take unnecessary risks.
Awesome.
And I guess another question to build on that is,
would you say it's harder for you guys to include a market
from a vault strategy or remove it from a strategy?
Are those guidelines kind of the same, you know, both ways there?
Yeah, good question.
I think technically it's like on the same level.
Like I think the inclusion is mainly like analytics.
So you can like pretty well like model things,
set a risk gap, start small,
and then gradually onboard it.
Whereas with offboarding, yeah, it's like a harder call to make,
especially if you're already like you spun up the market yourself,
if you're doing like marketing around it, et cetera.
Like then the things get a bit more muddled, you know?
So that's something, yeah, that we do,
we do like the hard work of saying,
okay, this no longer meets the standards
and we're gonna remove it.
I'd say like that is harder to do
than onboard something gradually and grow it.
Awesome.
And I know you brought up some more kind of RWA strategies and kind of how sometimes it can be a black box to verify what these entities are saying is true.
And Patrick, I know on the DeFi Llama side, we just released the RWA dashboard, which gets into a lot of this data um curious if you can talk a little bit about some
of the i guess hurdles um in getting that data to present um and be confident that what we have
there is accurate yeah totally so with rwas the big challenge has always been not necessarily just
not necessarily just like finding the list of RWAs and capturing them.
It's actually been, how do you assess what is classified as a tokenized asset on chain?
And how do you measure that?
Reason being that you see this frequently where, semi-frequently, I should say,
where some company will come out and say like, oh, we've tokenized, you know,
$100 billion of this.
And it's like, okay okay what does that actually mean because uh doesn't mean that you're like recording
the ownership um like you're writing it to a ledger doesn't mean that it's actually a digital
bearer asset which is like what what most people think of crypto what it really is is is a digital
bearer asset like is it actually a digital bearer asset that's what um like tether gold is for
example where you can transfer it you can use it in an application and then also like even if it is issued is it actually
being used so basically what the big thing that we did with our rwa dashboard is we introduced
two new metrics one is active market cap and active market cap is the portion of an RWA's market cap that is actually outside of the issuer's wallets.
So either it's deployed into a smart contract or it's in the wallet of a user.
Because there's some of these, for example, and there's reasons to do this, but it's worth differentiating there somewhere basically like all of the tokenized assets sit in a single
wallet and then actually actual ownership changes are like transferred or tracked in some off-chain
dashboard right or off-chain database and then the second metric we introduced is active tvl and
basically this is the portion of the the assets that are deployed into uh into defy and and that's
really interesting because if you want to track whether an asset, for example,
is being used in vaults,
whether it's being used as collateral,
whether it's being used in liquidity pools,
then ActiveTVL is going to tell you actually like,
okay, are these assets being productive?
Which is of course one of like the core unlocks of DeFi
or are they kind of just like on chain
for the sake of being on chain?
Awesome. I know, uh, nomadic, I saw you emoji reacting a few times. They're not sure if you
had, um, any insight from like a capital allocators perspective on, you know, the whole kind of RWA
scene right now and verifying and not just trusting what's out there? Just very quickly, I've kind of been very confused by how much TVL is out there. And I think,
you know, like rwa.xyz might have a different methodology. But I was happy to hear how Patrick
and DeFi Llama are quantifying it. Because what I care about is tokenized value that's usable on chain.
That's what I care about.
Like, can it be looped?
Can it be borrowed against?
Because I think that's what sort of expands the pie of DeFi.
Not this like mirrored, I don't know, whatever some tokenized funds have been doing.
It's like, I want that shit on chain.
Yeah.
No, totally agree.
And we're actually doing a state of RWA report where, you know, we're going to focus entirely on that.
So like what's actually active, not just what's mirrored on, you know, these different blockchains.
But super cool.
Getting into kind of the next section here um talking a little bit more about
deterministic agents and kind of automation as a whole with boundaries um seb just want to check
if you're okay to speak here it should be good i'm sorry i lost connection. Yep. X-progged me, basically.
Yeah, should be good now.
No worries.
I know in the report we mentioned that there were, you know,
over I think 4,500 automated rebalances since October 2025
in some of the KPK vaults.
So curious what kind of takeaways there are from the cadence
that we've seen there and just, I guess, an overview of their operational philosophy.
Yeah, I think here that number tells us a lot about how the system works in the back end.
It shows that the vault is not waiting for big problems
to appear before acting.
Instead, it makes small adjustments to keep exposure and utilization within the ranges
that were defined initially.
