Curve：穩定幣換匯，怎麼把流動性擠在一起？

6:27:11開場

6:27:11大家好。Michael，你好。今天我们要稍微談 AMM。Michael，你要不要先介紹一下自己、你在做的工作，以及當時為什麼覺得需要一種新的去中心化交易所？Hi everyone. Hi, Michael. Hi. So today we're gonna speak about AMMs. A little bit. Michael, do you want to start by a bit presenting yourself and maybe your work, and why did you feel the need to create a new type of decentralized exchange?

6:27:36好，可以。我想……它現在不算新類型了，但在當時看起來有點怪：為什麼要用美元換美元？不過我大概是最早一批用自己的交易所的人，因為我在借 DAI，又不想賣 ETH，大家都知道 ETH 會漲。Indeed, yes. So, well, I guess... It's not the new type anymore, but at the time it was, yeah, it looked a little strange. Why do you need to exchange dollars to dollars, right? And, well, I was probably one of the first users of my own exchange, though, because I was borrowing DAI because I didn't want to sell ETH, because ETH would go up.

6:28:09所以你創建的 AMM 有什麼特別之處？為什麼需要一種新的交易所來換穩定幣？對。當時我在借 DAI，實際需要 USDC，因為 USDC 可以提到銀行帳戶。Everyone knew that. So, yeah. So, and what's special about the AMM you created? Why was there a need to create a new type of exchange to exchange stablecoins, or...? Right. Yes. So, well, back then I was borrowing DAI, and I needed essentially USDC, because it's something I could withdraw to the bank account, right?

6:28:39這種兌換可以在鏈上或鏈下做。鏈上可以用 Uniswap v1，鏈下可以在 Coinbase。Coinbase 是普通訂單簿，Uniswap v1 則幾乎一定要經過 ETH，因為它把所有資產都配 ETH，而且 bonding curve 也最简单，所以兩邊效率都不高。And this exchange you could do on-chain or off-chain, right? On-chain you could do on Uniswap v1, right? And off-chain you could do on Coinbase, right? And on Coinbase, it had like normal, normal order books. On Uniswap v1, you could, well, you would necessarily go through ETH, essentially, right?

6:29:11我覺得可以做得更好。同時我也在思考 liquid staking，想讓人能進出 liquid staking 倉位。Because it pairs everything with ETH, and it was also the simplest bonding curve. So, well, it was not very efficient, and it was not very efficient on Coinbase. And I thought, well, okay, it can be done better. And at the same time, I was thinking actually about liquid staking, right? So I thought how to make exchange, like make it possible to go in and out of liquid staking, and . .

6:29:39從 NuCypher 到穩定幣 AMM

6:29:39當時 liquid staking 還不存在，但我在做 NuCypher，他們的代幣本來就要質押。我意識到人們會想在質押狀態下進出。liquid staking was not yet existent, but I was doing NuCypher, where you could essentially—well, their token was supposed to be staked. And, well, I realized that people will want to go in and out while it's staked.

6:29:56於是我設計出這套 AMM 演算法，覺得也許該先用在穩定幣上，因為那是我當下需要的，別人可能也需要；liquid staking 的需求也許將來會來，但穩定幣兌換的需求當時就在。So, well, I came up essentially with this AMM algorithm, and I thought that maybe it makes sense to apply it to stablecoins because this is where I need it, and maybe other people also need it there, and maybe the need for, like, liquid staking will come in the future, but the stablecoin exchange need came right at that moment.

6:30:222020 年初我先以穩定幣上線，後來證明沒錯，liquid staking 的兌換需求也來了，用的是同一套 AMM。我覺得這算是 AMM 世界裡最早的集中流動性（concentrated liquidity）。So, well, I launched it with stablecoins in the beginning of 2020, and indeed I was right. The need for exchange for liquid staking also came by, and it was exactly the same AMM. So it all indeed worked out this way. Well, and I think it was the first concentrated liquidity in the AMM world.

6:30:46Uniswap 的集中流動性要晚很多。能解釋一下「集中」是什麼意思嗎？對交易者有什麼意義？是不是更有效率、能不能用到池子的全部流動性？對，集中指的是……So, yeah, Uniswap with concentrated liquidity came much, much later. Yeah, and could you touch on that a bit? Like, what does it mean to be concentrated? Like, do we have, like, what... does it mean, like, for the traders? Are you in some way more efficient? Is it like, can we use the full liquidity of the pool? How do we differentiate a bit this? Right, right, right.

