Crypto has always been the perfect storm of planning fallacy. Things take ten times longer than anyone expects. In mid-2020, Loopring, zkSync (later zkSync 1.0, now zkSync Lite) and StarkEx with DeversiFi (now RhinoFi) went live in production. Optimism and Arbitrum were running testnets, the projects claimed and it seemed like 2021 would be the year of the rollups. Few remember, because few were following then - which is precisely why I started writing about rollups. It was pretty easy to follow because there were so few projects to follow. Today, I have very little idea of what’s happening in the rollup space. For example, TIL that BitDAO (backed by Bybit) is building an L2, and Consensys is building Linea - don’t even know how that’s related to their zkEVM project. OK, maybe I don’t have anything interesting to write about rollups anymore. But I do remember the past, and I believe I can attempt to provide a hopefully interesting perspective as a historian. I’m going to skip the early research days, and head straight into the products.

By the end of 2021, the situation had changed dramatically, and everyone was talking about rollups and how 2022 would finally be the year (remember #L222?). I was only cautiously optimistic and was looking towards late 2023 or early 2024 instead. Inevitably, the planning fallacy grim reaper came for smart contract rollups and even my then-pessimistic expectations have turned out to be optimistic.

App-specific rollups have done incredibly well, though, and were in a good state already by 2021. Many of these projects have clear solutions for sequencer, state validation and validator failures. Some have timelocks and security councils to address contract upgradability, while one is immutable. Only now is Arbitrum catching up to these standards, while the other smart contract rollups are probably another year or two away.

Of course, the meat of the demand was for smart contract rollups - not app-specific rollups. In late 2020 and early 2021, the situation had reached a boiling point with L1 gas fees, and combined with extreme shilling from alt-L1s, led to a lot of pressure on rollup teams to deliver early. This caused Optimism, zkSync and to a lesser extent Arbitrum to rush MVPs to production.

In 2020, there was a pre-eminent leader in the rollup world - Optimism. Yes, ZKRs came first with application-specific rollups, but everyone wanted smart contract rollups, and it seemed pretty obvious Optimism would be first, with the support of big projects like Uniswap and Synthetix behind it. A common thesis was about how Optimism would be the defacto standard Ethereum rollup, where a majority of the activity would be.

To be blunt, Optimism blew it, is today at least a year behind Arbitrum (IMO) and incredibly, zkEVMs have launched in production with provers before Optimism.

In 2019, zkEVMs seemed a pipe-dream and at least 5-10 years away. This is the one area that has exceeded all expectations, and today, we have two zkEVM projects live in production.

Back to 2021, the ZKR game was all about Starknet and zkSync. The thought was it’s so difficult to zk-fy the EVM that it would be much better to start from scratch. Starknet went the Cairo route, and zkSync chose a middle-ground with making their VM Solidity-compatible. Turns out, bootstrapping a new VM and building execution clients is far harder than making an already proven VM validity provable.

I had anticipated 2022 would be the year of decentralization, and then 2023, 24 onwards the focus would shift to performance. However, it turned out 2022 was the year of performance and getting the base architecture right. Optimism, Arbitrum, Starknet and zkSync alike all significantly overhauled their architectures, with most/all of them effectively starting from scratch. With the bear market and the pressure for scalability receding, this may have been the right approach.

Both Optimism and Arbitrum went back to the drawing board, and returned with a similar design, reusing Geth and other battle-tested execution clients. This has allowed Type 2 zkEVM designs like Scroll and Taiko to catch up, similarly reusing Geth as sequencer clients for the most part. Sure, their circuits and proof systems might be more complex than, say, Starknet (I don’t know) but evidently this complexity is less than building and maintaining execution clients.

With Arbitrum Nitro, Polygon zkEVM and zkSync Era all live, and Cairo 1.0 on Starknet and Optimism Bedrock releasing very soon, and the likes of Scroll later this year - 2022 is indeed finally the year of getting the base architectures for rollups worked out.

Today, there’s a clear leader in the rollup space - Arbitrum One. Even before the airdrop, and the dust after it is settling, it’s pretty clear Arbitrum One is in the same ballpark as BSC for economic activity. BSC has been the dominant alt-L1 for so long, it’s quite the achievement. Yet, Arbitrum One has not earned this, and is only now getting to the point where I’d say it has achieved minimum viable decentralization. It does go to show how incredibly valuable building a chain with a secure, validating bridge on Ethereum is - you get immediate access to $430 billion (at this time) in assets and liquidity. The good news is, as per L2Beat’s Risk Rosette, there are no more red signals on Arbitrum One, and the only two remaining yellow ones also come with honest-minority assumptions. I hope to see both of these improve through 2023, and at that point, I’ll finally be ready to recommend people use a smart contract rollup.

Optimism has made very strong moves culturally, however their technology has fallen well behind their peers. With the much-delayed Bedrock finally on track, it’ll be a big step forward. Optimism developers will tell you that with this base in place, the rest of the upgrades will be smooth sailing, but I’ll believe it when I see it. Arbitrum Nitro went live in Aug 2022, and when the Cannon upgrade comes - after Bedrock - they would have finally caught up to Nitro. This is why I believe Optimism is at least a year behind Arbitrum. Yet, culturally, Arbitrum is just as far behind Optimism. Their token distribution is even more atrocious than Optimism, and their transparency and communication has been lacking. They are well behind on licensing and building with the community - which I believe is the main reason Coinbase chose Optimism’s tech for their Base rollup.

