Thread by @mjayceee

All indications are that the Uniswap team is working on a V4.

In this thread, I’m going to discuss what Uniswap V4 might look like and why.

1/🧵

This thread is 100% speculation. I don’t have any inside information.

2/🧵

As a bit of background, the Uniswap V3 “business source license” expires in April, two years after the V3 release.

To deter another Sushiswap-style vampire attack, the BSL explicitly forbids forks of the V3 code for a period of two years.

3/🧵

With the BSL expiring soon, you could argue that Uniswap is under pressure to direct a path forward.

In just a few months, we’re bound to see a parade of low effort UV3 forks - and maybe a few good ones too - chipping away at Uniswap’s volume dominance and revenue profile.

4/🧵

I would not be someone making this argument tho. DeFi has moved past the stage where incremental innovation moves the needle with users or investors.

In fact, projects have been soft-forking UV3 and its most popular feature -concentrated liquidity- essentially since launch

5/🧵

Concentrated liquidity is the focal point of newer DEX competitors like @crocswap.

And according to a recent Twitter search, even projects on BSC, Cardano and Solana now consider CL to be table stakes for a competitive DEX.

6/🧵

The motivation for UV4 is not the impending BSL expiration. It’s really much more straightforward.

In all its dominance, Uniswap V3 still has a lot of holes in it.

7/🧵

This framing is a bit simplistic, but I see three main areas of concern:
- LPs Rekt
- Traders Rekt
- UX Limited

8/🧵

#1: LPs Rekt

Every few months, a couple of CT heavyweights reinvigorate the classic debate: are Uniswap LPs profitable?

Since UV3 launch nearly two years ago, the discourse has seesawed quite a bit.

9/🧵

For a while, UV3 liquidity managers were advertising atmospheric APYs on concentrated liquidity positions, with complete disregard for impermanent loss risk.

10/🧵

At @xtokenterminal, we were the largest manager of UV3 liquidity for about 6 months after launch, peaking at $25 million AUM.

But we only offered like-for-like vaults (e.g. USDC/DAI, WETH/sETH) before renouncing the active management space altogether.

11/🧵

The reason is simple: providing unincentivized ambient liquidity on volatile UV3 pairs – absent a larger strategy – is simply a bad idea. It doesn’t matter how frequently the position is rebalanced.

12/🧵

[NB: if you’re not pulling and replacing orders many times per day like a traditional market maker, you’re providing “ambient” liquidity]

13/🧵

You may get lucky, but chances are you will lose money. The expected return in any case is not worth the risk or opportunity cost.

14/🧵

This analysis from @thiccythot_ is the most authoritative and conclusive on the matter. Impermanent loss is very real and very potent.

https://twitter.com/thiccythot_/status/1589022229983293440

15/🧵

And this analysis by @0xfbifemboy confirms that rebalance frequency doesn’t improve matters.

Providers of ambient liquidity get burned.

crocswap.medium.com/benchmarking-t…

16/🧵

Even this supposedly bullish analysis shows ETH-USDC LPs registering **checks notes** a 30k ETH aggregate loss compared to just holding the base assets (h/t @0x94305)

https://twitter.com/charliemktplace/status/1606380020213104640?s=20

17/🧵

Anyway, it’s been fascinating to watch the conventional wisdom on liquidity provision evolve over Uniswap’s history. I made this meme

18/🧵

So, what to do?

I see two paths.

19/🧵

The first option is to abandon the fabled XYK model altogether. It’s gotten us this far, but it may have reached the limit of its usefulness.

I think there is a non-zero chance Uniswap leaves the XYK model behind completely.

20/🧵

But ultimately, I don’t think they will.

21/🧵

XYK works especially well for long tail, volatile pools, i.e., most pools.

22/🧵

There’s a notion that bootstrapping this liquidity via incentivization is unsustainable, but #actually paying LPs, even in perpetuity, for the service of making a market is totally normal and makes obvious sense.

This is a prime XYK use case.

23/🧵

The second potential path for UV4 relies on a basic insight about the XYK model:

By definition, LPs accrue impermanent loss during times of volatility. But currently they are not compensated for that additional market-making risk.

24/🧵

The solution: dynamic trading fees.

If ambient LPs are ever gonna have a shot, trading fees need to be higher during times of volatility. It’s not a panacea, and LPs will still lose money in many cases, but it at least levels the playing field.

25/🧵

This is not a new concept. @kybernetwork has implemented dynamic fees in their latest generation AMM, although they since seem to have de-emphasized it.

26/🧵

https://twitter.com/KyberNetwork/status/1379067393025077251?s=20

Both @traderjoe_xyz and @synthetix_io have implemented variations of volatility-based fees into their AMMs as well.

27/🧵

Interestingly enough, the Uniswap team and investors have been examining this idea since at least 2019, well before UV2 even launched.

Uniswap VC @_charlienoyes discussed the possibility of dynamic fees at a 2019 conference

28/🧵
drive.google.com/file/d/1r8YaSw…

In two concise bullet points, he zeroed in on what an AMM should aspire to.

29/🧵

The dynamic trading fee model could come with another benefit as well.

If the current swap fee is derived from some volatility measurement, there only needs to be one pool per token pair, de-fragmenting liquidity in a nice callback to UV1/UV2 design.

30/🧵

Building an effective internal volatility oracle will be no easy feat, but Uniswap certainly has the resources and track record to pull it off.

31/🧵

Because an on-chain system like this has never been operated at this scale, it may require some fine-tuning post-deployment.

32/🧵

Although it’s inconsistent with Uniswap’s DNA to allow upgradeability to the protocol, it may be necessary to design some modularity or configurability into the contracts. Modeling a system like this without real world feedback would be challenging indeed.

