Alchemix卷土重来：Vault、固定收益与10倍循环 | DeFi Frontier

1. How Alchemix V1 worked

So, we started really trying hard to think about how we could fix this problem. How could we make this more scalable? How could we balance the protocol between the depositors, borrowers, and the LPs who actually provide value for these assets? With that in mind, we started working on our next version.

Thanks for joining us. How are you doing?

I'm doing great. Thanks for having me on. I'm pumped to talk about V3.

Yeah, so are we. This one's really cool. I feel like there was this magical moment in early DeFi when everything felt new and exciting. Those early protocols—for me, Uniswap was one of them, and Alchemix V1 was another one. I think I shared Alchemix with more people than maybe any other DeFi protocol at the time. This idea of self-repaying loans was just so simple for people to grasp.

2. Alchemix V1 and the idea for self-repaying loans

So, yeah, I'm excited to dive into everything that you've been working on for a long time with V3. But first, I think it's worth it to go back down memory lane quickly here. It's been a long time since Alchemix V1 launched. Maybe remind people what the original insight was here, and what problem you were essentially trying to solve with Alchemix V1.

When we started coming up with the idea for Alchemix, it was a little bit before DeFi Summer, and there were some proto-yield vaults that were coming out at the time. They weren't anything amazing. It was like 5%. This was in 2020. We were thinking, "What can we do with this yield? This is DeFi. This is Ethereum. We can compose things. We can build apps together with other apps."

We started working on an idea where you could lock up some DAI for a few months and get this other token. Then this token could be sold, and you could realize your yield right away. After your lockup period for DAI was over, you would come back and get your DAI. The yield from the DAI would be constantly buying this token off the market. That was the very proto version of Alchemix.

As we started seeing the yields go up more and more, and Yearn came out at 70% on DAI and 50% on USDC, we thought, "Ooh, I wonder what else we could do with this?" One of my other co-founders came up with the first version of Alchemix. He said, "Wait, instead of having this whatever token with that floating price, why don't we just turn it into a stablecoin?"

We deep-dived into that and realized that we could essentially create self-repaying loans. I think the only thing that was somewhat similar on the market at the time was 88mph, which is now long gone. But they had a lot of things that we didn't like about it, such as the lockup periods and different things of that sort.

We had the idea, and one of the first contracts we had was the Transmuter contract. We named it that before we were even named Alchemix. Then we had this little inception idea of an Alchemix-themed protocol. At the time, I was a big fan of Synthetix. One day in the shower, I was just like, "Alchemix! Alchemix! Alchemix!" You know, "Eureka!" I came running out of the shower and told the guys.

A couple of months later, we launched, and we were thinking it was just going to be some little boutique DeFi app. I'd be happy to get maybe 10 million in TVL because things weren't so huge at the time in DeFi. Then it blew up. It went viral. We ended up getting close to a billion in TVL at one point, at the height of DeFi Summer, around when we launched in early 2021. But that was kind of the second wave of DeFi Summer.

During that time, it was amazing, but it was also terrifying, seeing that number go up and up and up. That's when I started having trouble sleeping at night.

3. Pillar 3: Fixed term redemptions unlocking fixed yield

Yeah, that was a wild time. It wasn't until Nomadic just mentioned how much he used to talk about Alchemix that it reminded me of that example of the self-repaying loan. I remember there was a meme of someone who had bought a boat—or at least posted that they were going to buy a boat with an Alchemix loan—and that it would pay itself off with the yield. I think that was DeFi Donut, the guy who did that.

We also had stories of people buying cars, somebody who paid for their grad school using an Alchemix loan, and people paying hospital bills. These stories were quite common in the early days, with people sharing how they were using it. That made me happier than anything. It was just like, "Wow, people are getting use out of this."

Well, that's because Alchemix was so early to leaning into the concept of being able to borrow on-chain. All of us were excited for Aave, and obviously Maker's kind of the OG of that use case. But what was really compelling was that Alchemix was using what we now would all refer to as a yield-bearing asset.

The concept of borrowing against something that is continually churning out yield is so common nowadays. Again, you were so early to this, and 2021 was a weird time because of all the high gas costs. I think we were the first to actually use yield-bearing collateral for loans.

4. Lessons from V1

So wild. Looking back in hindsight, what do you think you got right with V1, and where did the design fall short? I think we can all agree that one of the things you clearly got right was storytelling and hitting on real product-market fit. But what else do you think you got right, and where did you see room for improvement?

I think one thing that we've always gotten right in Alchemix is that our protocol itself has been secure. We've never had—I think the biggest loss we ever took was $20,000, which was a Yearn strategy on Optimism, and they recompensated us anyway. Oh, also Jones DAO, which was also recompensated.

None of the core protocol ever really got hit. There were a couple of external yield providers that had some minor losses that were easy to cover. We've also had a hack on the Curve LP a couple of years ago, but that wasn't our code or anything like that. It was something that everyone assumed to be safe, and it was just some deep compiler bug. Other than that, I think security is one thing that we've really, really taken seriously.

We've had a little bit of draconian measures to do so. We've blocked smart contracts from interacting with it because of the potential for oracle manipulation with flash loans. In that regard, we've taken our duty of custodying people's funds really seriously. Not exactly custodying them, because we're not touching them directly, but any depositor—we treat them like their deposits are sacred and that we must protect them at all costs. I think that's one thing we've really gotten right.

In V1 or V2, I think the simplicity of it was that you just put in money, then you borrow, and then you can walk away. You don't have to come back; the loan will service itself.