For the longest time, Starkware was on a roll, and right up to the release of Starknet Alpha in November 2021, they seemed like they could do no wrong. Unfortunately, Cairo 0.x never really gained adoption, their sequencer was very slow, and they have spent nearly 18 months now rebuilding their stack with Cairo 1.0, new Rust-based sequencers etc. Meanwhile, since the focus was on Starknet, StarkEx has also failed to evolve. zkSync has also seen many challenges - they released the first alpha testnet for their zkEVM project way back in May 2021. At the time, it seemed like an MVP was happening in 2021. Instead, they have taken nearly 2 years to deliver on that promise.

Outside of Arbitrum, the decentralization landscape is quite poor across Optimism, Starknet, zkSync Era, Polygon zkEVM etc. It’s clear my “early 2024” hope won’t happen, and I’m instead looking at late 2024 or even 2025 for some of these projects, and upcoming ones like Scroll or Taiko to decentralize and mature. It could be till 2026 or 2027 till the zk circuits are fully battle-tested and as proven as Ethereum L1 itself. Indeed, at the forefront of rollup decentralization lies Tezos - who recently released their take on enshrined rollups. I believe Ethereum core devs should see this as inspiration and work on upgrading L1 to zkEVM - but this is a whole another topic!

Then there are projects building novel solutions for the smart contract space like Aztec and Fuel. Aztec, in particular, is extremely exciting - building fully private smart contracts, but this is of course a very, very challenging task that may be several years away from maturity.

I don’t care for crypto tribalism, and I avoided crypto twitter for the longest time to evade this. 99% of L1s in crypto are ghost chains. Some of the early ones from 2013-15 are now starting to buckle under the weight of economic unsustainability, which I’ve written about at length. It seemed too obvious that many of these L1s would pivot to being L2s and benefit from the network effects, economic security & sustainability, and liquidity of Ethereum. I was very wrong about this, ego, cultism and tribalism run so hot in crypto that they would rather remain in irrelevance than build a pragmatic solution. Indeed, I’m only aware of one L1 project that has made this transition - Mir, now Polygon Zero.

Fortunately, if the L1 cults do not do it, someone else will. Again, forgive me for not being up to date nowadays, but I believe there are projects forking all sorts of different VMs - including stuff like Solana VM being built as a rollup by Eclipse. Other L1s with a decent social layer - BSC, Cardano, Polkadot, Tezos, Cosmos etc. - are also building rollup or rollup-related solutions, and even anti-rollup ones like Solana are exploring adjacent fraud proof solutions. Of course, in most cases, they would be better off collaborating rather than competing with Ethereum - it’s a win-win-win for everyone - but like I said, tribalism runs hot in crypto, and it’s still a step forward over unsustainable monolithic tech. Special mention to Polygon - I wrote many critiques about their PoS chain, and they have responded in the best way possible, by investing heavily on rollups.

Another thing I greatly underestimated was data availability sampling. Turns out this is an extremely difficult engineering problem, and is going to take many years to build an uncompromised solution.

On the flipside, I also overestimated what the application space innovation looks like. In 2021, applications were still being built, but there was a good understanding it would be mostly remixes and evolution of the 2017-20 class. The ICO mania funded pretty much every blockchain app possible, and as expected, 99.9% made no sense and went nowhere. Nevertheless, the 0.1% made for some really great apps, mostly built in 2018 and 2019. Still, I kept an open mind, and looked forward to what would be built in the next bear market. As we head into 2023, it’s now very clear to me that the great wave of app layer innovation happened in 2017-20, and most things that made sense for public blockchains have already been built or conceived. To the 6 of you who are mad enough to have made it all the way to here, I’d recommend aborting on this rant and instead go do something better - like watching Death by Hanging (the 1968 movie directed by Nagisa Ōshima) or tuning in to Electric Bath by The Don Ellis Orchestra. But if you are still mad enough to continue… shout at me on Twitter or Reddit! Anyway, so, happily, my underestimation for data availability sampling and overestimation for app layer innovation cancel out. More limited data availability layers like EIP-4844 and expanded 4844, EigenDA, Celestia, Avail, zkPorter will suffice for the foreseeable future (I’ve made multiple posts arguing why - this one’s the latest), while uncompromised data availability sampling is figured out.

As some of you may know, the polynya blog (and later twitter profile) is carefully crafted. For a year, I wrote about rollups in a neutral tone, but few cared. So, I started a blog called polynya and went on the offensive with bombastic statements and the like with a fair degree of creative liberty. In May 2021, I wrote a Reddit post about how “Arbitrum is the best smart contract chain since Ethereum in 2015” at a time when ADA was worth $100B and the prevalent narrative was “third-generation L1s will be as big as Ethereum, if not bigger”. Today, it’s evidently true - Arbitrum One is the #2 chain after Ethereum by multiple economic activity metrics, as mentioned above, and most of the largest non-Ethereum chains (with few exceptions) are heavily researching fraud/validity proof tech. It was all a show, in reality I had very little confidence, and I’ve suffered from strong imposter syndrome and gaslighting throughout my time on crypto twitter. It could have gone either way - we just got lucky the technically superior and long-term sustainable solutions are seeing greater meaningful adoption over the shiny new things. Of course, it could just as well be Ethereum is the largest cult in a bear market, which Arbitrum One has leveraged to its advantage, and . I hope the crypto space will learn to value sustainable economic and technical models, fair governance schemes, useful applications and UX, over shiny new things, memes, cults and plutocracies going forward. One can hope. (Lol who am I kidding)