Ethereum

The future is 4337 on L2.

But on L2, the main cost is calldata, and 4337 can use a lot of that.

Enter what @vitalik.eth's called the “infamous 1600-byte transfer”: a bundle (~800 bytes) containing a single Daimo transfer (another ~800). Too fat!

We solved it with bundle compression.

1. RESULTS. 4x less calldata, 2x cheaper, room to go further.

Here’s what a userop looks like before/after: https://tinyurl.com/compression-test-ts

Here’s a prod transfer. ~$1 before, $0.38 now. We can squeeze out another 2x, then pray to the danksharding gods. https://tinyurl.com/2s365wjk

Real World Ethereum 🌎 
building /daimo ✳️ 

2. How does this work?

- Compress ops in arbitrary app-specific ways
- Bundler sends via a proxy contract
- Proxy inflates > calls EntryPoint

In our prototype, any app can register their own onchain inflator contract.

3. Do you have to change your contract wallet?

No! The ops submitted to the EntryPoint are exactly as before. You’re still signing over the standard op, not over any compressed form.

The only thing that changes is less data = less cost = lower preVerificationGas.

4. Why still use 4337?

All benefits still apply. Shared tooling. Paymasters. Security. (The proxy has no security responsibilities: those stay with the extensively-vetted EntryPoint.)

And critically: you keep your option for censorship resistance. Pay a bit more gas, submit the uncompressed op to public mempool.

5. Here’s our prototype: https://github.com/daimo-eth/bulk

If you're running AA on L2, let us know if you want to try it. Help us navigate toward 1cent/1second transaction Valhalla.

+ big thanks to @kristof, great source of ideas & runs Pimlico, the best bundler we've seen. More coming soon. /end

PS. Here's the full breakdown. @ansgar.eth 's original 1600 byte analysis with a second tab showing compression.

PS. Here's the full breakdown. @ansgar.eth 's original 1600 byte analysis with a second tab showing compression.

https://docs.google.com/spreadsheets/d/1rf3AJMmm9BCkBsmQiRq4yP3Zhne9vIW-ib0Dp4QD2Tw/edit

Mini apps will eat the internet | Latest drop: /clankermon | Building : /large /fctrivia /miniapps /flappycaster /jeeves /source @degentip @playsource /jump

We can do better

Maybe not in the current degen case where every 4337 bundle tx contains exactly 1 op

But in the near future when every new L2 block = new 4337 bundle = multiple ops

We can beat legacy eoas on cost…in the (likely) event that major rollups don’t implement native stateful compression

Massive 👏 squeeze another 2x and we are almost at cost parody with EOAs 🤝

founder @explorelabs

don't take my casts seriously please (most of them at least)

They’re not alternatives. 3074 is a backcompat mechanism to upgrade legacy EOAs into contract accounts

(In practice: those contracts will implement 4337)