I think that was one of the best things: people didn't have to come in and do all these manual clicks and pay all this gas to do it. We could do the yield harvest as an O(1) operation, distribute that yield, and repay the debt for everybody in the system. I think that was definitely really important at the time.

Because if people had to constantly do stuff, any gains they would have made in Alchemix would have been eaten by gas costs.

5. Ideas leading into V3

Scoopy, I'm curious what you've been thinking about the most since, say, V1. I know V2 was launched as well, but there's been, I guess, a long period of time for V3 to come to market. What were the main things you were thinking about? What were the pieces of the puzzle you were trying to crack while waiting to launch V3?

After we went from V1 to V2, we thought the big missing piece of the puzzle was the diversity of collateral and options for users. At first, only Yearn DAI and Yearn ETH were acceptable collateral in Alchemix V1. In V2, we expanded that, so we had more stablecoins: DAI, USDC, Tether, and Frax made the cut.

Then we had wrapped staked ETH, Rocket Pool ETH, Frax ETH, and Pyrex ETH, along with a few other yield strategies that made the cut for ETH as collateral. On that front, I think we did a good job of creating a secure system that gives people the opportunity to take their assets with them into Alchemix.

But there was one giant blind spot we had in going from V1 to V2. We were a victim of our own success and a victim of the bull market. When times were good, there was lots of liquidity, and securing liquidity was relatively easy, so the alAsset pegs were always pretty much rock-solid.

One of the things was that our redemption mechanism was just the Transmuter. It was never really at capacity, so everything seemed like it was rock-solid in that environment. Then the whole FTX thing happened. The Luna/Terra Luna thing happened. Liquidity started drying up, token prices started getting depressed, and all that horrible stuff that came with that blowup happened.

As redemptions started flowing in, as people were exiting DeFi and getting out of Alchemix and DeFi in general, we realized that our pegging mechanisms were wholly insufficient. What was once very capital-efficient became a grind to secure LPs. LPs were a little bit sad, or complaining, because they were kind of an afterthought in the V1 and V2 protocols.

If they wanted to get out and the pools were not in favorable conditions, they would have to wait a very long time in the Transmuter to get their money out. It was not a very efficient process. As a result, especially alETH's peg has suffered, because it's a very popular product even to this day.

Because we had to secure $1 of TVL for $1 of loan, essentially, to back it for people to swap and trade, it became prohibitively expensive. One thing we had to do to survive was find this equilibrium of how much TVL in our protocol was sustainable and profitable for us. If we went over that, we would start bleeding costs to LPs and things like that to incentivize them to stay.

That capped our growth and had us stuck in this $30 million to $60 million TVL range for alETH. With this in mind, we knew this limitation back in late 2022 and early 2023, so we started trying really hard to think about how we could fix this problem, make it more scalable, and balance the protocol between depositors, borrowers, and the LPs who actually provide value for these assets.

With that in mind, we started working on our next version.

As we were planning and preparing to steer the conversation with you, we were thinking about V1 and where all of this started. I want to run through a few of the basics of V1 before we introduce all of the work you've been doing with V3.

Just in case someone is newer to DeFi in 2026, talk to us about the idea of when folks would deposit something like DAI or another stablecoin in V1. What were they borrowing, in terms of how that use case worked? We're not putting up a volatile asset and then borrowing a stablecoin. We're borrowing a synthetic asset that is correlated, or should follow the price almost to a certainty, in terms of the underlying collateral.

Just remind folks: there's alUSD, there's alETH. What were some of the examples in which folks were using Alchemix V1?

I'll walk through how V1 worked. V2 is very similar, except it has more collateral options. The idea is that you put your deposit into Alchemix. Let's say we go with DAI: I put in 1,000 DAI.

Upon deposit, the protocol will mark it to market. I put in 1,000 DAI, and then it will wrap it in a yield token. It will then detect, through the yield token, that there has been enough of a gain in that yield token, and it will do a little liquidation of that extra money.

It will keep that mark-to-market value globally. Everything extra will be liquidated and then used to repay people's debts. The part that gets liquidated goes into the redemption mechanism, the Transmuter.

The yield was what would basically back the currency. If alUSD was off-peg, you could put alUSD into the Transmuter and get DAI 1:1. You put in 100 alUSD, you get 100 DAI back, so you can arb any peg differences there.

In the beginning, we had this crazy phenomenon where we had debt caps, but we didn't have deposit caps. There was also a period where the alUSD peg was above 1, so people would self-liquidate themselves in Alchemix. They'd use their collateral to liquidate themselves, which would bring down the debt cap, and then they'd borrow more.

As people did this more and more, they could get into the LP and take advantage of that arb. It stuffed our Transmuter with tons and tons of money. At one point, I think there was like $200 million just in the Transmuter. Because of that, we thought our pegging mechanism was great: the Transmuter had everything it needed to handle redemptions. But eventually, that did dry up.

At a 50% LTV, people could loop it and essentially get double the yield by multiplying their principal. A lot of people did this kind of save-and-spend deal, where they would put it in there and then borrow alUSD as they needed it.

Other people used it to go long other assets. You put in your DAI, borrow alUSD, and then buy ETH with the alUSD. So you have this kind of hedged long position. People used it a lot. That's the way I used it, and I did increase my ETH stack by a decent amount doing that.

People have lots of different use cases for it.

6. V2’s weaknesses

If we go back to 2022, when you mentioned that at some point things were not working as intended, we were living through an apocalyptic time. There were all sorts of centralized players in the space that were failing. That was putting pressure on DeFi because a bunch of them, like Celsius, were actually deploying capital into DeFi, and suddenly they were trying to sell all their assets and pull their money out.