So the way to see it for me is as maintenance rather than emergency intervention.
And this is kind of unique, let's say, in this curation approach that KPK has.
So the small corrections, they happen frequently so that there is no large imbalance in the system.
There is no large imbalance in the system.
So in the end, this approach reduces the stress
because the system is always close to equilibrium
and close to the parameters that were set initially.
It makes the behavior more predictable
and the adjustments are gradual rather than sudden.
So yeah, I think this number shows that really,
that there is adjustments all the time going on to respect the policy and automation here
is key and protect the liquidity that is in the vaults.
Yeah. And I guess from a capital allocator perspective for DeFi Dad and Nomadic, does this constant rebalancing, I guess, instill more trust in this vault?
So like from your perspective, if you know that this vault is constantly rebalancing, constantly checking against their internal metrics, does that make you trust it more?
Do you feel more confident
deploying capital in there? I guess overall, how do you guys perceive it from your perspective?
So I think I've come from a background of assuming that the best strategies were the less
action that was being taken, basically keeping money in sustainable lending, staking,
LP type opportunities, but then doing the necessary rebalancing, especially with something
like an LP. So this is new to me. This is kind of a new era where I'm really excited to see
these strategies prove themselves and to see the vault. When I say strategy, see the vault actually execute on being able to continuously rebalance as defined by the vault strategy and to do that in a very profitable way. So this is kind of, again, that next era of vaults that we're excited
for coming from, again, a rather conservative history of allocating capital. The more actions
that I would hear vaults take years ago, the more cautious I was thinking, you know, it's kind of like with, it's, it is a
different, it's a different approach, a different risk profile. But if someone said to me, I think
that I make one trade a day, you know, I'm the type of person who's making a few trades in a year,
maybe, you know, and I'm mostly just staying long over the long term i've applied
a similar mindset to my my uh my liquid portfolio where i'm you know trying to earn as much yield as
possible so this is a huge unlock you know for for vaults as as a service if if the track record can
show that that the deterministic agents here can execute uh profitably uh despite
you know taking again more actions than i would take with a liquid strategy like that that really
opens up new opportunity nomadic am i missing anything there is that does that accurately
reflect how how we've talked yeah yeah totally and and i just wanted to share a quick i promise
i'll kind of answer the question as well just a a quick detour here. I had a bit of
an epiphany reading the report and just another shout out to the DL research team, amazingly
comprehensive report. But what it sort of unearthed for me while reading through it was we've now
gotten metrics, we've got KPIs for the curation space. And I didn't even know those
existed. So you see this crop up in every industry. Take for example, perp dexes. What are they
competing on? They're competing on trade execution, liquidity, spread, all that stuff that even I
barely understand. But I'm sort of an analytical nerd sometimes with sports,
with betting in a past life.
But also with seeing this kind of emerge
with the curation space is interesting
because I didn't know there was a way to kind of quantify
who's doing better than others.
And I think that was like a big takeaway
from this whole report is we now have
metrics to quantify performance. And it's going to be interesting to watch those over time and see
like who's actually doing the best with certain measurable things. So to get back to your question,
the rebalancing thing really came on my radar with, there's actually a guy in the audience,
Archer from the Tokamak team, I guess now auto
finance with what they started doing with auto pools and how often they were rebalancing. And I
thought that was really cool because they're showing all the rebalances, like fully on chain,
you could see when they did it. And sort of the methodology, I think it all comes down to like,
hey, as long as it's being done strategically and there's no diminishing
returns, like you're not rebalancing just to rebalance. But I think the report even had an
example of somebody had like a million dollar deposit and the pool rebalanced itself within
like, I think it was like between 30 or 50 seconds based on that new TVL that came in.
So yeah, I mean, that's just like an upgrade right like whether you know now
now you don't have to rely on a human that's maybe asleep to like oh shoot we got a deposit hey let's
do a rebalance oh let's do it tomorrow actually and then you lose efficiency right so yeah i think
it's a pretty cool future yeah that's a that's a really great example there. And I guess taking this convo to a slightly different spin for deep crypto, like is there some scenario where automation could actually become like a risk surface where maybe, you know, the vaults are rebalancing too much or maybe it it makes an incorrect decision around how to rebalance.
I guess, are there any guardrails in place to actually monitor this with some human input?
Yeah, good question.
Yeah, indeed.
It's only as good as the, like the system is only going to be as good as the inputs you provide to it.
So yeah, we do have a lot of guardrails in place.
