If you are interested in high performance system-building, read the code at There are ~40 different components, each modularized into its own library. - monad-consensus implements the core BFT algo - monad-eth-txpool implements the local mempool - monad-dataplane implements the core networking logic - monad-raptorcast implements the block propagation mechanism - monad-archive implements local archival of transaction artifacts - monad-event-ring implements the real-time data system - and so on You can learn a lot from reading the code. The engineers at @category_xyz aren't messing around. Everything is purpose-built and optimized, no fluff. That's just how they roll. The Monad Way.

WebSockets are live on Monad Testnet! This includes the following subscription types: - newHeads - logs - monadNewHeads - monadLogs newHeads and logs behave the same as in Geth, but return updates as soon as blocks finalize. Since updates wait for finality, you don't need any code to handle reorgs (very convenient) monadNewHeads and monadLogs are the same thing, but they return updates as soon as blocks are received and speculatively executed (very fast)

Testnet is back up after the scheduled upgrade - MonadBFT is live! - JIT EVM is live! - WebSocket support is live! and more

Vote now, compute later

When we began working on Monad, one of our goals was to bring HFT techniques to the blockchain. For example, high performance apps should be able to run a node and query blockchain data in realtime from a shared memory queue. Real Time Data is the start of that!

Today, Tue Aug 12, Monad Testnet will halt at 13:30 GMT for a brief upgrade to the consensus mechanism. The changes that happened to testnet-2 two weeks ago (linked in the next post) are going live on the primary testnet. This is a big upgrade, as it brings the following to the primary testnet: - MonadBFT - JIT EVM - Support for real-time data via websocket and shared memory queue - Reduction in block times to 400 ms - Support for the BLS12-381 precompiles (EIP-2537) - and more!

Doing some writing today about asynchronous execution (aka deferred execution). This is one of the "big ideas" of Monad that will seem obvious in hindsight: that as soon as transaction order is determined, state is determined. You might not know this, but in most blockchains including Ethereum and Solana, the nodes evaluating a list of transactions (i.e. a block) must also agree upon the result of that list of transactions (i.e. the state root after those transactions) as part of the voting process. That requires them to execute those transactions prior to responding with a vote, which obviously greatly shrinks the amount of work they can be expected to do. This expectation is too strict, and is one of the big reasons why blockchain throughput is so limited.

monad testnet will be upgraded on Tuesday. for a preview of what will land then, the changelog describes the changes (which landed on testnet-2 about ten days ago)

What did you get done this week?

Michael Chen (@indigo4043) is live now!