In terms of what you were seeing with Alchemix, when does this peg design suddenly reveal that it's not working as intended, in the most apocalyptic of times? Markets are selling off, stablecoins are depegging, and you've got the Terra collapse, where $40 billion evaporated over 5 days or so.

Just tell us whatever you can recall from that and, to dumb it down for folks who weren't there to experience Alchemix V1, because this is a great setup for us to talk about V3. We understand that a lot of this ultimately influenced the work that you guys have been doing on V3.

It was just the loss of confidence in DeFi, seeing a lot of liquidity exit, and more and more competition entering the space. Instead of people having maybe 10 options on Curve to choose from for LP pools, now there were 200 options. People were playing the Curve Wars games and stuff like that, and that started getting more competitive.

At one point, we could incentivize Convex and Curve and get a 1.5x multiplier on that stuff. Now it's compressed to basically 1:1.

There were lots of little things that happened. But overall, I think it was the crisis of confidence that happened, where people started leaving in droves, liquidity dried up, and as people were exiting, they basically used everything that was in the transmuter. Once the transmuter was empty, we found that people would dump their stuff on the Curve markets. More and more, it stressed the protocol, so we had to find some different ways forward.

7. Audits

One way we were able to stop the bleeding was to institute something called an AMO, which is a design that Frax came up with. It essentially allows us to repurpose some of the stuff that was still left in the transmuter and deploy it into Curve pools and things like that. As the peg has been drifting, we've been slowly decreasing the size of these AMOs. But I think there's still maybe $8 million in total TVL in our AMOs, which is definitely a good tool that we have for balancing things and recapitalizing. We can also make some money on the float at the same time, which makes us a bit more sustainable.

8. Overview of three pillars of Alchemix V3

Scoopy, I want to get into all the details of how V3 works, but maybe first let's start at a high level. I'm curious how you're thinking about V3: is it a totally new protocol, or is it just further refining all the things that you've learned across V1 and V2? Maybe just start there at a high level: what is V3?

I think when users come to V3, they'll think it feels very familiar. But at the same time, it is a complete redesign. It could actually be its own protocol, with its own name, whatever, but we're just going with V3.

The idea was that V2 and V1 were not very capital-efficient as far as us running a protocol goes. For every dollar of debt or synthetics that was generated, we needed to have that much in the LPs as well. We didn't really have good token sinks or financialization. One reason for that was that people looked at the protocol and thought the peg mechanisms were not as resilient as they should be. We knew this deep down: if it weren't for some of the early moves that we made in our treasury, with the AMOs and things like that, our protocol would not have been sustainable.

We really went back to the drawing board and tried to design it in a way where, even if there were no token to incentivize yields and reward different behaviors, it would work anyway. What if ALCX went to zero? Could this thing still work? That was kind of the design philosophy behind V3. If you want, I can go into a deep dive starting right now, or if you have any other questions before I do that, let me know.

Yeah, actually, we've got an early version of the V3 docs. I think they're a work in progress, but there are these three pillars in there: vault yield, synthetic borrowing, and fixed-term redemptions. Maybe walk us through each of those and how they flow into the operations of V3.

All right. There's a lot to unpack. Let's go.

I'll start with the vault yield. In our V1 and V2 systems, we would plug into other yield providers, like Yearn being one of them and LSTs being others. Especially in V2, these were all siloed away from each other. If you had wstETH, like wrapped staked ETH, you could put it in there. If you had Yearn ETH, you could put it in there. These would be different vaults that you could borrow from.

There's a problem with this, though: some of the yield vaults would become deprecated. There's a Yearn USDT vault that got deprecated. We updated it and made a new one that corresponds to their new vault, and we even have a migrator tool for people to get over there with one click. But we still have around 30K in that deprecated USDT vault.

We also have other yield strategies whose yields have been underperforming compared to other yields. When you have this kind of global, multiplayer system, and some users are paying more into the transmuter than others, it's not very efficient. So what we wanted to do was bring all of the yield into one token.

We're running a customized version of Morpho V2 vaults now, which allows us to deploy up to 15 strategies simultaneously. That will allow us to diversify the yield, control the amount of risk in the system, and make sure that there's no single point of failure in the system—a kind of strength-through-diversity idea.

It also improves the user experience. When users come in, one of the things they liked about V3 or V1 is that it's dead simple: “Okay, deposit. Cool. All right, borrow. All right, I'm done.” Whereas in V2, you go to the grocery store and see 50 different brands of pasta sauce, and you just have this paralysis of decision. What should I do? So we really wanted to simplify that user experience.

Us being a DAO and wanting to keep this decentralized, one of the customizations we made to the Morpheus V2 vault is that our DAO is the risk curator for it. There are votes for the different weights of the different strategies that are in the yield token. On top of that, anticipating a lot of degens wanting to enter this space, we have risk classifications for different strategies. We have low, medium, and high risk. High risk is capped at 10% of the makeup, medium risk is capped at 30% of the makeup, and low risk is uncapped.

9. Pillar 1: Mix-Yield Token (MYT), Alchemix's own vaulted yields

That way, we don't have too high a concentration in any one risk category, including any single high-risk strategy. If one high-risk strategy gets a 5% allocation and goes to zero, the yield token is only going to take a 5% hit. It's not going to be super catastrophic. That's how we're trying to have a risk-adjusted yield: something better than Aave, something better than Yearn, but not quite so degenerate that the whole thing is going to blow up.