6:31:12它最適合在很窄的價格區間運作，並把大部分流動性放在那些窄區間裡，有點像訂單簿把掛單擠在一起，而不是攤在整個價格範圍。So concentrated means that it supports—well, it's best working in some narrow ranges of prices, but it keeps most liquidity in those narrow ranges of prices. ranges of prices, but it keeps most liquidity in that narrow range of prices, like, well, imagine an order book keeping the orders, like, squeezed rather than spread all across.

6:31:30現在大夥應該都理解集中流動性為什麼好。同樣價值的資金可以做出更深的流動性。沒錯。So that's, well, that's essentially what it is. But I think it's now universally understood why concentrated liquidity is good. Yeah, so it's going to provide deeper liquidity for the same value stored in the pool. Exactly. Yeah, yeah.

6:31:49穩定幣池的集中流動性

6:31:49穩定幣方面，我一開始相對 Uniswap 那種全區間流動性，密度大約 100 倍；USDC/USDT 現在最優可能是 2,000 倍甚至更高。And for stablecoins, I think I started with 100x density compared to, like, Uniswap, essentially. But then for stablecoins, it appeared that— like for USDC, USDT, you have now optimal to be around 2,000x or maybe even higher.

6:32:06你把流動性集中在兩種穩定幣上，如果其中一種脫錨、價格哪怕只動一點，比如從 100% 到 99%，會怎樣？好問題。這和 Uniswap 現在的做法不同，因為流動性可以延伸到很遠，只是兩翼較薄、中心較厚。Okay, and so you try to concentrate the liquidity on two specific stablecoins, let's say, and what happens if, like, one of them depegs or the price moves even just a bit, like to maybe 99% instead of... Yeah, yeah, that's a good question. And this is actually where the difference with what Uniswap is doing now comes, because you essentially have liquidity going up to infinity.

6:32:33穩定幣脫錨時仍有一定流動性支撐。我們見過好幾次，USDC 在 SVB 倒閉時脫錨，價格跌到大概 90 美分，週末大家真的很恐慌，但事後看也是一次很好的壓力測試。You just have less liquidity in the wings and more liquidity in the center. So if some stablecoin depegs, you still have some liquidity to support it. And we already saw that multiple times. We saw that when USDC depegged, like when SVB collapsed, for example, and price went down to, what, 90 cents.

6:32:59Terra 則是完全崩潰，但池子仍長期提供流動性支撐。你現在還能在 Curve 上找到含 Terra 的池，雖然幾乎沒在用，但還在。I guess people were really scared over the weekend. But yeah, it was a good play, I would say. But, well, there was also Terra, which completely collapsed, but the pool was supporting its liquidity, like, until well, like forever, essentially. I think the pool, you still can find a pool with Terra on Curve and probably several Terra tokens in it, but, like, it's not really doing anything, but it exists.

6:33:39假設有新發行方推出新穩定幣，想在 Curve 上建池，他們該知道什麼、該注意什麼，才能有好的深度與執行品質？Good, good. And now, let's say there is a new issuer coming with a new stablecoin, and they want to maybe have a pool on Curve or anything. What do they need to know? What do they need to be careful about to have a good pool on Curve with, like, a good liquidity, a good depth of execution? Yes, so I think

6:33:59放大係數與手續費

6:33:59他們要選集中度或放大係數（amplification factor），就是一個數字，直覺上可理解成比全區間流動性好多少倍、區間窄多少。they need to choose the concentration or amplification factor, how we call it, right? It's a number, and this number has a good intuitive understanding of, let's say, how much better it is than full-range liquidity, right?

6:34:17如果發行方知道自家穩定幣波動大約正負 1%，也許該設 100；若是正負 0.1%，也許要 1,000，諸如此類。How many times better than full- range liquidity and how many times narrower. So let's say if an issuer knows that their stablecoin is fluctuating by, I don't know, plus minus 1%, then maybe you need to put this number to 100.

6:34:34另一件事是手續費，這真的可以是商業決策。If it's plus minus 0.1%, then maybe you need to put this number to 1,000. So something like that. So that's one thing. Another thing is which fee to put, and this is really, this really can go as a business decision, right?

6:34:51手續費太高可能沒交易、池子賺不到錢也幫不到人；太低，比如零費率，對交易者有用但池子也賺不到。應該有一個讓池子收益最大化的費率。But, well, you just need to know that if you have the fee being too big, right, then maybe you will get no trades, and the pool will earn nothing, and it also will be useless. If the fee is too low, let's say zero, right? If you had zero fee, then the pool would be probably useful for traders, but it would also earn nothing, right?

