Today in DeFi：國債代幣也脫鉤，退出前看什麼？

0:00開場：今年資產損失已超過 10 億美元

0:00好。我是 Danger，今天要講怎麼安全使用 DeFi。我是誰？我是職業 DeFi farmer，從 2020 年做到現在。我有一個媒體叫 Today in DeFi，做很多 DeFi 研究，研究結果會回頭餵我們的 farming。我們做風險研究，也做 alpha 研究，找出好的農場。為什麼要盯風險、做風險研究？大家都知道，AI 正在加速駭客與攻擊。Cool. Yeah, my name is Danger, and today I'm going to teach you how to use DeFi safely. Yeah, so who am I? I am a professional DeFi farmer, so I've been farming since 2020. And I have a media called Today in DeFi, and we do a lot of DeFi research, which informs our farming, basically. We do risk research, and we also do alpha research, so we find good farms. Okay, so why do—why pay attention to risk? Why do risk research? As everyone knows, like, AI has been accelerating hacks and exploits.

0:45今年資產被攻擊損失已經超過 10 億美元。跟往年比，有意思的是：即使 AI 參與很多，今年多數駭客與損失，比較不是智能合約漏洞，而是團隊被駭，例如 USR；還有交易對手帶著你的錢跑掉。Stream Finance 的例子是，交易員在鏈下虧了一大筆。結論是：風險研究能幫你省很多錢。風險研究的核心原則是什麼？So this year, there's been over a billion dollars lost in asset exploits. So compared to previous years, one kind of interesting thing is that even though there's been a lot of AI involved, most of the hacks and losses this year have not been, like, smart contract exploits as much as things like teams getting hacked. So that was like USR. And also counterparties running off with your money. So in the case of Stream Finance, a trader lost a bunch of money off-chain. Yeah, so basically, the moral of the story is that risk research can save you a lot of money. So what are the essential principles of risk research?

1:33今年有兩個資產，剛好示範兩大原則。一個是 Stream Finance。它是代幣化避險基金，一個叫 0xLaw 的人收下你的錢，去做進階的 degen 收益耕作。一部分策略在鏈上，你可以核驗；一部分不在鏈上，是鏈下。他把一些錢借給交易員，拿去中心化交易所交易，你核驗不了。後來那個交易員虧掉了，鏈上那塊沒事，核驗不了的鏈下那塊直接翻船。這表示資產要從頭核驗到尾，你得知道錢在哪。And I think there's two big assets this year which really demonstrate the two major principles. One is Stream Finance. So Stream Finance was this tokenized hedge fund where this guy called 0xLaw just took your money, and then he did advanced degen yield farming with it. And part of the strategy was actually on-chain, and you could verify that. But part of the strategy was not on-chain; it was off-chain. He lent some money to a trader who was doing, like, centralized exchange trading with it, so you couldn't verify it. So Stream showed that trader lost his money, and then even though the on-chain component was fine, the off-chain component that you couldn't verify just, you know, went belly up, and that showed that, you know, you have to verify assets end-to-end and make sure you know where all the money is.

2:25deUSD 也很有意思。它來自 Elixir，一開始相對單純，是跟 Ethena 類似的基差交易幣。若你在它還是基差交易幣時存進 Elixir，其實說得通：他們在 Ethena 之後推出，Ethena 已經驗證過概念，風險輪廓很單純。後來市場轉弱、資金費率下來，基差交易沒那麼好做，Elixir 轉向也借錢給 Stream，想在空頭多賺一點收益。deUSD is really interesting because deUSD is from a protocol called Elixir, and deUSD started out as a really relatively straightforward asset, a basis trading coin similar to Ethena. So if you had deposited into Elixir, in the beginning, when it was still a basis trading coin, that would have made a lot of sense because it's just like, you know, they launched after Ethena. Ethena already proved the concept, and it's just a very simple risk profile. But as the market turned down, as the funding rate came down, and, you know, basis trading wasn't good anymore, Elixir, they pivoted to also lending Stream some money to earn some extra yield in the down market.

3:10Elixir 之後：不能只做一次盡職調查

3:10Elixir 告訴我們：不能只做一次初始盡職調查。不能看資產一開始沒事，就當一切都好。你得持續盯著、持續監控。好，來看我們的第一個資產。看到這些資產被抽地毯之後，我們開始做更多鏈上核驗與風險分析。Stream 與 deUSD 出事後不久，我們開始耕一種叫 THBL 的資產。一開始看起來很漂亮，像美國短期公債的包裝。短期公債基本上是美國政府債務，大概是你能拿到的最高品質抵押。And what Elixir shows us is that you can't just do initial DD. You can't just look at the asset in the beginning and think everything's okay. You have to actually pay attention to it over time. You have to continuously monitor it. Okay, so our first asset. After we saw these asset rugs, basically, we started doing more on-chain verification and risk analysis. And one of the assets that we started farming soon after Stream and deUSD came down was this asset called THBL. And THBL is a really interesting asset because we looked at it at first and it just looked amazing. It's like this T-bill wrapper. And a T-bill is basically U.S. government debt, so it's probably like, it's like the highest-quality collateral that you can have.