Scoopy, just before you continue, I want to call out these vault yields. This is a net-new feature, right? I may be misremembering, but there's nothing like this in V1 or V2. You used to allocate assets to a third-party yield provider, but this is your own vaulted yield now.

Yes, that's correct. This kind of vaulted yield, we're calling it the mixed-yield token, and this is something that we're going to be rolling out. It's going to be our product—a standalone product. If you don't want to touch borrowing or anything like that and you just want passive yield, this is something that you can get into. It would be suitable for that, but it also doubles as collateral in the Alchemix V3 system.

Sorry, before we move on to that, I think we're going to get into borrowing here, but do you have any early napkin-math benchmarks of what yield you think you can get with the risk makeup that you mentioned? I'm trying to think what this collateral could post as yield.

We did a lot of our modeling back in Q4 of 2025, and we were looking at something around the 5%–7% range for USDC, based on the makeup of the yield token. Then, around 4%–5% for ETH, because its yields are a little more compressed compared to USDC.

One thing that I really like about this is that a lot of the yield strategies we're hooking into are similar to the strategies we're hooking into for V2 anyway. It's just putting them all under one roof. We can take third-party yield tokens into our own yield token, or we can develop custom strategies for it. So we have that flexibility.

One protocol that we're going to be working with is Tokemak, which I guess is now Auto Finance. One reason we want to go with them is that they have an elegant way to use LPs and turn LPs into a yield token. The infrastructure for that is actually quite complicated and a little bit on the riskier side, particularly when it comes to protecting against oracle manipulation and things like that.

By being able to write our own strategies and integrate other protocols that already have their strategies built out, it gives us maximal flexibility in our yield token.

And we can actually—we’re not competitors directly with anyone. So, if we integrate Pendle into our yield token, we’re a net benefit for Pendle. We’re not taking anything from them. The same thing applies with Yearn: if we have a Yearn token in there, we’re going to be benefiting Yearn and not necessarily be a direct competitor with any of them.

10. Pillar 2: Synthetic borrowing up to 90% LTV, zero interest, 10x leverage

Okay, I think I’m clear on the collateral side and this new vault infrastructure. So, why don’t we talk about borrowing against this mixed-yield token, this yield-bearing collateral? I think that’s probably where you were going to go next.

Yeah, the next 2 pillars are very intertwined: the synthetic borrowing and the fixed-term redemptions. I’ll just start off with the synthetic borrowing. It’s quite simple.

When you get in the system, you have your mixed-yield token, and it will read the value using its internal oracle for what its cost basis is. Let’s say it’s saying 1 mixed-yield token is worth 1.1 USDC, right? Then, however many of those you put in, you multiply it by that 1.1 multiplier for its price-per-share value, and that’s your principal basis. As it goes up, the value of your principal will go up as well. So, there’s no more mark-to-market in the V2 or V3 system.

Based on that collateral, you can borrow up to 90% of your LTV. You put in 100, you can borrow 90. With this, if you borrow 90, then you can sell it for 90, put that 90 more into Alchemix, and borrow 81. Then you sell the 81 and put it into Alchemix, and you can loop this. It can essentially approach 10x leverage for your position. You put in 100, and you can get close to 1,000 in principal.

You will have a lot of debt in that position, but now you have 1,000 collateral earning that 5% yield instead of 100. Since the debt is denominated in essentially the same asset as the collateral, it’s neutral in that regard. The debt doesn’t really matter because you now have 10 times more collateral earning that same yield. It’s like you got 10 times the yield.

You would think, “Okay, if we have this system, it would probably be very stressful for the pegs because people would be borrowing and dumping, borrowing and dumping, borrowing and dumping,” right? To get around this, or to balance the system out, we have a new redemption mechanism that employs fixed-term redemptions.

People will go to our Transmuter, which is our redemption mechanism, and they can put in, let’s say, 100 alUSD. It’ll say, “Okay, it’s 2 months to convert this to USDC.” They put it in, and it’s deterministic: at 2 months, they will be able to hit that button and get 100 USDC out of it. At any time in the middle of it, since it’s maturing block by block, they can do an early redemption and get the amount that has actually been converted in that time frame, then get the remainder that’s unconverted back in alAssets. So, it has a lot of positional flexibility for doing so.

There’s 1 key innovation that we had to do to make this whole system work: we have to guarantee that these funds can be paid to these people using the Transmuter. Where does that come from? It has to come from users’ collateral. That’s the only way that this can actually work.

The key innovation is that these are fixed-term redemptions, and we have a system where the debt in the system is actually 2 kinds of debt. There’s normal alAsset debt—that’s your alUSD and your alETH debt. But as these redemptions start to mature, the Transmuter will derive a redemption rate. You could think of this as a reward rate from a Sushi MasterChef contract.

This will apply to your debt, or all the debt globally, but on an individual basis, it applies to your share of that debt. It will slowly start to convert your debt into earmarked debt. This earmarked debt is a subset of the debt that is slated for redemption in the future.

While it’s being earmarked, it is not taken from you. Your collateral is not taken from you, so you get to earn on your full collateral even during this earmarking process. We’re calling this a temporal advantage because, unlike Liquity, it’s not being taken immediately. The same is true with ReSupply—it’s not being taken immediately. You can see, “Okay, in the future, I’m going to have this much debt repaid, but also the same amount of collateral taken to repay that debt.” So, it’s like an automatic deleveraging system.

