What BIP-110 Actually Proved: How to Read a Bitcoin Fork Fight

A piece of Bitcoin broke off in August and got stuck. Almost everything you're hearing about why is backwards.

The claims are already leaking out of Bitcoin Twitter and into normie crypto spaces, usually delivered with total confidence. Here's the machinery, the record, and four checks you can run yourself the next time somebody tells you Bitcoin is about to change.

On Saturday, August 8th, at 19:35 UTC, AntPool found block 961,632.

So did a mining outfit called Roughnecks, with a different block at the same height. From that moment, two versions of Bitcoin existed.

One kept going. By Tuesday it was at block 961,959. The other produced exactly two blocks, stalled at 961,633, and hasn't moved since. Roughnecks walked away within hours (they've since come back to "solo mine"), and OCEAN, the pool they mined through, watched its own dashboard hashrate fall from roughly 36 exahash to about 1.25.

Here's the trap that chain is in. It inherited Bitcoin's difficulty at the instant of the split but took along a fraction of one percent of the machines (Michael Saylor put it at about 99.85% of hashpower staying with Bitcoin). Difficulty only recalculates every 2,016 blocks, and this chain has mined two. One live monitor now puts that adjustment 6.3 years out. The day after the split it was 350 days. The exit recedes faster than you can walk toward it.

Two blocks, then silence, then a slow-motion trap. That's the whole event.

The interesting part isn't the fork. It's the story being built around it, in real time, for an audience that mostly wasn't watching.

So what actually happened at that block?

Every Bitcoin node (yours, mine, an exchange's) keeps its own copy of the chain and checks every block against a list of rules before accepting it. Not "trusts." Checks. That's the entire point of running one.

Those rules only change when essentially every node agrees to change them at once. That's a soft fork: a new rule tightening what counts as valid, rolled out carefully enough that nobody gets stranded. You coordinate it by signaling. Miners set a flag in their blocks saying "I'm ready," and when enough are ready, the rule switches on and nothing breaks.

BIP-110, formally the Reduced Data Temporary Softfork, wanted to cap how much arbitrary data you could stuff into a Bitcoin transaction, for one year. It needed 55% of blocks across a two-week window to signal on bit 4.

It got 51 blocks out of 2,016. Call it 2.5%.

So its authors had built in a fallback, and this is the hinge of everything that followed. Starting at block 961,632, BIP-110 nodes entered mandatory signaling: they would reject any block that didn't flip bit 4. Not wait for it. Reject. Treat as invalid. Refuse to build on.

Read that again, because it answers the question everyone's arguing about. Bitcoin's rules didn't change on August 8th. One group of nodes adopted a new rule about which blocks they'd accept, almost nobody flipped the bit, and those nodes rejected the chain the rest of the network was on and started their own.

A group added a rule, then stopped recognizing everyone who hadn't.

Most of my examples below come from Matthew Kratter's Bitcoin University channel, which ran heavy BIP-110 coverage to a large retail audience. I'm not picking on him. He's just the clearest and most-watched version of arguments that are now everywhere, and unlike most people in this fight he argues on the record, on camera, with dates attached. That makes him checkable. Being checkable is a virtue, so let's check.

Check one: which side changed its rules?

The claim now circulating is that the miners split off from Bitcoin.

Here's the thing. Before the fork, that same channel explained the mechanism correctly and on camera: BIP-110-compliant blocks "will be accepted by all Bitcoin core nodes and other non-BIP 110 enforcing nodes," while "Bitcoin core node blocks will not be accepted by BIP 110 nodes unless those blocks have flipped the bit and are signaling."

That's the split, described accurately and in advance, by an advocate. One side's blocks were acceptable to everyone. The other side's criteria had narrowed. Narrow your criteria, and if the world doesn't follow, you're the one who ends up somewhere else.

The same video reached for the Protestant Reformation as an analogy, describing a unified church that split "when Martin Luther decided to change the consensus rules."

Now, we've used that metaphor ourselves, with Satoshi as Luther, the central banks as the church selling indulgences, and the internet as the printing press. It's a good metaphor. The trouble is what happens when you aim it at Bitcoin instead of at the Fed, because the analogy is exactly right about who moved. The reformers changed the rules and left. Nobody thinks the Vatican seceded.

The check: when two chains diverge, ask which side's software changed and which side kept validating blocks the same way it did yesterday. The party that adopts a new rejection rule is the party that diverges. That's not opinion, it's a git log.