4:00THBL 也跟既有的受監理實體合作：代幣化方、發行方、實際辦理贖回的，都是新加坡受監理機構。紙上看是很好的資產。當時我們在耕，是因為 Pendle 的 PT 包裝在付 12%，現在聽起來很誇張。但 2025 年那個時間點空投很熱，12% 在 TBILL 這類空投協議裡其實不算沒聽過。我們耕了，很高興，因為有加槓桿循環，大約賺 30%。一開始都好。我們監控所有農場，後來發現損益比預期差。THBL was also working with established regulated entities, so the tokenizer, the issuer, the people actually redeeming the funds were all established Singapore-regulated entities. So on paper, it's a very good asset. And at the time, we were farming it because the Pendle PT wrapper was paying 12%, which sounds insane these days. But back in 2025, at this time, airdrops were really popular, and 12% was actually not that unheard of amongst airdrop protocols like TBILL. So we farmed this, and we were super happy. We were earning like 30% because we were looping it. And everything was good at first. And then we were monitoring—so we monitor all our farms—and we noticed that the P&L was actually worse than we expected.

4:49我們查下去，盯借款利率，不是因為借款付太多。借貸市場運作正常，但掛鉤、TBILL 的價格在掉。這很怪，短期公債應該是極好的抵押。資產本身很好，也跟受監理金融機構有贖回。對聰明的參與者來說，若能九毛買進、贖回拿回一塊，就會去做，把資產套利回掛鉤。我們做鏈上分析，拿掛鉤對上它應有的淨值，發現 TBILL 沒有交易在它該在的位置。So we looked into it, and we were monitoring the borrow rates, and it wasn't because we were paying too much borrow. The lending markets were working fine, but the peg, the price of TBILL, was dropping. And that's really weird because, like, T-bills should be extremely good collateral. This is a very good asset, and this asset also had redemptions with regulated financial institutions. So for any smart participant, if you can buy an asset at like 90 cents and you can redeem it and get a dollar, you're going to do that, and you're going to arbitrage that asset back to the peg. When we did on-chain analysis and we looked at the peg versus where it should be, the NAV, we found that TBILL was not trading where it should have been.

5:36我們覺得這很妙。這張圖上，黃泡泡是贖回發生的位置，旁邊是價格。綠色是資產在掛鉤或掛鉤之上。我們又往下挖，做更多鏈上追查、問人，找到的東西都有點臭。我們去申請 KYC，因為贖回要 KYC，申請了好幾次沒回音。去 Discord，團隊對掛鉤相關問題完全不回應。很怪。我們有追到贖回，贖回確實發生了，但以這種資產掉到掛鉤以下該有的速度來看，節奏對不上。So we found that really funny. So on this chart, you can see the yellow bubbles are where redemptions are happening, and then, yeah, you have the price chart there. So green is when the asset is at peg or above. Okay. So we looked into it more. We did more on-chain digging and talking to people, and everything we found was kind of stinky. So we tried to apply for KYC. So basically, you have to KYC to redeem, so we applied for that. No answer multiple times. We went to Discord, and the team was completely unresponsive to any questions about the peg. They just— Yeah, it was really weird. And we found that we did track the redemptions. We saw that they happened, but they just didn't happen at the pace that would make sense for this kind of asset when it's under the peg.

6:28未揭露的贖回條件，後來脫鉤約 2%

6:28我們也問人、打聽，發現其實有贖回條款，只是沒公開。贖回有費用，也有能贖多少的上限。贖回資產收費用本身不怪，怪的是沒揭露。我們就退出了。後來證明是對的，因為今天它大約脫鉤 2%。TBILL 的教訓有幾條。一是：很多資產不像穩定幣，穩定幣你只要看是不是該在 1 美元；很多資產你若沒盯著，可能根本沒發現它已經脫鉤。And we, so we also, we talked to people and we asked around, and we found that there were actually redemption terms and stuff, but they weren't publicly disclosed. So there was— There were fees on redemptions, and there were also limits on how much you could redeem. And yeah, the fees are not weird on a redemption asset, but they weren't disclosed, and yeah, it was just really weird. So we just got out of that asset. And that was a good idea because today it's trading like 2% off peg. So yeah. So what are the lessons from TBILL? There's a few lessons. One is that you actually, for a lot of assets which are not like a stablecoin—a stablecoin you can just check it's supposed to be at one dollar—but for a lot of assets, unless you're paying attention, you may not realize it's off peg.