6:35:15現在這類費率大概 1 bp、2 bp 左右。以前更高，但穩定幣波動變小了，它們更穩了，也許是因為……有 Curve 吧。更穩就要降費率。So there should be some fee at which the pool earns maximum, and well, it's usually these days such a fee is around, I don't know, one bp, two bps, something like that. So it used to be higher, but volatility of stablecoins decreased because they became more stable, because... ...

6:35:39誰控制這些？Curve 還是發行方？發行方可以全選，但池子建好後可用治理投票改參數，隨時能改但要投票。配置出錯誰負責？機構跟中心化交易所有什麼不同？我不覺得參數會「出錯」，最壞是表現不如預期；這是 permissionless 協議，不像借貸那樣有那麼多風險面。I don't know, because Curve exists, I guess. But yeah, when they become more stable, you need to reduce the fee, I guess. And so who controls what? Like, is it you, Curve, choosing, or is it the issuer choosing those fees and this amplification factor? Yeah, issuer can choose all of them, but after the pool is created, you can put up a vote to change these parameters.

6:36:07不像我們也有借貸業務，但提供穩定幣流動性時，能出大問題的情況少很多。So you can change it at any time, but that should go through a vote. Okay, and if anything goes wrong with those configurations, does anyone take responsibility? How does it work for institutions? Maybe compared to centralized exchanges and things like this? Well, I don't think something can go wrong with these parameters. Like, the worst you can have is the pool not performing as you want if you choose parameters wrong, right? But, like, if you...

6:36:39參數風險與外匯方向

6:36:39放大係數設太高，池子可能不小心變成 99% 一種幣、1% 另一種，因為稍微脫錨，但不算危險。這是 permissionless 協議，協議本身也沒有那種責任。choose amplification too much, then maybe you may accidentally get the pool having 99% of one coin and 1% of another coin because, like, it's a little bit off peg, but it's not dangerous, really. So, but yeah, it's a permissionless protocol, so you don't really also have, like, any responsibility of the protocol, right?

6:37:07跟借貸不同，你提供穩定幣流動性，能出岔子的情況不多。我想聊一下外幣、FX 兌換，你們現在也在做，能說明一下嗎？Well, I mean, it's unlike lending, which actually we do have, right? But unlike lending, if you provide stablecoin liquidity, you don't have you don't have many things able to go wrong. Okay, and now I want to touch a bit on foreign currencies, FX exchanges.

6:37:29有意思，從一開始就很清楚長期目標不只是 USD 換 USD。You're now doing a bit more of that. Do you want to explain a bit? Yeah, yeah, that's actually interesting because it was kind of clear from the beginning that the long-term game is to get into, well, maybe somewhere beyond exchanging USD to USD, right?

6:37:52我們試過 EUR，第一個 EUR 池和第一個 crypto 池同年或晚一年，2021 或 2022，跟 STASIS EURS，現在可能不太活躍。And it's, we did try that with EUR. I think the first EUR pool was created the same year as the first crypto pool, well, or maybe a year later. I think it was either 2021 or 2022 with STASIS EURS, which probably is not very active anymore.

6:38:162023 年有跟 Monerium EURe 的池。Monerium 有意思在於可以贖回進銀行帳戶，而且錢是從你名下的帳戶出來。But in 2023, a pool was created with Monerium EURe. And Monerium is interesting because it can be redeemed to your, well, it can be redeemed to a bank account in a way that money comes from an account on your name.

6:38:32大家開始用，過了一段時間我們注意到 EURe 跟美元池之間交易量不錯，有時池子賺很多。規模仍小，但也看到 fintech 推動的其他非美元穩定幣池，比如 crypto 卡之類。So people started using that, and it took actually time, but we started noticing quite good trading volumes for pool exchanges between EURe and dollars, and sometimes this pool is earning a lot. So it's, well, it's still small, and we also started noticing multiple, multiple other pools being created for exchanging non-USD stablecoins and just driven by fintechs, essentially like, you know, crypto cards and stuff.

6:39:12非美元穩定幣兌換

6:39:12才剛開始。2023、2024 用得不多，2026 年非美元穩定幣兌換這塊開始有動靜，我覺得終於值得認真看，我們也有合適的技術。It's just starting, I would say. So it was not used in 2023 or in 2024. 2025, but now in 2026, something started happening in this area of exchanging non-USD stablecoins, and I think this finally deserves attention, and we have actually good tech for that.