I'd say one of the main things is we are watching a lot
of on-chain events.
So in Morpho, for example, we will categorize each activity
that happens in the underlying lending markets.
So if utilization changes and that yield changes,
the agents will be aware of this.
Of course, it's quite important to you,
the data you're using is solid.
So we try to avoid any external APIs,
like external databases or data sources for it.
So either we can query it directly on chain, or we typically, we just ignore it.
But yeah, there's a lot of other guardrails in place as well.
So for instance, with the automation, it's really bounded, like what can take place.
So it can only act inside strict limits.
This includes like within the existing allocation caps
or maximum changes in each rebalance.
Within the approved marketing universe,
you have a very limited scope of actions you can take.
So it's not that we're suddenly going
to allocate a lot to a market that is not whitelisted
because the transaction just revert.
Or we're also not going to execute a transaction
that would get a worse yield or allocate into a market
that we're already too exposed to from a risk perspective.
So you do have a lot of bounds that we operate in,
and the agents will just not execute the transactions if they try to, or
like by accident, go outside these bounds.
We do have a lot of on-chain, let's say permissions that are encoded as well, using like the Zodiac
roles modifier on top of the diagnosis saves.
So even if something would go horribly wrong in our internal tech stack,
it's on-chain encoded that certain actions are just not possible.
So for instance, we limited our rebalance agent to only, let's say,
reduce exposure to certain markets for what we call exit agent.
It cannot be balanced and add exposure to specific markets.
So we do have a lot of these circuit breakers and rate
limits in place as well, especially if there's things
going on like Oracle's not updating, et cetera.
We will have systems in place to shut everything down
rather than continue flying blind, so to speak.
So yeah, there is definitely, there is a risk surface,
but it all matters how you handle these type of edge cases.
Because that's maybe something important to remember as well.
So the agents we use, they're not AI agents.
Like we don't have any hallucinations happening, et cetera. Like it's very predictable behavior.
So it's our deterministic systems where if you put in identical inputs,
you will get the same outputs within the bounds that are configured.
So even though they're running autonomously in a way,
they are like listening to hundreds of different on-chain events,
but can only do a
certain number of actions and there's always some human oversight like we're each day we're going
through like rebalance transactions to see if there's a parameter we need to change or to
optimize for for certain use cases well yeah there's more than meets the eye there
certain use cases well yeah there's more than meets the eye there
yeah i'm sure and uh thanks for that thorough breakdown there and i know uh in the report we
talk about the zodiac um use a bit more and um for anyone listening along um definitely check
out the report it's super thorough and and really breaks down everything we're talking about
to a really granular level.
And within that report, we have a case study on the URC market.
And I would love to pass it over to Seb for just kind of a general overview
before handing some questions over to Deep Crypto and the Edge podcast guys.
Yeah, absolutely.
You want me to review this case today quickly?
Yeah, just a quick overview.
Okay, so basically what happened is that there was an unchained deposit
of roughly $1 million into the KPK USDC prime vault.
This is the one you discussed, yeah.
And this was followed by a rebalancing transaction around 45 seconds later.
And on Arbitrum, the same workflow has been shown in under 30 seconds, reflecting faster confirmations.
This is one illustrative example that there were many
such rebalancing transactions as shown, let's say, on the dashboards. And again, we get back to this,
you have more than 4,000 automated rebalancing transactions that took place since KPK launched this one.
So yeah, clearly what it means is
the automation is really efficient
and going extremely fast to rebalance things
according to parameters.
That all set.
Yeah, really great stuff. it looks like deep crypto i don't know if you can still
speak i see you're listed as a listener but in case you wanted to add anything on on top of that
there yeah i think he uh i think he got rubbed uh Seb, could I ask Seb just a follow-up to what you were sharing there?
Because, yeah, this was a great example in the report of how vaults are becoming that much more reliable.