The genius behind this is that we were able to do this globally in an O(1) operation. For the non-computer nerds out there, this is just 1 operation. I don’t have to loop over it for every single position. It’s just 1 operation, and it’s super gas-efficient.

As your debt gets earmarked, it can now differentiate between your earmarked debt and your unearmarked, or normal, debt. It does this lazily, so it doesn’t need to be updated at all. It always knows how much your ratio of normal debt to earmarked debt is, and it applies only to your unearmarked debt. Anything that gets earmarked is almost like it’s not being counted toward your share of the debt anymore.

If new people get into the system while you’re being earmarked, you’re actually going to have a smaller and smaller share of the debt. This is a really cool innovation, and it has lots of use cases beyond self-repaying loans. We’ll probably be building other protocols with it as well because it’s kind of like black magic. It’s really crazy how this thing even works.

I want to talk about some of those other use cases and where you’ve drawn inspiration from in terms of other DeFi primitives. But just to go back to the basics for a moment, can you remind us: if we are borrowing alUSD or alETH, there is no borrowing cost, right?

Yes, there’s no interest paid on the debt at all. It’s 0% interest, so you’re getting free leverage. There is a fee on the yield that the mixed-yield token generates. When you do get redeemed against, there’s a small fee applied to your collateral.

Let’s say you have 100 redeemed. There would be maybe 100.5 of your collateral taken and 100 of your debt repaid. So, there’s a small fee applied to that. In a heavy redemption scenario, you might see a maximum 2% interest rate effectively. Either way, it’s still ridiculously low and much better than anything else in DeFi in terms of the cost of borrowing.

11. Liquidation risk in Alchemix V3

With V1, I used to describe, whether I was writing a blog post or talking to friends, that there was no liquidation risk. Is that the best way to frame up V3, or what would you warn users of? Is there any sort of liquidation risk?

V3 is at 90% LTV, so we do have liquidations in the system. But since it is a delta-neutral system, it’s not reliant on price oracles. There’s really only 1 scenario that can cause liquidations, and that’s if yield strategies take losses. That’s just to safeguard the solvency of the protocol itself.

It will only liquidate as much as it needs to make your position solvent. We have a 90% LTV and then a 95% liquidation LTV. If you cross below that threshold, it will essentially auto-deleverage you and liquidate you until you’re in a healthy position. It’ll get you to 89% LTV.

In that regard, it’s not quite as ironclad as V2. But with V2, if there was a major loss, the loss would be socialized anyway. So, it’s just a different way of dealing with it.

Overall, as long as the DAO does a really good job of risk curation and we, as developers, integrate these other protocols correctly, the plan is that there will never be liquidations in V3. But they’re there in case they’re needed.

12. Tracking where V3 strategies deploy money

Is there any dashboard that users will be able to use to track where these yield strategies are deploying all of this?

Yeah, it’s actually built into the V3 webpage itself. I can show you a little prototype of it.

If you go into a vault, it’ll basically have the strategy information right there. You can see what the makeup will be. This is just a little prototype, so things aren’t hooked up in it just yet. But you’ll be able to see the full list of strategies. There’ll be up to 15 per vault, and you can see their allocations.

If you want to get more in-depth about each of the strategies, you can go into the DAO voting page for them. There’ll be a dropdown here with explainers for the different yield tokens. We’re going to build it out so there are external links and contracts that you can see as well.

It’s going to be fully transparent. People will know exactly where their money is going, and there’ll be little descriptions explaining how it makes its yield.

13. V2 to V3 migration

Scoopy, in terms of the alUSD and alETH from V1 or V2, will these be new tokens, or what happens to what you had from V1 or V2? Can you use those interchangeably? Can I deposit alUSD into V3?

Yeah, the reason why we’re doing a migration from V2 to V3 is so we can keep the same assets, because fundamentally, V2 and V3 are not compatible with each other.

But if we shut down V2 and just take the assets with us over to V3, then it works just fine. We’d much rather do that because we’ve made lots of headway with oracle providers and LPs and all that sort of thing. We don’t want to start from scratch with new synths.

14. A new fixed yield primitive

That’s awesome that everything’s so transparent and seemingly fully on-chain. Where should we go next, guys? I have one more thing about fixed-term redemptions.

One side effect of having fixed-term redemptions, and the fact that you can get one-to-one asset value for the underlying, is that this creates a fixed-yield primitive in DeFi—a new fixed-yield primitive.

Imagine you see alUSD at 98 cents and there’s a 2-month redemption period. You can make a 2% arbitrage in 2 months. If you annualize that, that’s a 12% yield right there.

In a sense, we’ve created something a little bit akin to Pendle, but instead of there being a bajillion asset-YT-PT pairs and things like that, we have one that’s integrated into Alchemix itself. If you see alUSD is cheap and the redemption period is short, you should see dollar signs in your eyes and get that yield. You can make that money.

The thing is, this is how we balance things with LPs now. When the alAsset price goes down, the demand for it goes up because people can now turn that into a yield or an arbitrage that’s guaranteed for them. Likewise, when the price of the alAsset goes up, it makes leveraging more attractive because the cost basis for each loop becomes cheaper.

In a way, this is going to be an oscillator—something hybrid between a stablecoin and a bond. You can come in when its price goes down, when it becomes very attractive to buy and then redeem. Then, when the asset’s price goes up, it’s very attractive to leverage. It’s kind of in the superposition between being a stablecoin and a bond.

For LPs, the wonderful thing is that this will create a lot of churn in the pools. That means a lot more volume for them, which means more natural APY. That means the Alchemix protocol won’t need to incentivize it with tokens as much.

