Latest

Ethereum

Base activates Denim with 200 ms blocks and its blockchain changes

Base activates Denim on November 5 at 18:00 UTC with five blocks per second, removes Flashblocks and forces apps that read pending state to migrate.

Base activates Denim with 200 ms blocks and its blockchain changes
Base activates Denim with 200 ms blocks and its blockchain changes

Base has set a date for the change it had been announcing for months. The Denim hard fork activates on Thursday, November 5 at 18:00 UTC and changes the network's unit of time: where there used to be one canonical block every two seconds, there will be five full blocks per second, each with its number, its hash, its state root and its receipts. The figure that sums up the leap is simple: 2,000 milliseconds become 200.

The detail that turns this into a story with consequences rather than a performance note is that Denim retires Flashblocks, the preconfirmation layer Base has used since July 2025. Applications that subscribe to those streams or query the pending block state will stop receiving a response after activation, and will have to switch to reading canonical blocks. This is not a deprecation twelve months out; it is the same day.

For anyone who only uses Base as an end user, the change will be invisible or nearly so. For anyone running an indexer, a bridge or a bot that queries state before it is confirmed, the week of November 5 is the busiest of the year. The network that bills itself as the reference L2 for consumer applications has just asked its developers to change the way they read the chain.

What exactly changes on November 5

Base's documentation describes it preciselyDenim moves from producing one canonical block every two seconds to producing five per second, and each of those blocks carries its number, its hash, its state root, its receipts and forkchoice updates. Moreover, each block goes through its own lifecycle: unsafe, safe and finalized. There is no longer one fat block grouping transactions, but five small, complete blocks per second.

Time also changes scale. The Ethereum header keeps its timestamp in seconds, so Base adds a metadata deposit called BaseTime that provides the sub-millisecond part. With that, RPC calls return timestamps with millisecond resolution and applications can order events that previously fell within the same second with no way to distinguish them.

The practical consequence is that the network gains granularity. A market that needs to know what happened before and what after no longer depends on position in a list within a two-second block, and instead has an explicit temporal order. For exchanges and for payment systems, that detail is worth more than the 200-millisecond headline.

Goodbye to Flashblocks and the pending state

Flashblocks was the intermediate solution. For more than a year, Base published incremental updates of a block's pending state so that applications could show results before confirmation. With Denim those blocks are canonical, so the intermediate layer is redundant, and the company has decided to retire it rather than maintain two ways of doing the same thing.

That leaves three broken cases if nobody touches the code. Subscribing to Flashblocks streams, querying the tag pending in calls such as eth_getLogs or eth_getBlockTransactionCountByNumber, and any logic that assumes pending state and the latest block are two different things. Base's migration guide lists every pattern and its replacement, and it is worth reading before the date, not after.

The operational advice is to take inventory before November 5. Any integration that uses the pending tag, any webhook that depends on a preconfirmation, and any dashboard that displays the "before confirmation" state must be reviewed. If the team does not know which parts of the system use Flashblocks, that is the first job and also the most urgent one.

What developers need to change

The to-do list is short and specific, which is the best thing that can be said about a forced migration. Base published a table with every affected call and its replacementand the overall pattern is the same in every case: stop asking for pending state and read the latest canonical block, which now arrives five times per second and with the same latency that preconfirmation used to have.

Affected pattern What happens after Denim What to use instead
Flashblocks subscription Stop issuing Canonical block flows
Query with tag pending No valid response Latest canonical block
Logs with range up to pending Invalid range Range over canonical blocks
Preconfirmation logic of its own Duplicated Unsafe, safe and finalized cycle

The fourth point is the one that causes the most work in large teams. Many applications built their own "I think this will be confirmed" layer on top of Base's preconfirmations, and now they have to decide whether to remove it or rebuild it on the new blocks. In most cases the right answer is to remove it, because the 200-millisecond block already serves that function.

200 milliseconds and the history problem

There is a side effect that almost no one is discussing and that the documentation does mention. By multiplying the number of blocks per second by ten, the history covered by a fixed number of blocks shrinks in the same proportion. A window that previously spanned several hours now covers a fraction, and any flow that depends on historical hashes or beacon roots has to revisit its assumptions.

That detail affects bridges and proof systems. A bridge that validates an event against a block from a day ago now has ten times as many blocks in between, and a proof that waits for a number of blocks as a proxy for time needs to change units. The documentation warns that the number of retained values does not change, but the time coverage does, and that difference is what causes silent errors.

The solution is to stop using the block number as a clock. Base now offers timestamps with millisecond resolution, and any logic that needs to "wait an hour" should rely on them rather than counting blocks. It is the same lesson learned by the networks that sped up their production over the past two years.

The race for the millisecond is no longer Ethereum's

What is interesting about Denim is not the number, but who is setting it. Base competes with Solana, which has spent months selling finality in hundreds of milliseconds, and with the L2s promising ever-faster confirmations. The fact that a network built on Ethereum makes the leap to 200 milliseconds without abandoning the layer one says a lot about where the battle for applications is being fought.

The cost of that leap is complexity for the ecosystem. Every acceleration changes assumptions that developers took for granted, and every change of assumption breaks integrations that no one is watching. The network gains speed and the ecosystem pays in maintenance, a trade-off that has become routine over the past two years.

If you have an application on Base, the calendar is already set and there is no room to wait and see what happens. If you only use the network to move funds, you will not notice anything except that everything runs a little faster. And if you work on another L2, the November 5 date is a good excuse to review how much of your logic assumes two seconds per block.

José Maldonado

Author

José Maldonado

Activist and blogger on technology, free software and blockchain.

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading