This post is taking part in the Lens x Kiwi Writing Contest

------

Right, L2 is NOT Ethereum

The illusive elephant in the room.

Is L2 a part of Ethereum?

Here's a totally different questions.

Is AWS a part of Internet?

Yes, AWS counts for 30% of global traffic.

Is AT&T a part of Internet?

Yes, AT&T has 50% of undersea cables.

Is Amazon a part of AT&T? NO.

Does Amazon rely on AT&T? YES.

DEPENDENCE =/= ALLIANCE

No, L2 is not a part of Ethereum.

Totally separate network, separate database.

L2s have the right to choose to settle to other L1.

Is L2 parasitic to Ethereum? Yes?

Activity stagnates after Arbitrum went live.

Ethereum Daily Active Addres since 21 May 2023. Source: Etherscan

But what if Ethereum was built like Mina at the first place?

An L1 optimized to verify proofs.

(Ethereum is going to that way with verkle tree)

Is L2 parasitic? I don't think so.

L2 is a business of blockspace reselling. Every L2 is free to choose which L1 to be the settlement layer.

Ethereum incidentally is the most secure L1 with highest client diversity, staking pool diversity, highest economic security, and capability to verify proofs. That's make Ethereum a sweet chain to be a settlement layer where in the worst case of failure, everyone is 'guaranteed' to be able to exit on L1.

Is EIP-4844 a Blunder?

About EIP-4844 / Proto-Dank-Sharding / PDS:

  • Introduction of data blobs that expire in weeks.
  • Blobs are 10x cheaper than immutable blocks.
  • This feature is very suitable for rollups DA.

The good news?

L2 fees after PDS fell by 80-90%.

The bad news?

Many L2s are moving to alternative DA.

Ethereum revenue fell by 99% after 4844.

Ethereum Revenue since 13 March 2024. Source: TokenTerminal

Because the revenue come from EIP-1559 burn.

High block demand - high gas fees - more ETH burnt.

Rollups demand move to blobs over blocks.

Is this a blunder?

Let's read EIP-4844.

Oh, look!

PDS is a PROTOTYPE of DANKSHARDING as a part of roadmap onward STATELESS ETHEREUM.

It's a side effect for EIP-4844 to be suitable for rollups.

It's a side effect for EIP-1559 to make ETH deflationary.

EIP-1559 WAS BUILT TO STABILIZE L1 GAS FEES.

NO BLUNDER HERE, IT'S ALL ACCORDING TO THE PLAN.

(yet I do agree with 0xLouisT to reclaim L1 pricing power)

L2 ALWAYS HAVE FREEDOM TO CHOOSE THEIR STACK.

THAT'S THE DEFINITION OF MODULARITY.

(but butt, how about ETH value accrual? [parroting sound])

Value Accrual Rank1. Store of Value2. Staking Asset3. Collateral Asset4. Fees (Revenue)5. Everything Else
-sassal0x

Bitcoin has zero revenue despite being in the first place.

Alt L1s have centi billions of valuation.

The doge is sitting at 15 billion USD.

Meanwhile Ethereum is at the 2nd place after Tron,

EVEN AFTER THE 4844 REVENUE CUT.

Perhaps we should bring back the Internet Bonds meme.

Now let's probe the L2s, shall we?

The Commoditization of L2

It's obvious blockspace is abundant.

What else is abundant? EVM chains.

In 2020, many L1s are EVM chains, even the apps are forks.

In 2024, many L2s are EVM rollups, even the apps are forks.

The switching cost for both user and dev are literally zero.

I won't be surprised if dozens of L2s end as zombie chains.

In this plenty supply of blockspace, where's scarcity?

What's the difference between an OP chain and an OP fork?

Is it the gas fees? Is it the blocktime? Is it the VM?

No, the major difference is the integrated apps in the chain.

The only way for a fork to surpass their upstream is apps.

Arbitrum had a headstart and quickly integrated by uniswap, aave, curve and other L1 protocols.

Arbitrum holds the most TVL among the L2s ever since.

Base is an OP fork integrated with coinbase through smart wallets, gas-free USDbC, and onchain summer.

Thus consumer apps choose Base L2 as a go-to.

We haven't see any superprotocol, or a superL2, if you will.

An L2 with its own VM, sequencer, DA layer, and prover.