If LPs ever want to get out, and alUSD’s pool is really heavy with alUSD, they might get out with a lot more alUSD than USDC in the pool. They can just turn that impermanent loss into a permanent gain if they’re patient enough to wait out the transmuter. So, it balances everything between the different participants in the system. LPs are now first-class citizens, just as borrowers are.

Wow. Honestly, I’m just trying to keep up. There’s a lot of new stuff in here, it seems like. One of them that I want to go back to is the looping and this idea of yield boosting, or boosted yield, that you’re referring to in the docs.

15. The tradeoffs of looping in Alchemix V3

If there’s no borrow cost, I’m trying to think about what the trade-offs are. You might have told us already, but what are the trade-offs or downsides to looping? Where are you running up against paying more fees by being in a position for a longer period of time? What’s actually hurting a looper? It seems almost too good to be true the way I’m thinking about it currently.

Like I said, the peg will oscillate based on user behavior and things like that. The timing of your looping actually matters a lot. If you’re looping when alUSD is 99 cents or above, there are very few scenarios where you won’t come out ahead—massively ahead.

Even at 98 cents, according to the simulations I’ve done in a high-redemption-rate environment, you’re still going to be better off looping than not looping if yield is your goal. But if the asset price goes down—let’s say it’s 97 cents—and it’s a very high-redemption-rate environment, it might end up costing you more to loop than just being a passive user.

So, there is some timing element to it. You can’t just blindly YOLO into it. You have to have a little bit of awareness of the environment that you’re in.

We’re going to try to make that pretty transparent and have tools. We’re working right now on a looper tool. It’s not going to be ready for launch, but in that tool, it’ll tell you your execution price and stuff like that as well.

There’s also simulation built into the UI, so you can see, “Okay, if I loop at this point, and if there’s this redemption rate, and I’m looping at this price, what’s my expected P&L after 2 years?” There’ll be a little simulator tool in the UI that users can use and play around with.

In the looper tool, it’ll tell you your execution price, and if it’s bad, I’ll probably put up a little warning saying, “Probably not a good idea to do this,” just so folks are aware.

Just so folks are aware, we’re recording this quite a few weeks in advance of it going live, so forgive us. We’re trying to navigate this future app and tell the story of how we might use it as users.

Going back to fixed-term redemptions, this is my favorite part of Pendle. I love the ability to earn a predictable yield. I’m definitely a PT bull. I prefer that side of the equation versus the YT trading.

This fixed-term-redemptions feature is what I’m most excited about, just going off of what we learned here. Is it a single option that will be available in terms of the current offering for fixed-term redemptions, or will we see an array of choices based on the different types of loans that have been created through Alchemix V3?

I’m trying to think through what I’m going to see as someone who’s shopping for those fixed-term redemptions. I’m wanting to earn, to your point, something like a 10% APY yield over the course of 2 months. What can I expect to see?

We’ll have a redemption page, and on that there’ll be assets for USDC on mainnet, Arbitrum, and Optimism, and likewise for ETH on those 3 chains as well. Each chain will have its own little bit of variance.

The fixed yield that comes from these redemptions is highly dependent on the alAsset price, so it’s a timing thing. If some whale comes in and 10x-leverages their position, you’re going to see really good rates. You buy alUSD at 95 cents and there’s a 2-month redemption period. That’s an implied 30% yield right there for doing this.

The thing is, there aren’t going to be different maturity dates and things like that. It’ll be one redemption period that’s global for the transmuters. Each transmuter might have its own, but each transmuter has one global redemption rate.

It’s not like, “Hey, we’re offering you 15% yield. We’re offering you 10% yield.” It’s more like, “Hey, the market is at a discount, and right now, if you were to buy it off the market and put it in, this is the yield you would get.”

When you put it in, you make that trade and lock in that price. But that fixed yield isn’t fixed itself. It’s only fixed once you execute that trade and put it in.

16. Example of how a loan, looping, yield works in V3

Given everything that you’ve shared, I do want to do one final example. If you’ve been getting lost in any of the detail of V3, keep your eyes open for when it launches. Of course, follow Scoopy and Alchemix on Twitter, and we’ll be making lots of noise and sharing our own experience using it.

Let’s talk about the example you just pointed out with the whale. Let’s pretend a whale deposits, I don’t know, $5 million or $10 million into an alUSD-type vault. Let’s also pretend that they decide to borrow and loop. Talk us through some of what happens under the hood there, and then what are some of the opportunities that could present themselves with the fixed-term redemptions you’ve covered?

If a whale gets in with $10 million and loops at 10x, essentially they’re going to be creating close to $90 million in alUSD as part of this. With each loop, they’re essentially going to be selling it on Curve, in Balancer, and in the field and drone markets.

If you know anything about DEXs, you know that sales do affect the price. This would bring the price down quite a bit. It would depend on how much liquidity was there in the first place, but I would expect that this would have quite an impact on the peg.

Something like this in our current environment would probably take the peg down to around 70 cents because we just don’t have that much liquidity to service it.

And so that depressed alAsset price then becomes an arbitrageur’s best friend because they know that they can turn whatever that price is back to 1. All you have to do is wait a little bit of time. So let’s say this does happen: It pushes it down to, let’s say, 80 cents. alUSD is now an 80-cent shitcoin.

