Hello, everyone. Thank you so much for joining our Twitter space this today. We will be chatting
with aftermath blue fin and Korea about perps and defy in general on sweet. We will get started
in just a minute. We'll give people just a little bit longer to join the space before
Hi, I'm Wayne Cunningham, and welcome to our space where we'll discuss perpetual futures
in defy from simple commodities trading the world of traditional finance developed a dizzying
array of financial instruments and contracts. Most notably, futures trading let someone buy
an option to purchase a good at a set future date. Typically, they're hoping that the good
will command a higher price at that future date. Crypto trading borrows a lot from the
traditional finance industry. However, one key difference is the rise of perpetual futures,
which have no set end date. Joining me to talk about perpetuals are representatives from sweet
defy protocols bluefin, cryodex, and aftermath. We'll start with some introductions revealed
from fluffin you want to give a give a brief intro about your platform and what you and
your role there. Yeah, thanks, Wayne. Yeah, I'm reveal. I'm one of the co founders, and we're
building bluefin, which focuses on derivatives, particularly perpetual swaps. My role is evolving
day to day, but my background's in physics. So I help with the formalism and design of
our marketing engine. And then I work with our investors, our market makers, our partners,
as well as support different functions within our team the best I can.
Thanks, reveal. And let's go to you, Aditya, from cryodex. Can you talk, give you a brief
intro about yourself and your platform? GM guys, I'm Aditya. I'm also one of the founders of
cryodex. Kriya, as some of you might know, is one of the most liquid spot DEXs on sweet right
now with close to 30 million in TVL. Our next stage of growth is more focused towards enabling
leverage derivatives trading, starting off with burps as sort of the first instrument.
I'm more of the non-technical co founder. So on a day to day basis, I mostly spend my time
interfacing with quants, tech leads, and designers in the team, trying to bring the whole product
together. Currently also working on building a strong marketing and BD team to prepare for
scale in the coming years. So yeah, that's about me. Great. Thank you, Aditya. And let's go to
aftermath with air TX. Can you talk a little about your role at aftermath and describe the
platform? Yeah, how's it going, everyone? Yeah, air TX here, I do a little bit of BD and help with
the on chain perpetual product, we have probably seven, eight products here at aftermath. Some are
live, some are currently getting on to testnet, you know, then building on silly before before
even went on mainnet specifically to build a fully on chain matching engines. Great, thanks,
air TX. And let's stick with you for a minute and talk about how your platform processes
perpetual futures. And what's the experience like for users? Yeah, so we're on set. So right now,
we already had, like, perpetual is what we call perpetual V1 internally, which is a fully on
chain matching engine, kind of a big breakthrough for us in terms of having fully on chain
matching engines, just in crypto in general, they're very compute intensive, and really kind
of hard to design to have like an efficient system, been really lucky with how SWE's consensus
enabled us to kind of have a matching engine design that's comparable to off chain order books,
and being able to see the book fully on chain have all the, you know, all the matches be done
on chain and essentially help build more robust markets. I think that's as a trader myself,
pretty, pretty fascinating and, you know, very complex in the sense that we have to
focus on not only security of the matching engine itself, but the effective way of making it
economical. And I think that building a robust, you know, fully on chain perpetuals system is very,
very complex and tough. And as you know, we've seen only some chains are able to sustain it
something like Solana, but the effect that we get from storage design, the gas mechanism on SWE
is really fascinating and really excited to bring it into mainnet in the next coming months.
Can you speak a little bit to the advantages of doing of having your order book on chain?
Yeah, so there's a few things with off chain stuff that, you know, we believe that if we're
gonna, if we believe that DeFi is this kind of race to equality, we believe that matching engines,
specifically things that make order that settle orders and kind of match orders should be fully
on chain and transparent. That means that no one, no participant in this system should have a first
look or be able to be first in the queue. And, you know, some would say is, you know, well,
what about Mev? And I think that the argument there is we're actually, you know, fighting against
Mev by creating more opportunities for people to see the book and kind of build more robust markets
since there's more people seeing everything that's going on and increases censorship resistance,
it increases people's ability to not get front run. And most importantly, it allows us to
be more hands off in terms of let users and the community decide which markets they want to
kind of build. And that is very fascinating. I think that's why we're all here. And very few
chains are able to give us this type of performance. And it really comes down to like
the consensus model and just how sweep does gas. And I think that's just so different.