An L2 as a whole integrated ecosystem, feels like the L1.

The Aztecs?

Vertical Integration.

However, Base is not an L2. Do we have a real L2 though?

Stages of L2. Source: L2BEAT.com

What's an L2 anyway?

L2 First Principle

A computation done anywhere and settled on L1 as lifeboat.

Computation: A state transition from a → a'.

Anywhere: Any computer, including local device.

Settled: A verified proof, either fraud or validity.

L1: A chain that compute verification function.

Lifeboat: An exit door from L2 to L1 in the case of emergency.

(by lifeboat I mean escape hatches)

There are three stages of L2s

  • Stage 0 : L2 can steal your fund and run away
  • Stage 1 : L2 can run away and your fund is frozen
  • Stage 2 : L2 can't steal nor freeze funds and you can exit

As of today, major L2s stay on stage 0 or stage 1.

The Stage 2 are DeGate (TVL $25M) and Fuel V1 (TVL $600).

L2s in the current state are pretty much AWS.
- 0xMert

The idea of L2 is to let Twitter, Paypal, Whatsapp, or any apps to have 1) Liveness through decentralization. 2) L1 as an exit in case of failure or censorship.

Sounds good on paper, but at what cost?

The Wall of Modularity

L2 unlocks unlimited possibility on building a blockchain.

Arbitrum launched Stylus WASM, Optimism stay an EVM maxi.

Scroll, Taiko, and ZkSync are building zkEVM at different levels.

Movement bet on MoveVM, Atlas and Eclipse prefer the SVM.

Aztec and Starknet built their own language and VM.

Meth is centralizing block building to get 100k TPS.

Fragmentation is a huge issue. If you don't feel that way then you're a web3 web2 normie.
- me

I want to mint on LiveArt, Zora, Sound, Pods.

There are dozens of clicks and pop-ups for 4 actions.

Not to mention the gas fees and loads of websites to visit.

I have to learn to use relays like Across, Decent, Relay, or DLN.

It's obvious that L2 UX is broken bad.

There's something as bad as fragmented UX.

The Forgotten Devs

Most apps today are written in JS, HTML/CSS, and Python.

Most chains support Solidity/Rust, both exotic for web2 devs.

Though Rust adoption is growing as main language of WASM.

Solidity isn't on the list

Another problem is the cost of rollups, relative to websites.

The monthly cost of deploying an AWS EC2 starts from $30.

The cost of deploying rollup is at least $4,000 every month.

L2 Deployment Cost. Source: Hashed Emergent

Million of rollups won't born if devs can't afford the cost.

1 Million Developers before 1 Billion Users

That's why,

Millions of Rollups as L3s

L3 is a rollup built on existing rollups.

Why L3:

  • Cheaper DA and settlement cost
  • Affordable forced inclusion and escape hatch
  • Max specialization with micro-rollups or mini-rollups

Arbitrum so far the most dev-aware team with the launch of Orbit L3s and Stylus WASM. It's clear they're inviting web2 devs to build on web3.

L3 = ONCHAIN APPS

For the time being, I haven't seen any RaaS that provide the real needs of developers. I don't think the next devs are going to care about choosing rollups stacks. They'll rather care:

  • Supported languages
  • Accesibility to other apps
  • Subscription cost
  • Modules and Plugins
Rollup Deployment Interface. Source: Conduit

The real mammoth in the room,

How do these L3s interoperate with each other?

Oligopoly of Superchains

Intents are great, but it doesn't solve fragmentation.

Intents only connect supply-demand between rollups.

We need a native unity between the rollups.

Each rollups have their own way to generate proofs.

Therefore each rollups have their own trust assumptions.

Fragmentation is caused by different trust assumptions.

The ideal world is to unite rollups into a proof of singularity.

However, rollups have different thesis, views, and systems.

This lead to the idea of clusters of L2s.

Starknet arguably had the first idea with Fractal Scaling in 2021.

Though we haven't seen anything live in practice. (soon™)

Optimism stated the Superchain thesis, dating back to 2023.

Arbitrum caught up quickly with Orbit Announcement.

ZkSync stay in the loop with Hyperchains.

(kinda like superchain parody, rebranded to elastic chain)

Polygon unveiled the AggLayer in early 2024.

Today, Arbitrum Orbit and OP Superchain are dominating.