DeFi Dad sees this and he’s like, “Oh, okay. I understand how the system works. I’m going to buy as much alUSD as I possibly can, and then I’m going to deposit that into the Transmuter.” Let’s say you buy $1 million of it. I don’t know, you’re a whale, I’m assuming. No, I’m just kidding. [Laughter] When you transmute it, you just have to wait, and then you get all that USDC—the same amount of USDC as the alUSD you put in.

So you’re buying a dollar for 80 cents, and all you have to do is wait however long the redemption period is. We think 3 months will be the maximum it’ll ever be. It’ll be 3 months or fewer. Then, if you do a little bit of napkin math, you can calculate your APY. At 20%, let’s go again with the 2-month redemption period: You’re getting 20% every 2 months. Times that by 6, you’re getting 120% fixed APR, in that sense.

After you buy that, maybe the price of alUSD is now at 85 cents. Then another one comes in and is like, “Wow, this is still a really great deal,” and buys a bunch and puts it in. Now the peg is at 90 cents. Then somebody else says, “Well, even though it’s not as good as it was just a second ago, it’s still a really good deal.” More and more and more arbitrageurs enter the system, compressing the yield and bringing the peg back up.

17. What’s a doomsday scenario for V3?

Okay, so we painted a picture of how this all works as intended. Is there anything you’d call out in terms of what a worst-case scenario here looks like? How does this play out if my fixed-term redemption, which I’m waiting a few months for, doesn’t return alUSD back to peg and create all this yield for me to ultimately earn a fixed yield?

How could this go south? What are some of the risks here to consider? How does the protocol overcome those challenges? Let’s say there isn’t enough buying pressure in the market. What happens?

I’m not worried about the buying pressure in the market. I think over time, longitudinally, it will self-correct. But let me walk you through a doomsday scenario.

Like I said earlier, the only time liquidations ever really happen is if a yield strategy takes a loss. Let’s say something catastrophic happens and the yield token takes a 20% loss, but the protocol was levered at, basically, 85% LTV globally. That would put Alchemix V3 into insolvency mode, essentially, because the liquidations would not have been able to overcome the gap there.

When insolvency mode happens, there is some risk for people using the Transmuter because the redemption rate will be re-rated to the level of solvency of the protocol. If we’re 90% solvent, instead of getting 1-to-1, you’ll get 0.9, or you’ll get 0.9 for it. In that sense, there is some risk involved. But again, this is a black-swan, worst-case scenario.

Once Alchemix is in maintenance mode or insolvency mode, it’s essentially going to block all deposits and all borrowing, and it will just heal itself over time. If it’s a small insolvency gap, it might only take a few weeks or a month to clear up. But if it’s a large one, it might be in that mode for a very long time. It’s hard to tell.

That is a real risk, and I do want to be upfront about that. If something like that happened, we would basically be a dead man walking for a little while. We might have to redeploy new vaults and new synths and stuff like that if we want to keep ourselves running at that point. But that is the doomsday scenario.

According to our test suite, it should be able to recover from this doomsday scenario, and people should be able to be made whole in due time.

18. Zero interest on debt but what do we pay?

Scoopy, that’s kind of a good segue into something else I wanted to talk to you about. I want to do a more rapid-fire section here. The first one is audits. What can you tell us about how many audits you’ve gone through? I feel like I’ve seen that you’ve gone through a lot, based on some of the stuff I’ve seen on Twitter. How do we know that this is secure?

We’ve been audited—let me count. We did Spearbit, Cantina, Immunify, and Nethermind. We’ve done Yearn, and we’ve also had a solo security auditor, a researcher, audit as well. So, 6 human audits.

We were going to be launching Alchemix V3 before ETHDenver. Yearn has this tool called Zero Cool, and they were like, “Hey, let’s just run this on V3 before you launch.” It surfaced a few medium-severity issues that we hadn’t seen before, and we were like, “Okay.”

I’m pretty far ahead of the curve on the AI stuff. I collected a bunch of auditing skill packs that I made a ton of myself, ran even more tests, and found a few more cases. We were furiously trying to patch things before our intended launch. My bots were finding things, Zero Cool found things, and we ran Cantina’s AI tool and it found things. We were like, “Oh, okay. This is a problem.”

So we beefed up the testing framework a ton. We made cases for the doomsday scenario I just described. We made the most insane test cases and invariant testing—all of the stuff, end to end, every single contract, every single line of code, making sure everything got hit.

After we did that, we found a few more little bugs and were able to make some more patches. Then we said, “Okay, we need one more human audit.” That’s why we went with Yearn and did their yAudit, which just wrapped up.

The good news is that no critical issues were found. The findings they had were kind of nitpicky, in a sense. Some of them were very useful findings that we’re patching up, but others were more informational-level findings and stuff like that.

We’ve now got the fixes in, and we’re going to run it through the gauntlet of AI tools again. If all that comes back clear and we’re really confident in things, we’re going to go ahead. I’m extremely confident in our core Alchemix and Transmuter protocols—our contracts. I think those are extremely rock solid.

Going forward, I think the biggest risk for the system is essentially external yield providers that are hooked up into the Alchemix yield token. That’s something we’ll always be continually monitoring and taking risk assessment seriously.

We have a lot of DeFi veterans and pros on our team. We have data scientists on our team, including someone who’s also an accountant and is really good at crunching all these numbers and assessing risk. We have our own internal risk-assessment framework and rubric that we use to grade different protocols and determine their suitability and, if they are suitable, what their risk level is.

Those will all be published and made publicly available for everyone to scrutinize as well. We’re definitely aware of where the weak link is in the system, and we’re taking tons of precautions to make sure that link is not a failure point in the system.