And I think that's just very underrated. I appreciate that. Thank you. And let's go to
you, Aditya, about the Cryodex. And can you describe how your platform processes perps
and the experience for users? So I mean, from a very zoomed out lens, if you look at it,
perpetuals are primarily a tool for gaining leverage price exposure to an asset. And how we
see it as is like there are four primary components to this flow, trade intent,
intent matching, position, margining, and P&L settlement. And how we've built Kriya is
most of the components here are decentralized and completely on chain. For instance, the orders are
settled on Sui's on chain order book D book, the order book is countless and completely on chain,
completely agreed to ethics point there. Like this is just a different form of MEV,
where everybody is able to see the order book and there's sort of no front running or people
partially matching the orders, then transferring into the final order book. We also have a custom
mark pricing record that makes things more transparent. However, like basic keeping the
risks of financial leverage in mind, we release new functions and ease trading rules incrementally.
For instance, we even would have 20 X isolated margin USDC settle contracts,
we too will have cross margin cross collateral support, we three will have a bunch of more
detail features, we'll introduce maintenance margin limits increase them, but that we'll do
only once we bootstrap like a solid liquidity base, a big enough insurance fund to support more pairs
and volatile market days. Essentially, we'll have to see how the system handles stuff like
liquidation spikes, funding rate accuracy and low volume pairs or to deal leveraging etc.
But as far as the influence Sui is concerned, I think this is one of the few chains where an
on-chain order book is actually practical in building. I mean, we tried a bunch of EVM chains
before we came here, we tried even looping and different L2s whatnot, because we wanted to stay
closer to the EVM liquidity. But I mean, here, it's possible to actually build a decentralized
on-chain order book because of basically Sui's consensus mechanism, and the high TPS it can
support and also like why we particularly built on Sui was the security of move like given the
recent keywords of hack, right? We saw like, even in the best of the blue chip EVM protocols,
there is a possibility of re-entrancy and with leverage instrument security ski. So yeah,
that's pretty much like how our flow works. We've kept most of the components decentralized,
and we slowly see how it compares to like places where some things are off chain,
some things are on-chain. Great. And of the user features you mentioned,
because user features are really going to be your differentiating factor,
which would you say is the most impactful or the strongest one that you offer?
I mean, first of all, pulling everything on-chain with an order book is also key differentiator.
Like if you see on the EVM side, only GMX or GAINS networks are a few DEXs that are completely
decentralized, and that too is also arguable that like they have keepers and all, plus the
price discovery happens just via the oracle. There's no native price discovery. And all the
DEXs that have native price discovery, like DY, DX, et cetera, till now had an off chain order book.
So this itself, we believe is a key differentiation. And later, I mean,
we can segregate the user experience into two parts, essentially, capital efficiency and UX.
So on the capital efficiency side, we are working on increasing the maintenance margin
limits, increasing cross collateral support, like initially, we'll have just USD margin contracts,
but we'll try to bring in BTC margin, ETH margin, SUI margin contracts pretty soon. So that is there.
Plus, we have sort of a bunch of like vertically integrated features that will help like bring in
more take the liquidity. For instance, like I previously started up before this also, I was
building a retail broking app in India. So my experience is pretty much around how to make
trading easy for retail. So we'll release a bunch of features around copy trading,
market making walls for retail. So essentially, market making on an order book for a retailer
would be as simple as just uni v3 LP, something like that. So yeah, basically, some features around
capital efficiency, some features around UX, the UX features will mostly enhance take the liquidity
and the capital efficiency will help the makers come on board. So yeah, that's the case.
Thank you. And let's move on to bluefin reveal. Can you talk a little bit about how
bluefin handles perps and what's the experience like for users?
Yeah, for sure. We've been live for a few months now with an early version,
and I've already processed a high level of scale. So we crossed, for example, 400 million in volume
yesterday alone. We use an order book and an on chain margining engine. So that is responsible
for validating signatures, collateral risk before executing the orders and transfers of balances.
We've built on the premise of remaining non custodial from the start. So users don't have
to take the same security risk with their assets as they do on a centralized exchange.
And now are working towards complete decentralization and permission list
markets by the end of the year, with rolling releases coming along the way.