Like, this is what's going to attract bigger capital allocators on chain. So the stablecoin that is pegged to the euro here,
someone makes a million dollar deposit, you were saying, let's say, and then that liquidity was
going to be allocated according to whatever the vault strategy was. But basically, let's assume
it was being lent out on, let's say it was something like Aave, I don't know. And let's assume it was being lent out on um let's say it was something like ave
i don't know um and let's assume that that that market then you're saying was was utilized to the
tune of 97 so basically nearly all the liquidity was being borrowed there and so the takeaway here
is like this would be viewed as just bad luck that would have happened in years past like
ah shoot we deposited a large amount of stable coins into a vault which we've been doing dd on
we've been super excited to get exposure to it how does this happen we just we just deposited, but it was unforeseen that one of the underlying strategies was using some pool that was basically nearly 100% utilized. Now, if we need to withdraw our liquidity from the vault, we're not going to be able to do that. Like it's, it's one of those things where it's been a norm that we, we need to get away
from. Am I dumbing this all down though? Like, am I, am I making sense? Like this is, this is how I
understood the, the example here in the report. Yeah. Yeah. Let me add some context there. Cause
I think there were two things a bit mixed up. Um, but yeah, indeed, like with your C1, there was,
um, a bit similar to Stream,
like there was a big liquidity crunch in the market.
So indeed, we noticed like, well, I didn't notice, like the agents noticed.
I was asleep.
But we had a very big utilization spike.
So going like to 97% and higher.
And this kind of signals internally to the system that we're in the danger zone.
So whenever the trigger was hit
on those underlying markets on Morpho,
the system started to prioritize withdrawable liquidity.
So the exit logic started
and this gradually shifted away the allocation
from like the most constrained lending venues and
rebuilt a type of idle buffer in the vault.
So withdrawals could take place and didn't depend on perfect market conditions.
It was like mentioned earlier, like we would do this in smaller batches and
over a large number of transactions.
So every time new liquidity became available in these markets, we drew a little bit back until the 97% utilization.
So that meant at that moment, like maybe sacrificing like a tiny bit of yield intentionally.
But the whole point of that whole exit policy is that the yield is optional, but like exits are not optional.
policy is that the yield is optional but like exits are not optional we have some bigger
allocators specifically in that vault that rely on these funds for their treasury operations
and they need to be able to like exit from there at all times so yeah like i think the key takeaway
is there like this type of behavior wasn't decided in the moment like it was decided
months earlier when we set up the vault
and we encoded the boundaries
and how we wanted the vault to respond.
So under stress, you're not caught into like a difficult situation
and you don't end up in internal debates,
but you really just want execution.
So yeah, speed really matters there.
So the goal isn't necessarily milliseconds,
but if you can do this within like one or two blocks,
it's for sure like beating human response times
and also avoiding any cues for depositors
that really don't want to be stuck in a position
that's not liquid.
And I want to call out the status quo for years has been,
even with the vault,
you were dealing with very simple,
hard-coded rules and strategies that, you know,
if you were a more advanced user, a more advanced investor,
you would say, I mean, that's great and all,
but like, why don't I just do it myself? And this is
such a big step for us in the space. Like for me personally, this has been the first year in,
this has been the first year since maybe, you know, 2020, 2021, where I'm starting to look at Bolt's strategies and how, again, how risk is managed
and saying, hey, this is getting to a point where I'm not reliant on deep crypto dive to wake up in
the middle of the night and save my liquidity. That is truly the kind of dependency that we've
had for many years. So it's really, really exciting because it's going to
ensure that more serious capital can come on chain. And I think more importantly, it's going
to prevent the horror stories that have been perpetuated over the years because we've seen,
when you see some innocent retail user, which I mean, let's call it what it is.
I think Nomadic and I, if you asked us, like, what are we? Like, I still, I think every investor
is, you know, retail, unless you're, you know, managing someone else's money. If you're managing
your own money, like I still, I feel like I have a certain mindset there you know it's i'm putting my own capital at risk and um you know
i i don't have the ability to to uh create the kind of automated risk management that the guys
here at kpk would create and you know others in the space other curators that we are really excited
for so so this is this is a this is a huge step forward for for all of us and this is this is a huge step forward for all of us. And this is causing me to go back to the drawing board and think, okay, how much longer do I allocate however much of my liquid portfolio myself here versus how much more exposure do I want through a vault strategy where I'm, again, entrusting these deterministic agents to take care of me while
I'm asleep so that a pool doesn't get wrecked overnight and I wake up and I take a 50% haircut
when I could have lost 2% or 3%. Everyone in 2025 was chasing the highest yield until we got
punched in the face on October 10th. That's what changed everything. Suddenly,
the highest yields weren't attractive anymore. It was about where's my money safe? And the answer
in the aftermath of it from the contagion of 10-10, I think was next to nothing. So no surprise,
this conversation we're having should happen all the time and it should happen during bull runs, but it tends to get listened
to in the bear market. So all we can do is build upon this more intentional design so that the next
major uptrend, we don't experience a stream finance blowout again. We've got automated
risk management through something like a KPK vault. We've got
an accountable proof of solvency type dashboard that's constantly verifying on chain where all
of the reserves are and whether something's pegged or de-pegged or not. So just so much
in terms of fundamental development for us to be excited about. Yeah, I fully agree. And I want to add a few things here
because the space has changed so much since 2021.