I’m obviously by no means an audit expert, but a lot of the names you rattled off there were ones that I’m definitely aware of, and some of them have what I think are the best-known reputations in the space.

What a weird time, too, where AI tooling is progressing so fast and you’re using it in tandem. That must be a bit of a relief—that now you can throw basically all of this at the code that you’ve built to test for any holes. It’s a double-edged sword because the bad guys can do the same thing, too.

Exactly. You might as well be using it, though. You have to use it now because others are potentially using it to find holes, right?

I’m relieved to hear that the audit was not done by ChatGPT.

It’s Claude. It’s Claude.

19. Could curators be involved in Alchemix V3?

Yes, it’s fine. Yeah. Scoopy, I also want to ask you about curators. Obviously, when V1 came out, the curator space wasn’t really a thing, but now it’s growing rapidly. I’m curious if you’re building this in a way that expects curators to build on top of what you’re building here.

The system that we have is a DAO voting for the weights of the different strategies in the system. We’re more than open to working with curators so they can give recommended weights, but ultimately it’s going to be at the protocol level, the DAO level, for controlling that stuff.

We have been in talks—I might be wrong about this—but I think we have talked to Gauntlet about possibly doing some risk curation. I think it’s probably at the strategy-selection stage that we would be consulting with them.

At the same time, there’s a lot of publicly available information about what the allocators and risk curators are doing on the Morpho markets. We can glean a lot of information from that as well to inform our decisions.

20. New ALCX tokenomics coming

And then, Scoopy, before we let you go, are there any discussions about redesigning the Alchemix token? The reason I ask is that I recognize Alchemix was launched during a totally different era.

We had this dark cloud hanging over the industry with the Gary Genslers of the world. Things have changed. There's a much more pro-crypto regulator mindset across the globe now, especially within the United States. Things have changed drastically. So, is there anything that you can share on that, or anything that we can look forward to?

The ALCX token is not going anywhere. We're not going to do a rebrand, a token migration, or anything like that. I think that's BS, and I think that's disrespectful to the token holders who are still there. Even if the chart looks bad, we'll make it look good eventually, hopefully.

We do have a whole new tokenomics design that's coming out, though. That will basically marry it to the protocol and enable full on-chain governance. And with that, revenue sharing as well. This is actually a very radical design. I'm not sure if you know who Block Enthusiast is on Twitter?

Yes. Yeah, I follow that account. Yes, I do.

He's a mad scientist. He came over from Reserve Protocol. He joined us from there, and he is working on this really wild system.

We're calling it the V-Cube, the vesting cube. It's kind of like you take the best parts of the Curve voting escrow contract and the ETH 2 staking cube. We have lots of control valves to work this through. But essentially, you would have to queue into the system, and then you'll have a vQALCX token. This would be kind of like a senate token. It does general governance things. But then you can recursively queue this vQ token for product-specific things in the DAO.

There'll be an alUSD DAO. There'll be an alETH DAO. And for any other future products, there'll be another DAO—a recursive DAO inside of it for that. It's set up so that wherever you recursively put your token, you get the revenue stream from that product itself. This is a way for our users to signal what is important to them and to double down on those things.

But also, if they want to have a new thing—let's say enough people in the DAO, token holders, want to have an alBTC product—if they can recursively, permissionlessly spin up their own sub-DAO inside of the system, then that will start giving some token inflation toward a path that will then fund the development of this new product.

This is a very radical departure from other things. This will be hooked into the Aragon OSx governance platform as well. So, it will basically transition Alchemix into a fully on-chain DAO. This is very far in development, so it's something that we're very excited to get out. As soon as V3 is done and deployed, we're going to have the rest of our smart contract team take a look at all of Block Enthusiast's work and get that thing ready for audit.

And that's going to be a public good. We want to push this far and wide. We want a lot of people to adopt it because we think it's awesome.

That's so exciting to hear that all of that is happening in conjunction with the work you've done on V3. It is so refreshing just to hear the focus and the dedication that you've had to building this protocol. There are just so many founders who have quit over the years, and it's rare to see a team stick to their guns here and keep building. So, we're so excited for V3.

We really didn't know much more than the surface-level details coming into the conversation today. But we all loved V1, and we really had great stories to tell about the self-repaying loan use case. I really look forward, again, to this idea of fixed-term redemptions and then building upon all of the DeFi primitives in 2026 that can assist in a better V3.

I think the truth is, when you go back to V1, there just wasn't a lot that was built. You guys were pioneers in the space, and now there's just so much more that's been built out that you can build on top of, and you can benefit from a composability standpoint.

21. Closing

It's wild to hear all the net-new ideas that you guys have packed into V3. So, for folks who are listening, hopefully in the future V3 will already be live. But you're really getting a first look at what's coming.

And then, Scoopy, thank you so much for coming on with us and for going into so much detail about everything that you guys have been working on. I want to give you the final word and wish you the best of luck with the launch.

Yeah, I want to touch on that point about why I'm still here. First of all, I love my team, and I don't want to let them down. I want to stand by them, and I know when founders leave, it creates lots of really bad scenarios for the people who are left. I don't want to do that to my team because they're freaking awesome.

Above that, I'm excited for what we're building with Alchemix V3, and I want to see it, steward it, and make sure it grows and it's healthy. But even bigger than anything else, I'm on a mission, man. The world is so up, and I think there are all these systems that are designed to keep people down and keep people in traps. I want to create a better system that allows people an exit—an alternative path that's not predatory and that works for them. That's, at this point, my life mission, and I'm not quitting. I'm not leaving.