Great, and let's stick with you for a minute and talk about,
I mean, you mentioned volume. Can you talk about how popular perps have been on your platform?
Yeah, I think it's still early days. And it's still an early version of bluefin with
a few markets. But it's still exciting to see that we've, you know, on average,
crossed $100 million in volume for the last two months, and are at a similar scale to GMX
and a few others in other ecosystems. And so I think it's great early traction,
and we're just excited to release the next version of our protocol and keep building from here.
I know bluefin specializes in derivatives. Are perps a portion of those,
or is everything a perp? No, so perps are the initial phase of our launch,
we will be introducing more financial instruments, including spot and then we're also looking at
options. Great, and I guess can you I mean, the perp process is pretty interesting, because I
guess people can take both sides of a position, you can short it or long it. Can you describe
that a little bit? Yeah, I think perps are really popular for a few reasons. One,
they're super, they're a lot cheaper than spot assets. So for example, many centralized exchanges,
it costs about 10 bibs to trade the spot asset, whereas it costs about four bibs to trade the
perp. It's usually a lot more liquid. So you can execute larger orders better. And then you
can take leverage, leverage exposure on both sides. So it's a lot easier to be able to go short,
which allows you just to hedge and also, you know, take a bet on the other side.
Great, thanks. And let's go to you, RTX, about an aftermath. I think you mentioned that perps
are still in testnet phase on aftermath. Can you describe how that process has been going,
getting everything working? Yeah, I know. You know, we've learned a lot in already perps v1.
And we made some couple breakthroughs in terms of this matching engine design that we've been
working on. I think it's important to note that what separates like, you know, suey from everything
else is that sweet sequences access to the same shared objects, rather than the transactions
themselves. And so building this matching engine has been, you know, able for us to like make
everything, you know, more optimized, especially accessing memory and computation operations. So
that's what we've been doing for the last, you know, two years, but especially the movement between
perps v1 to perps v2 has given us probably like four to five x more throughput design,
sorry, throughput in terms of orders being matched. And that, again, as someone who's loved this
stuff for many years is super exciting, because this is all happening on sweet using sweet
consensus. So just from like the testnet phase, it's been pretty remarkable in terms of just
getting this type of capacity that we just haven't seen anywhere else.
You mentioned you have 3x and 4x throughput. How have those stress tests gone? Have you really
pushed the system with a limit? Can you comment on GPS? Yeah, so it's orders per second, the
it's difficult to comment about, you know, orders per second, because they're, it's not
straightforward. But essentially, what we're talking about is we could scale today, not even when
miss a CD comes out, but today, the level of how many machines you can throw in the network,
and that's how much more throughput we can get per market. So you could, you know, again, very,
very simple, please, like very simple design, you can see upwards of over 2000 orders being
matched per second on chain, you know, in any market. And again, these are, you know, not
limit orders, like those have, you know, different types of variables, like, you know, we're seeing
upwards of that, of that number. And it's like, pretty insane to think about this is all happening
on suey. So yeah, I mean, things like that is, you know, me personally, has been pretty,
pretty, you know, pretty impressive to see and just shocked that like, this is actually possible,
because this two years ago was just a dream, to be completely frank with you.
Yeah, that's great. That's, you know, we, I was there when missed and was developing the
that not early days, but you know, relatively early days, you know, and there's a lot of
promise about the platform. And it's just amazing to see it work out in production. That's obviously
the real test. Aditya, I want to go to you on in Cryodex about, can you talk about how
popular perfs have been on your platform? So Kriya's perpetuals are also not live on mainnet yet.
However, we did an incentivized testnet for two weeks in December, that's so close to
50,000 transactions being performed by about 1200 users. And like given our seven day active users
on the spot, AMM are around 3000, 4000 right now, seeing that almost half of them tried out
Perpetuals is a sign of like good demand from even the retail side of things. And yeah, as
I just mentioned, like, TPS is tough to quantify, because there are many things happening under the
hood, people are placing market orders, they are matching with multiple limit orders, stuff like that.
And mostly this Perpetuals testnet phase was to habituate people more with the UI,
get their feedback on like what's working, what's not in terms of the manual orders.
So we didn't get to experiment with a lot of traffic spike yet. But that we have done some
internal tests, seeing some good numbers, but yeah, don't want to promise on the testnet stats.
