How would you go “Long DAI, Short Tether” in a DeFi way, since no DAI-hards have any Tether to lend you in a DeFi lending pool to sell short? Here are 3 ideas.


One of the first DeFi applications has been secured lending and borrowing. Dharma, Compound, and Maker are the most well-known examples in the space, allowing peer-to-peer lending of assets via lending and borrowing pools (Dharma and Compound) and  “peer-to-contract” lending (Maker).

Maker has used the overcollateralized lending model to create a crypto-backed stablecoin, DAI, such that DAIUSD = 1. DAI is created when users deposit ETH into a “CDP” and take out a loan (denominated in USD) against the USD value of the ETH. The loan’s accounting is provided by the use of DAI as a stablecoin.

It can be argued that DAI is the “ideologically purest” stablecoin at this time, in that it’s managed in a fairly decentralized way (there are some notable caveats, namely in the centralization of the oracles used to determine ETHUSD prices, the incentives for holders of the governance token, etc.) As a result, DeFi and DAI evangelists have integrated DAI into many other projects and use cases, enabling more users to get on-ramped directly from fiat into DAI.

On the other hand, Tether (USDT) was supposed to be a fiat-backed stablecoin backed by USD reserves held in a bank account. This turned out to not be entirely true, and was pointed to as an example of what can go wrong with centralization.

Traditional Shorting

Given the DeFi community’s love of DAI and skepticism of fiat-backed stablecoins, USDT would seem like an asset that lots of DeFi and cryptonative evangelists would like to short (i.e. make money if the price of USDT goes down).

Typically, the way to short an asset is to borrow it from someone else, sell it in the market at price PS, buy it back later at a lower price PB < PS, and then give it back to the original lender of the asset.

But that cultural wholesale rejection of USDT seems to also mean that DeFi community members who are strong advocates of DAI don’t want to hold USDT, not even to supply it to short in DeFi borrowing pools. It doesn’t appear in Compound or Dharma, probably because no one who is in the DeFi space wants to have USDT/USD price exposure, even if they could earn some yield by lending it out in a pool.


Instead, here are some derivatives that might enable this risk exposure without requiring people to buy and then loan out their USDT. Let’s assume Alice wants to be long DAI/short USDT and Bob wants to be long USDT/short DAI. But remember – Alice wants DAI and doesn’t want to hold USDT, and Bob wants USDT and doesn’t want to hold DAI.

  1. “Jimmy-Joe” swap
    • Jimmy Song, BTC maximalist, and Joe Lubin, co-founder of Ethereum, have a bet regarding the state of the Ethereum dapp ecosystem in 2023. Fuzzing the conditions of the payout for a moment, the payout stipulates that if Jimmy wins (there “aren’t a lot of dapps”), Joe gives Jimmy 69.74 BTC. If Joe wins, (“there are a lot of dapps”), Jimmy gives Joe 810.8 ETH.
    • In USDT/DAI terms, we can structure a swap so that if DAI/USDT goes up, Bob pays Alice DAI and if DAI/USDT goes down, Alice pays Bob USDT. They have to agree on a strike in advance (say, DAI/USDT = 1).
    • Pros:
      • Alice and Bob each get more of the currency they want if DAI/USDT goes in their favor.
      • This could totally be built on something like the state channels of the Rainbow Network!
    • Cons:
      • How do Alice and Bob find each other?
      • Who decides how much margin Alice and Bob have to put up?

2. “Bitmex-style BTC/USD in BTC” swap

  • Bitmex offers BTC/USD futures margined in BTC. Each contract is worth $1 of exposure to BTC, so someone who wants to go long $100 worth of BTC/USD buys 100 contracts.
  • We could create a similar product for DAI/USDT – someone going long DAI/USDT would buy 100 of these contracts, getting 100 USDT exposure to DAI/USDT. Any PnL is then paid out in DAI terms.
  • One of the problems with this product is that it has convexity. This means that if DAI/USDT goes up from say, 1 → 2, then Alice should get 100 USDT (assume they bought 100 contracts). However, because DAI/USDT is now 2, the 100 USDT margined back into DAI is only worth 50 DAI. If the notional had been in DAI, she would have had 100 DAI instead. This is all because the notional here is denominated in USDT, not DAI, resulting in convexity.
  • Pros:
    • This is an established product type that Bitmex users might be familiar with. Bitmex transparently shows users that there is convexity in these products.
  • Cons:
    • Alice would denominate their exposure to DAI/USDT in USDT terms – a very unnatural unit of account for her.
If DAI/USDT goes up from 1 to 2, then Alice gets 100 USDT, which at a DAI/USDT price of 2, equals 50 DAI.

3. “Bitmex-style ETH/USD in BTC” swap

  • As an alternative to the earlier “Bitmex-style BTC/USD in BTC” swap, we can offer a DAI/USDT swap where the PnL is paid out in a totally neutral, “third-party” currency, like USDC. Everyone’s exposure is then denominated in USDC, which everyone should feel equally “meh” about. This is like Bitmex’s existing ETH/USD swaps, which settle in BTC. Someone who wants to be long or short ETH/USD shouldn’t care about the BTC/USD price, so long as they receive enough USD-equivalent BTC to match their USD PnL from the ETH/USD exposure.
  • For example, let’s say Alice buys 100 USDC worth of this DAI/USDT swap from Bob. If DAI/USDT goes from 1 to 2, she’ll get 100 USDC. If DAI/USDT goes from 1 to 0.5, she owes Bob 50 USDC.
  • Pros:
    • Easy to understand. Bitmex users are probably already familiar with this product type.
  • Cons:
    • How do Alice and Bob find each other? There is also some “quanto” risk that Bitmex has also previously blogged about.
Alice and Bob should both be equally “meh” about being paid out in USDC.

Some rankings

Here’s a ranking of each of the 3 products for different criteria. The “Jimmy-Joe” swap is the coolest and gets the closest to what DAI-hards probably want – and shouldn’t be much harder to create than the others!

  • Liquidity: How do Alice and Bob find each other?
    • The “Bitmex-style quanto swap” (3) has standardized trade terms, and Alice is fully funded, so Alice and Bob can list their orders on an exchange and connect fairly easily (we just have to set margin terms for Bob). (1) might have bespoke margin requirements for both Alice and Bob, that they might want to negotiate. (2) has nonlinear risk – which makes it hard to pitch to Alice and Bob.
  • Closest to reality: Which can be built?
    • This is a 3-way tie. All 3 could probably be implemented with state channels (like Rainbow Network!) – we’d just switch up the price feed that determines margin movements, and use a different margin currency, depending on which one to implement.
  • Needs additional players: Do you need market makers/hedgers?
    • Products with weirder risk (like the convexity and quanto features in (2) and (3)) are harder to hedge for market makers, so they’re less likely to step in when markets are imbalanced. (1) probably would feel more like a Veil/Augur market, in that Alice and Bob would have to find each other and offset exactly.
  • Reflects DAI-hard view: Do they really give better exposure?
    • The Jimmy-Joe swap in (1) is the most explicit “Long DAI/Short Tether” view. Each party is getting more of the coin they want when the price feed moves in their favor. (3) and (2) try to get there, but fall short.