Monad, after the first year: a parallel EVM that finalises in two slots

Monad is a layer-1 blockchain. It runs Ethereum virtual-machine bytecode, so a contract compiled for Ethereum can be deployed without a new compiler, and the remote-procedure interface is the one operators already know from geth. What it does not share is the way a block is agreed, executed and stored. Those three jobs are split, pipelined and, where the work allows it, run side by side.

Public mainnet went live on 24 November 2025, with MON as the native coin. The chain id used by wallets on that network is 143. Through the first half of 2026 the public parameters were a 400 millisecond block and finality two slots later, about 800 milliseconds. In July 2026 the MIP-12 hard fork cut the block time to 300 milliseconds. Finality of block N now lands when block N+2 is proposed: 600 milliseconds. Speculative finality, which reverts only in rare cases of leader equivocation, is one slot.

Four systems, one receipt

The Foundation’s own account names four changes. MonadBFT agrees an order of transactions and resists a leader quietly dropping the previous block, a failure called a tail fork. Asynchronous execution lets consensus move on while execution of the last block is still finishing, which is what buys the execution engine its time budget. Optimistic parallel execution runs transactions together and re-runs any that touched the same state. MonadDb is the state database built for that access pattern, not a stock key-value store asked to do something it was not shaped for.

The stated throughput is 10,000 transactions a second. That is a design target, not a promise that every quiet Tuesday looks like a stress test. What an outside observer can check is narrower and more useful: blocks arrive on the published interval, a transaction either lands in one or it does not, and the receipt is final two slots later. A blog post that quotes “10,000 TPS” without that distinction is advertising.

The receipt is the product. The white paper is the reason the receipt can be that short.

What the year actually changed

CertiK, writing on 20 October 2026, put total value locked near $1 billion as of 16 October, and described a network that had stayed up since public launch, with more than 60 million uninterrupted blocks through the end of August and a finalisation rate it put at 99.999 percent. Those are an auditor’s figures, not the chain’s own dashboard, and they should be read that way. They are still the right order of magnitude for a chain that is no longer a testnet with a marketing site.

A proposed consensus design called Cadence aims at 100 millisecond blocks and 200 millisecond finality. It is not the network you send a transaction to today. MIP-10, a RaptorCast change in which validators can vote before the full block has arrived, is likewise ahead of the software most clients are running. The desk’s rule is simple: write the live parameter, then the proposal, and never the other way around.

The state root arrives late

On a chain that executes while it agrees, the block can carry the result of the transactions inside it. Monad separates the two clocks. Consensus orders a block. Execution of that order finishes afterwards. The Foundation calls this asynchronous execution, and it is the reason a 300 millisecond slot is not also a 300 millisecond budget for every call in the block. A node can still be voting on a new proposal while an earlier block is being written to the database. The receipt is real. It is simply not born in the same instant as the vote.

What a twelve-second habit gets wrong

Bytecode compiled for Ethereum deploys here. Habits do not travel with it. Waiting “two blocks” before treating a payment as settled is a wait of well under a second, and it is still the wrong wait if what you needed was full finality rather than inclusion. Timestamps step on the slot, not on a twelve-second clock, so a contract that used the timestamp as a coarse timer is now a finer timer whether the author intended it. An oracle that writes once a minute is the slow machine in the room. The chain will not pause for it, and a bot that assumes it will is early by hundreds of slots.