So we have a phase two of incentivized testnet two, probably early Feb. So that's when like,
let's see, that's the plan where we'll run a lot of bots, see how much gas surge happens. And let's
say even at a scale of 1 million wallets, are we able to deduct funding rates or not stuff like that?
Yeah, it's impressive. And I want to pivot a little bit out of about cryodex. And obviously,
you're putting a lot of resources behind supporting perps. What? So you've seen popularity in the
industry for them. Can you comment on why you think perps are so popular in web3?
Yeah, so honestly, I think there's no one single reason why perps exist in crypto and not in
trade fight. It's maybe a combination of multiple things. I mean, correct me if I'm wrong. So one
is the history. So Bitmex introduced perps for the first time in practice, and this data we
introduced them in crypto. So perps came for the first time in practice in the world in crypto,
there was some theoretical papers, but of how to do it, do it in trade fight, but no one implemented.
Second, I think is maybe the demand. Like for any futures instrument, there are two components
speculate on the price exposure, or I either hedge for some event risk on some particular day.
So like for commodities, I usually buy a dated future assuming, like in my supply chain,
if something goes wrong, some sort of blockade happens, I'm safe against that risk.
So dated futures as an instrument in trade fight are used mostly to hedge the risk of the underlying
asset, or to take physical delivery and crypto, it's all magic internet money, right? So there's
not much concept of physical delivery, people mainly want the leverage part of things. And
given the market is 24 seven, and auto rolling over futures, that's what a perpetual is basically
every six are an expiring futures is a perpetual right. So there was more need for crypto in
there, there was more PMF, then it got introduced here first. That's why I guess Bob's exist more in
crypto and not in trade fight. Like I've heard of some examples like in Southeast Asian markets,
maybe in Hong Kong or Singapore. They've tried one day expiry futures on some commodities,
but it didn't work for some reason, maybe the clearing process sector and trade fire
to habituated to the T plus two settlement, etc.
Great, thank you. And to reveal, let's go to you. Can you talk a little bit about the popularity
of perps on web three as opposed to traditional finance? Yeah, I think that the, you know,
generally covered most of the points, then the other pieces that you know, perps use funding
rates to converge to the underlying, which usually helps keep them very, very closely tied
to the underlying price. And then, you know, given that they're more liquid, cheaper to trade,
you can go long and short, you can take leverage. I think these are the elements that have taken
them off. And then also, I think you can trade crypto 24 seven. So just having an instrument
that's super liquid converges to the underlying price and is very easy to trade is really why
it's become, you know, crypto's most popular financial instrument. Now, I know with supporting
your position on a perp, you have to keep the funding rate going. Can you, as you said,
can you, I mean, is there a point in the future, like maybe one year, two years where it just
becomes ridiculous to keep your position when the price isn't going to change enough to compensate
for the funding, right? I think with funding rates, it's just really a function of liquidity
and having enough market makers, I guess, you know, quoting on on the order books,
if you know, if it is competitive up there, and they're quoting, it's funding rates are going
to be very, very small. And I think based on market movements, if you know, it's super volatile,
it's going in one direction or the other over time, you know, the goal should be that it's
it's very minimal. And that's what we're working towards. And if you use bluefin today,
we have a very tight top of order book, it's about one tick wide, a lot of the times,
and our funding rates, especially in the core markets are very, very small. And that's kind
of the goal that we have been working towards, you know, over the course of last year.
Great, thank you, Reveal. Eric, you've got a lot of experience investing in crypto. Can you
can you talk about the popularity of perps? Yeah, no, you know, I am, you know, personally
very cynical. I think that, you know, with there have been a ton of hacks, especially in like,
you know, in pool two and people farming, I think more and more we people have switched
over to basically trying to, you know, get leverage on whatever they want, you know,
whatever they want to trade or, you know, whatever it is, I usually just call it trading vapor.
But, you know, people are looking for leverage because it is a similar way of like kind of
degenning on whatever they want, you know, whatever they believe is going to go up. So
instead of farming where you have this risk of smart contract risk, especially on something like
solidity, you know, more and more have switched over to kind of, you know, leveraging up using
perps. And I think that's another reason why we wanted to build what we want to build. We want to
give access to perps that are, you know, not only ones that are listed on centralized exchanges,
but ones that don't are not on, you know, on centralized exchanges. That's where, you know,
the on chain kind of matching engine and these on chain vaults, market making infrastructure
that we built is to get those people on our platform because we think that's where a lot
of the volume, like real organic, real retail volume is coming from is those types of users.
