Web3 is now too large and the stakes are too high to launch on the Ethereum mainnet. The Ethereum roadmap pivoted in 2020 to focus on supporting L2 protocols as the primary scaling strategy. Optimistic Rollups are a type of L2, and are production-ready today. Web3 developers should build their applications on one of these rollups, such as Optimism or Arbitrum.

Choosing a specific Optimistic Rollup is not a difficult decision, because users will be able to bridge across all L2s cheaply and securely, and these L2s are likely to converge on similar code. The difference between L2s will be like the difference between Mastercard and Visa – as far as most consumers are concerned, there is no real difference.

L2 networks are starting to incentivize adoption with governance token airdrops, and supporting applications and infrastructure are compounding strongly. This post explains why devs should start building on an Optimistic Rollup today.

Over $100 million was spent on gas for the YugaLab’s Otherdeed mint yesterday, with only 27,500 users participating. The $APE community assumed that members would wait for gas prices to stabilize during minting, but the fact that they did not highlights the problem with L1 – that although there are efficiency improvements that can be made to support more transactions (like different auction types, or code optimizations), what makes L1 untenable is the fundamental limit of concurrent users. Many users want to participate in drops at the same time, each applying independent balance changes – so the only way out of this concurrency conundrum is to batch transactions, which is what L2s do.

Two L2 protocols have proven to be production-ready, with high-quality applications in use: Arbitrum and Optimism. NFT marketplaces such as Quixotic on Optimism and Treasure.lol on Arbitrum show these systems are working.

On Mirror, we recently let anyone mint an NFT of the Optimism governance token announcement – but only to mint on Optimism itself. 520k tokens were minted in one day. The take-away was that Optimism handled a half-million token drop easily, and it was the “web2” parts of applications like Mirror, Quixotic, and Dune that struggled to keep up with the volume. These applications are now working to improve their infrastructure from this feedback, and will become more resilient for the next drop. We will likely see flawless, million-token NFT mints on Optimism in the near future.

Arbitrum is also starting to have robust applications built on that network.

The big decision is to move from L1 to L2. Once this decision has been made, picking an L2 isn’t difficult.

Firstly, L2s can be abstracted, because they theoretically share the security of the L1, and it will be extremely cheap for users to bridge across different L2s. End users will care less about their L2 network for minting than they can about whether they’re using Mastercard or Visa to shop on Amazon. Only bridging from L1 to L2 will be expensive, but in the future users will typically never do this – they will be able to buy L2 tokens directly on exchanges. So bridging from L1 is a short-term problem to solve, and we shouldn’t blow it out of proportion when considering moving to L2s.

Secondly, and more technically, what makes any Optimistic Rollup a good block-producer is very similar to what makes the base-layer good. So Optimism, Arbitrum, Polygon, and other L2s might simply converge to the same code over time – a tight adaptation of the L1 code. This is exactly what Optimism did by forking the most popular Ethereum client, Geth. This approach makes sense to me, and I think it is a good pattern for new Optimistic Rollups. So there probably won’t be a substantial technical difference between L2s; assuming the code is bug-free, it’s fine to choose either. I like that Optimism is very similar to Geth.

The main reasons to stick with Ethereum is that it will continue to be a robust and stable network to build on, it already has a wide distribution of usage, and bridging to L2 for cheaper transactions will prove to be more secure than bridging to other L1s. I expect the majority of quality web3 development to continue to be in Solidity, and for many existing Ethereum users to to bridge to the L2s. The more this happens without a major incident, the more the network effect will grow and development and investment will favor L2.

In the long-term, L1→L1 bridges might become perfectly secure, but today they are not proven. The recent Wormhole Solana-Ethereum bridge exploit for $250 million is a hint of the scale and consequence of insecure bridging.

Like many others, I am obsessed with the promise of zero-knowledge across all domains of technology this century. I believe ZK-proofs are the key to scalability and privacy in the long-term. But I still think that the Optimistic Rollups are the best choice today for development.

Firstly, ORs are fully production-ready, while ZK-rollups are in an earlier stage of development. Programing in Cairo is difficult, and only an acceptable burden for some stellar technical teams (like dYdX), but it is not approachable for most other product-focused teams that want to move quickly. To make this concrete: consider that Cairo doesn’t support for-loops, and so the contract developer must use recursion.

Secondly, (to say it again) bridging across different L2s will be cheap and easily abstracted in the user-experience. The bridging and migration path across L2s (even OR to ZK) will not be not as burdensome as the migration off of L1 in the first place.

Ethereum is becoming a base layer for L2 solutions – serving as a common root of trust to move assets between different rollups safely and cheaply.

This was reflected in a new roadmap, introduced in October 2020 – a week after Optimism announced their testnet launch and proved that rollup ecosystems were developing.

Cross-L2 transfers will be cheap and seamless for the consumer, and prototypes are developing that integrate existing L1 state with new L2 data, such as ENS names. This will support a user-friendly experience when interoperating across L2s, to the point where users will not care about the specific L2 network for a given application. This removes the stakes when deciding on an L2 network. Since Optimistic Rollups are production ready and easy to develop on, and Optimism’s code is very close to the base layer’s code, it’s viable to choose Optimism to build on today.

L2s are also going to launch their own tokens, which will incentivize adoption and funding. Altogether, this makes L2s very attractive for development.