33/🧵

#2: Traders Rekt

According to a recent analysis by @delitzer, Uniswap emitted roughly one fuckton of MEV over the course of the last year.

There is no protocol that leaks more to MEV searchers than Uniswap.

34/🧵
medium.com/nascent-xyz/th…

This is really bad for traders. MEV bots are competing to see who can extract as much value from Uniswap users as possible.

35/🧵

But, by drawing some insight from another project in the space, UV4 has an opportunity to flip this dynamic upside down.

That project is @cowswap, and my feelings on it are summarized here:

https://twitter.com/mjayceee/status/1587918168626192384

36/🧵

<<<Disclosure: I hold $150 of mostly unvested $COW, tucked away for safekeeping on Gnosis Chain, where I’ll forget about it and never be tempted to sell>>>

37/🧵

Cowswap is not a DEX in the traditional sense. The Cowswap contracts hold zero liquidity and abide by no pricing formula.

38/🧵

Users sign orders with trade params (input amount, min output amount, etc.) and an army of “solvers” compete against one another to deliver the best fill, BOTH in terms of swap efficiency and gas efficiency. More on the gas part later in the thread.

39/🧵

The solvers source liquidity from any on-chain source, competing to propose the most optimal combination. So far, this is not *that* unlike any other aggregator.

40/🧵

Cowswap has brought something new and very innovative to the table, however. The protocol’s “coincidence of wants” (C.O.W.) model enables solvers to match liquidity on opposite sides of a trade within a batch period.

41/🧵

For blue chip pairs, this is incredibly powerful. There are many ETH/USDC trades in each direction during a batch auction period (~a few min). COW generates massive efficiency b/c this liquidity never needs to hit a conventional AMM, bypassing slippage costs and swap fees.

42/🧵

Unless a user needs instantaneous execution, the batch auction model is clearly superior. A protocol with Uni’s market share could actually run these batches much more frequently than Cow. The difference in confirmation speed compared to a regular trade would be minimal.

43/🧵

If you squint just a bit, a batch auction could be described as a bidirectional dutch auction, maximizing value for both sides.

44/🧵

Uniswap VC @danrobinson, known to have taken an active role in UV3 design, has publicly expressed a high degree of interest in dutch auctions on multiple occasions.

45/🧵

This man may very well be tweeting out the UV4 roadmap right before our eyes.

46/🧵

If you need any further confirmation that the Cowswap model is the way of the future, after I started writing this thread, the market leader in DEX aggregation @1inch, pushed an upgrade that looks an awful lot like Cowswap

47/🧵

https://twitter.com/1inch/status/1607015053307379712?s=20

Good ideas spread fast in DeFi. Uniswap has experienced this on both sides of the coin.

48/🧵

[NB: To be clear, Cowswap still has some hard work to do on the decentralization front, a would-be tall order for the Uniswap team, assuming that they don’t compromise on longstanding values]

49/🧵

#3: UX Limited

This section is not so much about UX weaknesses of UV3 as much as weaknesses of Ethereum that Uniswap, as the most used app in the space, will solve for in their bid for world domination.

50/🧵

And by “world domination,” I really just mean vertical integration.

In a sense, Uniswap’s acquisition of NFT aggregator @geniexyz tipped their hand. They want to own the aggregation layer that will soon tap into the mobile app that I can only assume is coming.

51/🧵

The Cowswap model of aggregation is especially appealing because of the UX interface it provides. The normies are simply not going to toggle gas settings when submitting a transaction. Cowswap abstracts this complexity, as the solvers bake in gas costs into trade output.

52/🧵

Stepping back a bit, V4 in the Cowswap model would be a paradigm shift for Uniswap, moving from a DEX to more of a decentralized execution engine. There would still be a DEX underneath the hood, but the complete stack figures to be much more intricate.

53/🧵

And with the primary user interaction being a signature instead of a direct transaction, the potential design space is a lot richer.

You could imagine UV4 including more complex transaction types like limit orders (of any duration) built natively to the protocol.

54/🧵

As for other UX considerations, the Uniswap team has been conspicuously shepherding EIP1153, pushing for its inclusion in an upcoming Ethereum protocol upgrade, so far unsuccessfully.

https://twitter.com/saraareynolds/status/1601345467736281088?s=20

55/🧵

I don’t have the answers for exactly how Uniswap intends to use transient storage to improve UV4 UX, but I bet that a better developer than me can piece it together from the numerous talks @sendmoodz has done on the subject

https://twitter.com/sendmoodz/status/1551451868521721856?s=20

56/🧵

Wrapping Up

Within this architecture, there are several ways that Uniswap could introduce a productive use case for the UNI token – one that doesn’t run downstream from trading fees, but instead one that governs the solver architecture or administers the volatility oracle

57/🧵

This is a subject for another day, but a non-fee-switch token model would be a welcome sight for the Uni+VC legal team.

58/🧵

There will come a day when the community revolts against a legal team that runs fee switch interference on governance forums and contorts themselves into increasingly conservative positioning.

59/🧵

For these last few tweets, I want to mention some things that I *don’t* think will be in V4, at least in the protocol layer.

60/🧵

Cross-chain logic:

There are too many security assumptions that come with cross-chain operations. I don’t see it. One chain, one instance.

61/🧵

KYC:

There certainly might be KYC apps built on top of UV4, but I don’t think Uniswap will corrupt the base protocol layer with non-native logic like KYC.

62/🧵

That’s it! Thanks for reading

63/🧵

P.S. h/t to @hosseeb, who I think predicted some of these ideas and developments in a blog post all the way back in August 2020.

medium.com/dragonfly-rese…

64/🧵