So that's what we have been very, very focused on. Great. Thank you. And let's stick with you
for a second. I want to pivot again to really the technology underline everything. We've talked
a lot about SWE's consensus process saying the fast transactions. Can you talk about the
SWE technology or what specific SWE technologies does Aftermath really rely on? Yeah. So being
on chain requires us to rely on everything. You know, the biggest issue when it comes to,
you know, building these things is obviously gas, you know, gas efficiency. And, you know,
being able to create a economic model of like, you know, having users pay for gas and whether,
you know, what happens and makers pay for, you know, makers always have to keep paying for gas.
The fact that like the storage rebates is absolutely like one of the biggest game changers,
I think, of SWE and like makes me think that it's just super misunderstood is that like when
you delete something from the chain, you get a gas rebate. And like more often than not,
that gas rebate could, especially during times of volatility, and we love trading during volatility,
that's where most of the money is made, could be more than the original transaction itself.
So you kind of can create this flywheel, we can get natural makers without incentives
to trade and being able to like, you know, you know, cancel orders, get paid for that,
and then re add orders. And it creates this like kind of flywheel that we personally just find
really fascinating and just you can't do this anywhere else. That, again, the sequencing,
the sequencing of the objects, I think, is also another game changer doesn't exist anywhere else.
And that allows us to create these markets that you can paralyze against each other.
And that is just so, you know, again, from a design space from like designing the stuff
for two years and seeing everything else doesn't exist. And the other thing I want to mention,
I mentioned before, we're talking about TPS. I think when we talked about perps, especially
order book based, not AMM based, is we want to talk about orders per second, not TPS.
I think orders per second is more relevant here. And achieving something like, you know,
like I said, just let's say 2000 or whatever, orders per second is much more insane to think
about. I know it sounds less, but it's bigger to me of a number than saying, oh, we're doing
150,000 TPS because we're essentially saying we're matching 2000 people together in a second.
And you can scale that. And that's like, you know, just from a derivatives, you know,
seeing everything that's just insane to me personally. And that's what makes me really excited.
Yeah, I agree. I mean, orders per second is really where the rubber meets the road.
That's the user experience and the number of users you're getting through your system
and platform. So that makes a lot of sense as a metric to look at.
Rabeel, I want to go to you to talk about what specific suite technologies really make
Bluefin work. Yeah, for sure. I think the consensus mechanism generally, as we've spoken about,
and soon we're moving to take advantage of suite's own objects. And, you know,
the key thing over here is we have about 400 millisecond finality that will improve user
experience significantly. There's a detailed thread that Chaos Labs wrote about it yesterday,
if anyone wants to delve deeper into what this means. I think it's both on their social and
our social. Apart from this, we're actually also excited about a lot of the account abstraction
work that, you know, suite has been putting in and building out. Primarily, there's ZK login,
and we're currently building a wrapper on top of it. So in two weeks from now, people will be able
to use Bluefin with just a Google account that's, you know, secured on the back end by ZK login.
We're obsessed with user experience. And, you know, while it's still early on suite,
on some of the fronts like bridging, we think a lot of the right effort and right priorities
have been placed on making it as easy, making the network and the protocols on top of it
as easy as possible to use. Yeah, bringing more people into this technology is, you know,
incredibly important because there's so much potential, you know, to talk about total
addressable market. It's in the billions. Let's move on to Aditya. If you want to talk a little bit
about the technologies you've been working, or that your team has been working on to
make credit X work. So initial reasons why we picked suite was because of move security and the
consensus mechanism. But in retrospect, one benefit we've seen of building on suite is the superior
on-chain tooling. Mr. and keeps building Post mainnet, as Rabil mentioned ZK login,
and with the bridging being dependent on wormhole previously, now there's a native suite bridge
coming as well. The gas sponsorship mechanism is at the chain level. So these are some benefits
of suite we're seeing. Then mainly, like our core perpetual protocol is built on top of Deepbook.
So we currently use suite's on-chain public good for matching or price discovery in our protocol.
