Three blocks on the first night

The page we opened this morning. Three steps, three blocks.

Yesterday’s hardware post ended with a used Goldshell SC Box solo mining BLAKE2b mainnet through our own Umbrel node at about 930 GH/s. This morning the payout address held 9.39316932 BTC. Between 01:10 and 09:17 UTC, while we slept, the box found three blocks.

The three blocks

Times are UTC; the clock in the screenshot runs four hours behind.

Each reward is the 3.125 BTC subsidy plus that block’s fees, a few thousand sats: with a block landing every three minutes or so there is not much of a mempool to collect from. Every coinbase carries the same tag, 8-30 NYPost Deride And ConquerBLAKE2bMiner.com. More on that below.

Two sources agree on all of it. The screenshot is mempool.guide’s page for the payout address. Our own node says the same: at 11:30 UTC the DATUM gateway on the Umbrel was building job 962,258 on top of the same chain, 15 hours 49 minutes since it started, 775 shares accepted, zero rejected.

The gateway this morning. Nothing on it was touched overnight.

The headline in every coinbase

The tag is DATUM’s primary tag, then its secondary, no space between. The first half is the activation headline, and nothing after block 961,640 needs it.

Satoshi put a newspaper headline in the genesis block to prove it was not mined before the paper existed. PR #359 turned that into a consensus rule for one block: rc4 refuses to start without blake2b_headline set, and rejects block 961,640 unless its coinbase carries the exact string. That is why the string reached the network before the block that carried it, and why every node operator typed the same line. No later block is checked. Luke said so in the Knots Discord the evening of 30 August, and the mainnet guide’s update walks the source.

So the tag in our blocks is a leftover from the DATUM config, not a rule, and the blocks are no less valid for it. Two things carry the string and only one should keep it. The DATUM primary tag goes to BLAKE2bMiner.com at the next restart, which the final release will force. The node’s blake2b_headline line stays, because a fresh sync still validates 961,640 against it. If rc5 resets the chain, a new activation block gets a new headline and that line changes with it.

How lucky was that

The same arithmetic tells you what to expect from your own box.

Mining is a lottery where every hash is a ticket. Difficulty says how many tickets a block costs on average: difficulty times 2^32, and at the current 30,393,776 that is about 1.3 × 10^17 hashes per block. The number looks tiny next to the SHA-256d difficulty it replaced because rc4 shifted the target 22 bits easier at activation (Blake2bTargetShift=22 in the mainnet parameters), which is what let a chain with almost no hardware produce blocks at all.

Our box does about 930 billion hashes a second. Divide, and the expected wait per block at today’s difficulty is about 140,000 seconds, call it 39 hours. That is an average, not a schedule. The lottery has no memory.

The box went onto mainnet at about 19:40 UTC on 30 August, so by the third block it had been mining for 13.6 hours. At one block per 39 hours, that window is worth 0.35 expected blocks. Three or more in it is roughly a one-in-two-hundred outcome (Poisson, for anyone checking). We got lucky. The same math says the next block could take a week.

For scale, the whole chain: 617 blocks in the 29 hours from activation to 962,257 is one every 170 seconds, which at 1.3 × 10^17 hashes each puts the network near 770 TH/s, about 830 boxes like ours. Chris Guida, who assembled the proof-of-work code, says the starting hashrate was set deliberately low and the block rate is about what they were aiming for. Difficulty still adjusts every 2016 blocks, by at most 4x per step. The first retarget lands at block 963,648, about 65 hours away at the current pace, and at this block rate it moves up. Expect the wait per block to stretch toward a week after it.

Why the explorer files us under “Unknown”

We hoped to see BLAKE2bMiner.com on mempool.guide’s mining pools chart. It is not there, and the reason applies to your tag too.

An explorer does not know who mined a block. It guesses, by matching the coinbase text against a hand-maintained list of patterns, one entry per pool or miner someone has bothered to add. mempool.space uses a public list on GitHub. mempool.guide runs its own, which is why it can name dozens of solo miners that mempool.space would lump together. Anything the list has not heard of lands in “Unknown”, which held 109 of the last 549 blocks when we looked. Our three are in there.

The tag is in the chain regardless; the chart is a label on top of it. We will ask whoever maintains that list to add ours, and until then the address page above is the record.

The caveats, same as yesterday

• Rehearsal. Luke Dashjr calls this run a rehearsal. The final release is due 1 September, tomorrow. If anything forces an rc5, the chain resets to 961,639 and these three blocks go with it, along with every other block since. We wrote that line in the mainnet guide before we had anything to lose. It reads differently today.
• Maturity. A block reward cannot be spent until 100 blocks have been built on top of it. The first two have matured; the third does at 962,293, likely before you read this.
• Worth. We are not pricing it and there is no market we would point you at. Nothing here is advice.
• Nothing changed. Same box, same node, same Ledger address from the guide. It kept hashing.

If you have a Sia-class box in a drawer, the two guides are the node and the hardware. The arithmetic above is the expectation: about a block every day and a half at today’s difficulty, longer after the retarget, and nothing at all if the rehearsal resets.

Discuss this post on X