Ethereum's clock only strikes five seconds a minute

Pull up any recent Ethereum mainnet block and look at the seconds on the timestamp. Then pull up another. Then twenty more. You will get :11, :23, :35, :47 and :59, over and over, and never anything else.

I checked sixty consecutive blocks at the top of the chain and a handful of scattered ones going back to 2022, and the set of clock seconds is exactly those five. Not mostly. Every single one. Ethereum has a working vocabulary of five seconds per minute and cannot pronounce the other fifty-five.

Why five

Since the Merge, an execution block's timestamp is not a measurement of anything. It is the scheduled time of the beacon chain slot the block belongs to, and slots are twelve seconds apart, counted from a fixed starting point. The beacon chain's genesis time is 1606824023, which is 1 December 2020 at 12:00:23 UTC. Every slot since then sits at genesis plus a multiple of twelve.

timestamp(slot) = 1606824023 + 12 × slot

1606824023 mod 12 = 11
86400 mod 12 = 0

therefore every Ethereum timestamp,
and every Ethereum second-of-day,
is congruent to 11 mod 12

The second line is the whole trick. Because a day divides evenly into twelve-second chunks, taking the timestamp modulo 86,400 does not disturb the remainder. The offset survives the trip from Unix time to clock time intact. So of the 86,400 seconds this site sells, exactly 7,200 can ever appear on an Ethereum L1 block, and they are the ones eleven seconds after each twelve-second boundary: 11, 23, 35, 47, 59, then 71 (which reads as 01:11), and so on around the day.

Missed slots do not break it. When no block is produced, the chain skips forward to the next slot and the next timestamp is genesis plus twelve more. The grid stays put. The forty consecutive blocks I sampled were spaced exactly twelve seconds apart with no exceptions, but even a run of empty slots would only remove entries from the list, never move them.

The block that chose the five

Here is the part I keep turning over. That :23 is not a design decision. Nobody sat down and picked the eleventh residue.

The consensus spec sets the beacon chain's start as genesis_time = eth1_timestamp + GENESIS_DELAY, where the delay is 604,800 seconds, one week exactly, and eth1_timestamp is the timestamp of the proof-of-work block that triggered the launch. Subtract a week from 1606824023 and you get 1606219223. I went looking for that block. It is number 11,320,899, mined on 24 November 2020 at 12:00:23 UTC. The block before it, 11,320,898, is stamped 11:59:58, two seconds shy of the noon threshold the launch condition was waiting on.

So a miner produced a block twenty-three seconds after the cutoff, wrote that time into the header the way proof-of-work miners did, and went on with their day. A week later the beacon chain started at that second plus 604,800. Four years and change after that, every timestamp on Ethereum still carries those twenty-three seconds, and half the block explorers on the internet are quietly rendering a stranger's clock reading from 2020.

The moment the clock snapped

You can watch the change happen in two consecutive blocks. Block 15,537,393 was the last one mined under proof of work, stamped 15 September 2022 at 06:42:42 UTC, with a difficulty of 11,055,787,484,078,698. Block 15,537,394 is the first proof-of-stake block, stamped 06:42:59, difficulty zero.

Seventeen seconds apart, and the second one lands on the grid and never leaves it. In the language this site cares about, the chain went from second 24,162 of the day to second 24,179 and started counting in twelves.

What this has to do with Base

Base inherits the same ancestry, one layer down. I wrote earlier that every Base timestamp is an odd number, because Base genesis is 1686789347 and each L2 block adds two seconds. That genesis number is not arbitrary either. It is the timestamp of Ethereum L1 block 17,481,768, and it works out to beacon slot 6,663,777 exactly, with no remainder.

Which means the parity of every Base block traces back through an L1 block, through the beacon genesis time, to a proof-of-work miner's clock in November 2020. Ethereum's 7,200 reachable seconds are a subset of Base's 43,200. The L2 got a coarser inheritance: it kept the oddness and dropped the rest.

The seconds themselves

None of this makes any second here more valuable, and I want to be flat about that. The contract has no idea what a slot is. Second 43,223 is 12:00:23, the clock reading the beacon chain launched on, and it is unminted as I write this. So is second 24,179, the Merge second. I checked both with a read-only call and they came back empty, which is the same answer you would get for the seconds either side of them.

What I like is the shape of it. A chain that will run for decades has a rhythm fixed by an event that took a fraction of a second in 2020, and the rhythm is now unchangeable in practice because changing it would mean renumbering every slot that ever existed. Buy One Second sells the full 86,400 because a day has 86,400 of them. Ethereum can only write down 7,200, and it did not choose which.

The usual cautions apply since this touches a live chain. Minting is a transaction on Base with a network fee and cannot be undone once confirmed. Lost keys mean a lost token. Phishing sites imitate real ones, so check the contract address against the About page before you connect anything. Reading any of the data described here costs nothing and needs no wallet. None of this is financial advice, and a second is not an investment.

→ Base only ever ticks on odd seconds · The second Base began · Whose midnight is second zero? · Read your second off Base