So that is one major benefit. Other is, I mean, I'm not a very technical founder,
but what I keep hearing from the dev team is smart contract writing has been modularized and made very
easy with dynamic fields and the object model. They can write complex settlement and margining
contracts with relative flexibility. Lots of UI integrations slash batch operations have been
made easy via something called programmable transaction blocks. So essentially, you can
batch operations together and execute them together on-chain. So that makes a lot of
stuff easy for us. So yeah. Yeah, programmable transaction blocks, that's like a key
suite primitive that lets you just batch things and make the whole process or the whole
processing way more efficient. Aditya, I want to stick with you for a second. We're just about
at Timers. I want to give some closing thoughts here. Can you give any final thoughts and
possibly talk about any future features that you're looking to implement in the next year?
So yeah, like in terms of features, what we are thinking is to lower the transaction cost from
the taker side or the gas fee side. We are thinking of introducing Shinami gas station
to sponsor fees or build a market for that in the background. We are thinking of adding
a keeper layer, but most of the transaction cost part is solved for by SUI's consensus
and low gas fee. In terms of reducing the liquidity cost, as in the maker fee or the
spreads, we are thinking of introducing market making walls for retail, thinking of a market
maker program that will closely be associated with our token launch also, but we'll have to
think through it given we already have a spot deck. So how the governance token aligns with
both the spot decks as well as the perpetual decks, that's something we are closely thinking on.
Then in terms of like there are a bunch of vertically integrated features we've thought of,
for instance, in AO, you can see there's yield on idle margin. So if your UOTC is kept in your margin
account, you at least keep earning interest on it till a position is not open. Some things around
copy trading like hyper liquid has really good feature called community walls where people can
build their own hedge fund basically on chain and people can copy their trades or invest in their
fund that is used to trade on the decks, stuff like that. And to bring in more institutions,
what we are thinking is like adding functionality for block trading or off-order book settlement,
custom margin for P2P orders or verified market makers, stuff like that. And yeah,
in terms of final thoughts, nothing much like thanks a lot Wayne for Sui Foundation for putting
the space together. Thanks for the listeners for their time. And for our spot users, stay tuned
for Kriya's season, Chakra campaign season two. And for the ones awaiting perps, the audit is in
progress. We are trying our best to launch in Feb, but yeah, there are a lot of risks associated
with synthetic leverage. So yeah, we are trying our best. Great, thank you. And AirTX, do you
want to give some final thoughts and maybe some future possibilities for aftermath? Yeah, I know,
really excited to finally get this onto mainnet. We've been building for a long time. We also
built the Elixir integration for the whole chain, not just for us. So we're really excited about
bringing those kinds of experiences to Sui, just getting more liquidity in there.
Yeah, I guess to sum up, I think the reason why we are excited the most to being on Sui
is we want to offer the most diverse markets, the most transparent markets and ones that create
the feeling of user experience. Since we have all these user experience tools like ZK login
and making it easy to sign up and have retail be involved, transparent. We don't want,
we don't want to sit in a situation with Robinhood. We don't want institutions or market makers to front
run or being able to see the book somewhere else, you know, some AWS system. We are here
to make sure that we are the most central, you know, censorship resistant and I guess
fairest access to markets we can build. And that's as a trader, as a DJ and someone who's really
involved in DeFi. I'm, you know, really excited to see Sui actually getting that kind of opportunity
to showcase all the amazing stuff that all that makes in technology that we can incorporate to
building all this. Great. Thank you. And let's give a final thoughts and closing to
Rabeel. If you want to talk a little bit about what might be in the future for Bluefin and any
final thoughts about Purps on DeFi? Yeah, for sure. I guess, you know, we have a current,
we currently have a few priorities that we're building around. An easier user experience-wise
secure account abstraction goes live by the end of the month. You know, performance and
merging engine upgrades with Bluefin V3, the token, the DAO, permissionless markets,
and full decentralization of our stack. And then finally, I think, you know, some of the others
were talking about Bolts for market making, Elixir, which is a great team, you know, after
after Matt was also talking about, we'll be launching on Sui and Bluefin on Feb 7th.
So that's right around the corner. And we're super excited, you know, what this will enable.
Great. Thank you, Rabeel. And thank you, everybody, for joining us today for this session. Look forward
to future sessions. We'll be conducting these on a weekly basis, and we'll be announcing
our next one very soon. Thank you very much. Thanks a lot, guys. Bye-bye. Thank you.