You also had vaults back then.
You had your own finance.
But the scope was way more limited.
So you had the iEarn rebalancer, which
would shift your USDC between Compound, Curve, Aave,
and I think maybe one or two additional venues.
But right now, the complexity of each of these underlying protocols has increased exponentially.
You can be exposed to 100 different things underlying to it. If you're allocating to
Infinify, you will get exposure to a bunch of other assets as well.
There's hundreds of examples like that.
And it's not so simple for, let's say,
even a DeFi power user to be aware of all these intricacies
and allocate yourself in this way.
Even doing this full time,
if I didn't have support from people doing risk monitoring, doing the due diligence research
on these underlying protocols,
it would be impossible for me to make
a well-informed decision to allocate.
Because at this point, you cannot keep up
with everything that is happening in the space,
like all the different developments,
changes in technical configurations, et cetera.
You really need this abstraction there on top. different developments, changes in technical configurations, et cetera.
You really need this abstraction there on top.
And I think that's where curation, in general,
can be a very useful tool for the end user.
Yeah, Seb, I saw that.
Yeah, I have a quick question again for KPK because you said something interesting regarding speed
of execution for automation that it's not necessary to have to be fast it can be done
in a few seconds as long as we are faster let's say than the human behavior but how
do you see that evolving uh in time with competition rising
automation maybe becoming mainstream somehow do you think that speed of execution will be
essential at some point yeah i would assume like the speed of execution like how fast
somebody or an entity can react will compress more and more,
like with more sophisticated players entering the space
as well.
I remember back in the day, I was doing manual arbitrage
between centralized exchange and a peer-to-peer marketplace.
And that was things that you see in early stages.
Like right now, DeFi is still in very much an early stage.
I like comparing it to that, like these early arbitrage
opportunities.
Right now, it's all abstracted away
of bots that take care of this in seconds, milliseconds.
And you have MEV searchers on top of this
that compete with each other to extract value from it.
So yeah, I think it's going to be important as time progresses
to keep the response time up.
But for some things as well, it depends which ecosystem you're
If you're in a vault with 100 or 1,000 depositors,
not all of them are going to be monitoring their position to such an extent.
So let's say an Aave or another major lending protocol,
I think there it's probably still going to be OK
to be a bit more slow and just do things in one or two blocks,
which still is pretty fast, I'd say.
So yeah, speed is going to matter more, but it's not going to be the end goal, I'd say. So yeah, it's going to be like speed is going to matter more,
but it's not going to be like the end goal, I think.
Great question, Seb.
And thanks for the thorough answer there, DeepCrypto.
And we are just coming up on time here.
So I want to just leave a moment for, you know, maybe the defy dad nomadic,
anyone from the KPK team, Patrick, if anyone has any closing thoughts. Otherwise, just want to say
a big thanks to everyone for listening in. Thanks to KPK for, you know, working together with us on
this report. And huge thanks to the guys from the edge podcast for joining us
today and bringing some really strong insights with them hey i'll just say one thing um yeah
first of all thanks for putting this together guys and thanks for the awesome report one thing that
stood out from the report uh this is a quote kpk deploys its own treasury alongside depositors
across its vaults i thought that was a really cool thing.
I don't know how many curators do that, but skin in the game is like important, I think as well.
So kudos to you guys.
Yeah, thanks.
Yeah, like I personally have like most of my euro savings in our euro vault, for example.
So yeah, we definitely like to, let's say, dog food our own products.
We have on the Dune dashboard also each depositor teched.
So you can see exactly how much is allocated where.
But I think from the 9 million treasury, more or less,
that we have, I think like 7 or 8 is deployed
into Morpho Vaults.
So yeah, we definitely believe in the product enough
to put our own money in there.
And I think we might be one of the curators
with the most money in their own vaults,
as a matter of fact.
That's a great stat.
Yeah, I was just going to say,
might be something we need to track on DeFi Llama
pretty soon, but great stuff, everyone.
Thanks again.
If you guys want to read the full report,
you can check it out on our website.
And yeah, really appreciate everyone coming out today.
We'll catch you on the next one.
Thanks, gang.
Thanks so much, everyone.
Thanks for having us.
Thanks, everybody.