7:17你得真的去算淨值。TID Research 這類風險儀表板能幫忙。這場另一個重點是：AI 除了製造很多攻擊，也可以用來做風險研究、監控投資組合來保護自己。但不能丟一句就指望 AI 一次做完，因為風險研究對每個資產仍很專有。TBILL 是短期公債包裝，Syrup 是要去看的貸款簿，Ethena 有基差交易。每個資產不太一樣。無論你跟 AI 一起做，還是在做風險報告或儀表板，都需要懂狀況的人在迴路裡。Like, you need to actually calculate the NAV. So risk dashboards like the ones we do at TID Research can help with that. And also, so one major point of this talk is that AI, besides, you know, creating a lot of hacks, you can use it to protect your portfolio by doing risk research, by monitoring your portfolio. But you can't just ask AI to, like, you can't just one-shot that kind of thing, because risk research is still really proprietary per asset. So TBILL is a T-bill wrapper, but other assets like Syrup is a loan book that you have to look at. You know, Ethena has some basis trading. So every asset is a little bit different. Whether you're working with your AI or whether you're creating a risk report or dashboard, you actually have to have a human in the loop who understands what's going on.

8:10進場前要查，進場後還要再查，因為這些資產會變。那能用什麼？可以自己做，但非常耗時。資產風險是我們比較專長的。我是 DeFi 專家，團隊裡也有傳統金融的人，財務風險這塊還可以。資產風險可以看 TID Research。若想知道協議安不安全，我們也做一些協議分析，但我們不是智能合約審計員，會交給別人：我們覺得 Yearn 的風險評級不錯，還有一個網站叫 Philidor。Yeah. And yeah, check before you enter, check again, because these assets can change over time. Yeah, so what can you use? You can either build your own, but it's extremely time-intensive. So for asset risk, we kind of specialize in asset risk. I'm a DeFi expert. We have, like— TradFi people on our team, so our financial risk is pretty good. So you can go to TID Research for asset risk. If you want to know if a protocol is safe, we actually do some protocol analysis, but we're not smart contract auditors, so we defer to—we think that Yearn has pretty good risk ratings, and also another site called Philidor.

8:55若要存進 vault、想看 curator 好不好，可以查 Credora 或 Philidor。這是我們的風險報告頁。你可以鑽進報告看更多，也可以只看總分。很趕的話就看總分。分數是一到十，我們偏保守，所以 PYUSD、RLUSD、USDC 這類品質很高的穩定幣大約七或八。七、八相對相當安全。六也不錯。多數人我覺得可以從五開始耕。五是多數人該碰到的下限。五代表還可以，但要每週檢查。And if you want to, if you're depositing into a vault, you want to see if a curator is any good, you can check Credora or Philidor. And this is one of our—this is our risk reports page. So you can either dive into these reports and get more information, or you can just see the overall scores. So if you're, like, really in a rush, you can just look at the overall scores. It's one to ten, and we're pretty conservative, so very high-quality stablecoins like PYUSD or RLUSD, USDC, they're around seven or eight. So seven or eight is, like, relatively quite safe. Six is also pretty good. I would say for most people, you can start farming around five. Like, five is as low as most people should go. Five, it means that it's, like, pretty good, but you should check it weekly.

9:47報告每週更新，儀表板多半每小時更新

9:47真的想衝高收益，可能下到四，但資產得每天查，也要有監控。我們的報告長這樣，用六個軸打分。除了穩定性，還有流動性。依賴關係很要緊。例如看 APYUSD 這種資產，也要看背後的 sTRC，還要看 MSTR。合約管理則是存取控制。這是儀表板。報告與儀表板的差別是：報告比較像資產的整體概觀。And then if you really want to get, like, high yield, you might go down to four, but you're going to have to check your asset, like, on a daily basis, have some monitoring set up. Our reports look like this. We judge on six different axes. Besides stability, we also have liquidity. Dependencies are kind of a big deal. So, like, for an asset... If you're looking at something like APYUSD, you also need to look at the backing asset, sTRC, and also you have to look at MSTR. So we do that kind of thing. And then contract admin is, like, access control. This is our dashboard. So basically, the point of having reports and dashboards is that reports are kind of like an overall overview of an asset.

10:33報告比較靜態、也比較主觀。儀表板這邊：報告我們每週更新，儀表板的資料多半每小時更新，可以看掛鉤、儲備、流動性等等。差不多就是這些。你今天在耕的資產，未必還是你進場時的那個資產。用風險報告，用儀表板。TID Research 有不少，也可以幫忙追投資組合、監控與警示。需要的話跟我們說。就是這樣。好。嗨。謝謝 Danger。It's a little bit more static, a little bit more subjective. And then for dashboards, this is like... So reports, we update weekly, and then for dashboards, the data is mostly updated hourly, so you can check the peg, the backing, the liquidity, so on. Yeah, that's mostly it. So the asset that you farm today may not be the one you entered with, so. Use risk reports, use dashboards. We have a lot at TID Research. We can also help with, like, portfolio tracking and monitoring and alerts. So yeah, if you need any of that, just let us know. Yeah, that's it. Cool. Hi. Thank you, Danger.