6:39:37技術是自動管理的集中流動性：USD 圍繞 1 集中很清楚，非 USD 價格會慢慢變，集中區間也要慢慢移；怎麼移才不虧？這是池子要解決的問題。And the tech is... automatically managed concentrated liquidity, because, like, for USD, it's clear that liquidity should be concentrated around price of one. But for non-USD, you have prices slowly changing, so you have to slowly move concentrated liquidity, and how you should move it to have it, like, not, you know, profitable, not lossy? It's a good question, and this is what the pools solve.

6:40:09能解釋一下怎麼解嗎？我們在 invariant 層面定義，讓你總能算：把集中流動性往這邊移，是賺還是虧、虧多少。Yes. And do you care to explain a bit how you solved this problem? Yeah, of course. So, well, one thing we... We essentially defined everything on the invariant level in such a way that you can always say, so if I move concentrated liquidity this way, will it be profitable or lossy, and how much do we lose if it's lossy?

6:40:37量測之後，我們會追蹤池子賺多少、花多少去移動集中流動性，不允許移動成本超過收益的一定比例。So it's very important to measure. And then we, when it's measured, we actually keep a budget of how much the pool earned and how much it spent to move concentrated liquidity. So we don't allow to move it if it costs more than a fraction of what it earned, right?

6:40:59最近 Curve 也用在 Yield Basis 上：可以有外部資金流補充這個預算，發行方可以 refill，用更高資本效率支撐自家資產。But also more recently, with Curve and actually used for Yield Basis as well, we figured out that we could actually have an external stream which refills this budget. So asset issuer, for example, could refill this stream, right, and it can refill the budget to move concentrated liquidity, and then, you know, more capital efficiency could be supported by that asset issuer.

6:41:35機構何時上鏈

6:41:35最後一個問題：我們已有相當有效率的去中心化 FX 市場、深度也不錯，要讓機構交易台、交易所願意上鏈在這裡換匯，需要什麼？And so maybe one last question I wanted to ask is that now we have this quite efficient decentralized market for foreign exchange, where we can have, like, good deep liquidity and everything. But what would it take for, like, the institutional desks, exchanges to actually, like, what would make them want to come on-chain to exchange here in this market, instead of staying where they are currently?

6:42:05我不認為現在就要拉大型機構上鏈，他們要的是大單很深的流動性，銀行間市場已經做得很好。Well, I think we should not try to pull large institutional players to on-chain markets yet, because what they want, they want really deep, big liquidity for their big swaps, right? They already have that with, like, interbank rates, and they are working really, really well.

6:42:30但他們對最小交易規模有下限，可能至少要一百萬美元；也不是誰都能進。小 fintech 進不了，只能走 Visa、Mastercard，換匯大約收 1%，發行方都不喜歡，所以它們會上鏈，我覺得它們會是第一批客戶。But they limit the minimal swap size on the downside, like it should be probably, I don't know, larger than a million dollars or something. And also, not everyone can have access to that, right? So small fintech companies cannot, so they choose to swap with Visa and Mastercard.

6:42:55不是自上而下，不是說服大銀行上鏈讓合規部門抓狂，要看小玩家的需求，它們先上鏈、流動性變大，也許有一天大機構也會用。Visa and Mastercard charge 1% per swap, and none of the issuers like that. That's why they try to go on-chain, and I think they are the first customers. So you don't start top-down. You don't, like, convince large banks to exchange on-chain, where their compliance will go crazy about it.

6:43:19可能要自下而上，而不是自上而下。You probably need to look at, like, smaller guys, their needs, they come on-chain, that will get liquidity bigger, and who knows, maybe at some point larger guys will use it as well if it becomes very efficient.

6:43:36你覺得會比較有機、自然發生，不需要什麼大變革？But, like, you probably go, we probably need to go bottom-up rather than top-down. Yeah, so you think it's going to happen a bit more organically, naturally, and that there is nothing, there is no big change that needs to happen.

6:43:51結語

6:43:51我覺得正在按該有的方式走，但是有機的。時間差不多了，非常感謝 Michael。是啊，謝謝這場好訪談。Yeah, I think it's going the way, the way how it should for that to happen, but it's organic. Okay, I think we're reaching the end. So, thank you very much, Michael. Yeah, indeed. Thank you for the good interview.