Both of them counts for almost a third of all rollups.

Rollup Framework Share. Source: L2Beat

We can see oligopoly of superchains is plausible. After all,

Miners, Relays, Builders, Restakers, Oracles.
ALL of them had fell into Oligopolies.
-stats

I'm not smart enough to find way for L2 team to collaborate. It's like requesting Nvidia, Intel, AMD, and ARM to do joint venture and build a chip singularity.

The only thing I can think of is for all eth stakers create a bounty of witness encrypted private key on L1 filled with 10 million ETH. Then all the stakers vote for the best L2 design. Maybe all L2 teams will be united for $20B payoff.

I have a slight of hope on Circle STARK built by collaboration from Polygon and Starkware. If two major L2 frameworks decide to unite their proof and make both assets are composable, maybe other L2s will join the party and lead to domino effect where everyone fall into proof singularity.

Though I'm sure there always be that monolith trying to be authentic.

The Aztecs?

Towards Sufficient Decentralization

Do L2s need contract immutability?

That'll be a hassle, imagine L2 migration in every upgrade.

Do L2s need the most decentralized sequencer?

That's a slug, imagine wait 12 seconds for based sequencer.

Do L2s need the most secure data availability?

That's expensive even with the blobs, thus L2s are using alt DA

Perhaps we'll have a maximally decentralized L2s.

Scroll and Taiko are leading that way with based sequencer.

Gwyneth and Facet are max decentralized as they live on L1.

For the most L2s,

Sufficient Decentralization is Sufficient.

BNB chain has been live with only 52 nodes.

Cosmos 180 nodes. Polkadot 297 nodes.

We don't need 4000 to 6000 nodes for L2s.

Some reasonable choice for L2 teams to make:

  • Integration with AVS infras
  • Upgradable contract through councils
  • Post data to multiple DA layers
  • Embrace multi-proofs
  • Add settlement to Bitcoin via OP_CAT

Another possible future: local computation, onchain verification.

Low computation usage like payment, ZK and TEE computation are possible to done locally, then the proof is aggregated, verified, and settled onchain.

  • ZK-ZK-Rollup: Aztec 2.0 and Polybase
  • NES Emulator: Dappicom by Tonk
  • Prototype of TEE-Rollup by InitC3

The Seamless Dream

Imagine it's 2026.

Super-orbit-elastic chains went live.

Espresso and Astria sequence hundreds of rollups.

Nodekit and SUAVE become the superbuilder.

Anoma and Khalani let anyone solve user intents.

(i don't know which are witch anymore)

Rollups are settled with naysayer proofs.

Nebra and T1 aggregates millions of zk proofs.

Aligned and Hyle verify every zk proofs.

Fabric and Ingonyama accelerate with zk hardwares.

All infras secured by restakers from Eigen and Symbiotic.

Conduit and Caldera L3 deployment costs drop to $30.

LiFi and Socket abstract the backend devex.

MagicAccount and 1Balance abstract the frontend ux.

How would this imaginative world look like?

I will only need e-mail or passkeys to access my wallet.

I'll give another try of the previous 4 actions.

I want to mint on LiveArt, Zora, Sound, Pods.

LiveArt on Polygon, Zora's Zora, Sound on OP, Pods on Base.

The [chain]-ETH in my wallet aggregated into 'ETH'.

That's nice, I'm done with balance fragmentation.

But I must get some MATIC on Polygon, or so I thought.

From now on, I can pay with USD balance on any chain.

Did I mention that everyone have weekly 5 gas-free tx?

Perhaps some wallets have business model like Revolut.

When I open liveart, zora, sound, and pods web page,

My apple id already connected to magic passkeys.

All I need is to put my finger on my mac to confirm.

Four actions done in 4 clicks and 4 touch.

If I activate my session keys, it would be 1 touch only.

The solvers will figure everything out for me.

Perhaps I'll only need a single sentence as a prompt.

Then some onchain ai agents will do the rest.

At least, that's the dream.

The chain aggregation thesis is not new.

Cosmos and Polkadot had it earlier.

They also have native bridge IBC and XCMP.

Apps went live, liquidity fragmented, and everything fractured.

L2s are even more complex.

There's no guarantee L2s aren't going to end the same way.

Still, I believe the buidlers are staying for the Infinite Games.

Thank You

------