How it went down

BIP-110 was a temporary soft fork to limit arbitrary data in transactions for about a year. It needed 55% of blocks in a two-week period to signal support, a bar its authors had already lowered from Bitcoin's customary 95%. If that was not reached, nodes running the BIP-110 rules, mostly Bitcoin Knots, would start rejecting any block that did not signal, from block 961,632 onward.

Support in the final period before the deadline was 2.53%. Fifty-one blocks out of 2,016, nearly all of them from the OCEAN pool.

Block 961,632 arrived late on 8 August UTC, Sunday morning here. AntPool mined it without the signal. Bitcoin Core nodes, and the overwhelming majority of the economy, accepted it. BIP-110 nodes rejected it, and a small mining group called Roughnecks, working through OCEAN, produced a competing block that did carry the signal. For a few hours there were two versions of Bitcoin.

The BIP-110 chain produced exactly two blocks. In the same eight hours the main chain produced 48. By the next day the gap was more than 200 blocks, Roughnecks had given up and told everyone else to stop, and the minority chain sat frozen at block 961,633.

The reason is mechanical, and it is the same reason we gave in July. A minority chain inherits Bitcoin's full mining difficulty but only its own sliver of hashpower, roughly 0.15% in this case. Difficulty adjusts only every 2,016 blocks. At two blocks a day, that adjustment was about 350 days away. The chain did not lose an argument. It ran out of blocks.

The price did not react at all.

Why it failed

Three things, and none of them is a surprise.

First, the miners never came. Foundry, AntPool, ViaBTC and F2Pool, between them most of the world's hashpower, never supported the proposal, and the only pool that did was too small to matter. A soft fork enforced by nodes still needs miners to build the blocks those nodes will accept.

Second, the economy never came either. No major exchange, wallet, custodian or Lightning implementation committed to the BIP-110 rules. Node counts were quoted at anything from 2% to 15%, and the higher figures included a great many nodes that were probably not real. Bitcoin's rules are ultimately enforced by the nodes that matter economically, and those stayed put.

Third, the design assumed the split would resolve in its favour. It shipped without replay protection, on the argument that a soft fork does not need any. That left anyone tempted to sell fork coins exposed: a transaction signed on one chain was valid on the other, so a buyer could rebroadcast it and take real bitcoin from the seller. The only sensible advice in this case was to do nothing, not to try to split your coins, and to wait for the minority chain to die. It took hours.

Michael Saylor, who had published 110 reasons to oppose the proposal, put it simply. BIP-110 was free to fork, and the network was free not to follow.

Then the hard fork

What happened next is the part worth reading twice.

Within a day, the proposal's author declared that the large pools had colluded in a secret hard fork against the node network, and that the community would change Bitcoin's proof-of-work algorithm to fire the miners. Luke Dashjr, who maintains Bitcoin Knots and co-founded OCEAN, put a date on it: 1 September. Every existing mining machine would be useless on the new chain. That is no longer a soft fork of Bitcoin. It is a new coin.

The algorithm chosen was BLAKE2b. A rehearsal ran on 30 August and the chain launched on 1 September, the date BIP-110 would originally have activated. It arrived almost empty. No major exchange, wallet or Lightning implementation supported it. The project's own FAQ told exchanges to pause deposits and withdrawals around the split. Light wallets cannot read its new block headers, and the replay problem is now permanent, addressed only by an opt-in signing mode most wallets do not have.

A coin with no name got a provisional ticker, BTCB2, on one small exchange, where the highest bid on launch day was about AU$115 and the lowest ask about AU$265, a spread that means nobody was really trading. Five days later, on two obscure venues turning over roughly AU$1.4 million a day, it printed a high of about AU$2,500. At that price its notional value would rank it among the ten largest coins. It is not tracked by CoinGecko or CoinMarketCap. Treat the number as a curiosity, not a price.

Meanwhile, OCEAN admitted that for about 18 hours during the August split it had routed hashrate from miners who had chosen the main chain onto the BIP-110 chain instead, through a default in its configuration. It refunded about 0.3 bitcoin. The pool's reported hashrate fell 96% during August. Miners demanded new leadership. At the end of August Dashjr resigned as OCEAN's chairman and chief technology officer and the company bought back his equity. A separate motion was filed to strip him of his role as a BIP editor.

What it actually tells us

Bitcoin's governance is slow, conservative and resistant to change, and August is what that looks like when somebody tests it. The block size war took two years to settle the question of who controls the rules. This one took eight hours to confirm the answer: not miners alone, not node operators alone, but the whole economy, and the whole economy has to agree.

That cuts both ways. It means nobody can push a rule change onto you that the network does not want. It also means proposals many Bitcoiners like, covenants among them, now face a bar that was already high and has just been raised. If you think Bitcoin should never change, August was a good month. If you have an upgrade you are hoping for, take note.

The underlying dispute is not resolved. Arbitrary data is still being written into blocks, node operators still pay to store it, and the people who cared enough about that to fork have now left to build somewhere else. Some of them contributed a great deal to Bitcoin over the years. None of this was fun to watch.

What it means for you

Your bitcoin was never at risk. The chain that has run since 2009 never missed a block. As we said we would in July, we paused on-chain bitcoin deposits and withdrawals as a precaution while the split played out, for about five hours on the Sunday morning. Lightning came back first, and buying, selling and AUD transfers ran throughout. Balances held with Bitaroo were never affected.

If you held bitcoin in self-custody at block 961,632, you technically also hold the same amount of BTCB2. Our advice is the same as it was in July: do nothing. Claiming fork coins means splitting your coins first, and doing it wrong can move your real bitcoin. It also means running the fork's software with your keys. Every past fork has spawned websites and apps offering to claim your free coins, and their only purpose is to harvest seed phrases. Nothing that asks for your seed phrase is claiming anything for you.

Everything else we said in July stands. Self-custody remains the destination for most people, Bitaroo Vault is there if you would rather leave your bitcoin with us behind a time-lock, and the best thing to do during a contentious event is very little.

BIP-110 set out to change Bitcoin. What it proved is how hard that is.