Conversation

Replying to

4) Well, before you can do _anything_, you have to trust the blockchain you're using: trust that its miners, or stakers, or nodes, will try to build the longest chain without ulterior motives. And I'll leave discussion of different blockchains to another day.

1

1

14

5) With that out of the way, the goal is: a) don't trust anyone, assume everyone is selfish and incompetent b) build great products that provide lots of value Let's first start with something that does pretty well: atomic swaps.

2

3

25

6) There are some details here (buying short-term options by reneging), but ignoring that you can swap 2 ERC20 tokens for each other in a way that's trustless and efficient. How? Well, you use smart contracts. The central smart contract can check that both sides have funds.

1

15

7) Both sides put the funds in the smart contract, and once it has both and both parties have agreed to a trade, it sends each person what they bought. You can do this via pre-programmed smart contracts with no middlemen who can fuck with things. And it does what users want.

1

17

8) What else works well in DeFi? Borrow/lending mostly works as desired; see Compound, Aave, etc. But venture far beyond that and the product quality quickly degrades. For example, take stablecoins.

1

17

9) Stablecoins are basically the simplest product in CeFi. You take an ERC20 token, and you take dollars in a bank account, and you map them 1:1, with creations and redemptions. There's basically nothing tricky about them (other than holding the bank account!). How about DeFi?

1

17

10) In DeFi, stablecoins are a total fucking mess. The problem, really, is that a stablecoin is a dollar, and dollars are not on the blockchain; they're centralized in bank accounts. So if you only have the blockchain, you can't redeem for $1. But then what makes it stable?

2

4

23

11) All over DeFi you see complicated attempts to solve this. Curve and mStable are entire DEXes built around stablecoin:stablecoin swaps. OTOH you can trade USDC:TUSD 1:1, no fees, infinite size using ftx.com/wallet, or you can do it yourself if you have banking.

Cryptocurrency Derivatives Exchange

1

18

12) But those require banks, and if you don't allow centralized elements then you're left with, basically, atomic swaps of TUSD for USDC--which depend on some outside mechanism to keep them in line and provide liquidity.

1

1

13

13) But it gets worse--if you really hate centralization, you might not like the fact that your stablecoin's value relies on a bank account. What if it gets frozen, or the creation/redemption facility shuts down? What is worth $1 but doesn't rely on banking?

1

18

14) There's a really fundamental tension here! So we end up with stuff like $DAI. And DAI is a total mess. It's not backed by dollars. That means you can't redeem it for a dollar bill. But in fact, you usually can't redeem it at all! What makes it "worth" $1?

2:53 AM · Jul 21, 2020Twitter Web App

Replying to

15) In fact you can't create it either, really--you can just borrow it. So it's peg is really soft. And it's backed by a leveraged position, so if markets move the stablecoin's holdings might get liquidated. So to recap: no creations, no redemptions, but yes liquidations.

2

2

24

16) Doesn't really sound so stable! (For a more in depth discussion of the dangers of DAI, see here: medium.com/@ministry_of_a)

A Little Unsteady

I think it would be fair to say that, over the course of crypto, three classes of coins have served as the backbone of the industry: BTC…

2

3

29

17) So you don't want to trust USDC's centralized redemptions, but you do want to be able to turn your stablecoin into $1. What are you left with? Well, you have mStable, and Curve, and lots of other attempts to build out a hybrid product.

1

2

19

18) And stablecoins are just one example. The tradeoff between centralization and usability is all over the place. Many DeFi projects need price feeds; so they use oracles, which.... usually draw from Coinbase (or Binance, etc.). Which is fine--that's what FTX futures do!

1

17

19) But it's not super decentralized; it's super prone to centralized failure or API issues. (See e.g. BitMEX and Deribit futures when Bitstamp etc. blew out.) And if you want to trade BTC on Ethereum, most attemps ultimately have a pack of judges custodying the BTC.

2

19

20) This even pops up when looking at blockchains. One way to make your chain faster and cheaper is to make it centralized--if it's just you saying what happens you can do so quickly. But of course that starts to cut into the whole point of crypto, and DeFi in particular.

2

14

21) So what do we do here? Many of the answers aren't obvious. But that doesn't mean it's hopeless. There are powerful procedures to create trustless cross-chain behavior; there are blockchains that gain speed without sacrificing security.

1

17

22) And there are ways to approach stablecoins that, while not perfect, at least do something better than DAI, and less centralized than USDC. In order for DeFi to grow, it's going to have to make a compelling case that the trustlessness justifies the hassle.

4

1

25

23) And that means finding pareto optima, making the best tradeoffs you can, and continuing to find ways to build out the power of the DeFi ecosystem without sacrificing its core principles.

4

1

35