Check two: is silence really a vote?

The second claim making the rounds is that by not signaling, you let the miners choose for you. That you needed to actively signal against BIP-110, and staying quiet was abstention.

Slow down on this one, because it sounds reasonable and it's the most load-bearing thing the argument has.

Declining to adopt a change isn't abstention. It's rejection, and it's the only form rejection takes. There's no "no" button in Bitcoin because you don't need one: not running the new software is the no. The status quo doesn't require a quorum to defend itself. Rules change when nearly everyone opts in and persist when they don't, which makes inaction and refusal the same act.

Now notice the design problem underneath. BIP-110 set a 55% threshold on miner signaling. If your position is that miners shouldn't determine Bitcoin's rules, why build a mechanism that asks them to?

The honest defense is that signaling was never a vote, it's a readiness handshake, so miners were asked to coordinate rather than consent. That's how BIP 9 was always described, and it's fair. But mandatory signaling breaks that defense. The moment "signal readiness" becomes "signal or we eject you," it stops being coordination and becomes a demand. You can't claim you weren't asking permission while treating a non-answer as grounds for excommunication.

After the fork, the argument sharpened into something more principled: proof of work is "completely pointless if the criteria for what constitutes a validity is decided by the ones doing the work." That line comes from an OCEAN-affiliated developer, read approvingly on the channel, and serious people hold the position behind it.

But look at what it's defending. BIP-110 made miner signaling the entire mechanism: 55% of blocks, or no activation. You can't spend eight months asking miners to decide and then, having lost, announce that miners were never entitled to decide.

The check: ask which way the default runs. If a change only happens when people actively adopt it, it needed your yes and never needed your no. Doing nothing already refused it. So when someone says your silence handed the decision to the miners, watch the move: your refusal is being relabeled as an absence, because an absence is easier to blame on somebody else.

Check three: can you audit the number?

Watch which numbers get used to carry the weight.

The pre-fork case leaned on node counts: "almost 17,000 plebs running BIP 110 nodes," about 15.5% of the network by one page, "almost 19%" by another. Then, in the same video, an acknowledgment that the figure had stalled between 15 and 18%, blamed on opponents spinning up fake nodes to match the growth, making the real number "closer to something like 20 or 25%." After the fork, a share of listening nodes in the teens became "broad support among the community."

Node counts come from crawlers that count publicly listening nodes and read a self-reported version string. Anyone with a VPS and an afternoon can add a hundred. Which cuts in every direction at once, and that's the problem. "They're faking theirs" and "we're really at 25%" are the same unfalsifiable move wearing different hats. Neither can be checked. Both were used as evidence.

Compare that to blocks. Fifty-one signaling blocks out of 2,016. About 99.85% of hashpower staying put. Two blocks on the fork in six days. You can verify every one of those from your own node in a minute, and so can the person who disagrees with you. Simple Mining, which mines through OCEAN and still refused to follow, put it better than I can: hashrate is a vote you cannot fake.

The check: ask whether you could disprove the number if it were wrong. Auditable numbers (blocks, hashrate, confirmations) are Bitcoin's native language. Unauditable ones (node counts, "the community," sentiment) are the language of everything Bitcoin was built to replace. When an argument leans on the second kind, that's information.

Check four: was the fallback disclosed as loudly as the pitch?

This is the one that should bother you most, and almost nobody is putting it together.

The sell was modest. A temporary soft fork, one year, expiring on its own. All existing coins grandfathered, so nothing you hold is at risk. "An obvious maintenance soft fork to accept." The channel's own risk video said it plainly: we're looking for a soft fork here, contrasting it with the hard fork that permanently divided Western Christianity.

Then it failed. And within a day the plan was a proof-of-work change, announced by the BIP's own author as a way to "fire the miners." By August 12th the replacement algorithm had a name, BLAKE2b, and a target date of September 1st, delivered as a modification to Bitcoin Knots.

A proof-of-work change does exactly one thing: it makes every SHA-256 machine on Earth worthless on the new chain. Billions in hardware, the accumulated work securing fifteen years of history, the entire industrial base of Bitcoin mining, discarded by design. The channel said so without flinching: it's "definitely a hard fork if you change proof of work."

So the least disruptive option became the most disruptive option, in about four days.

Except it didn't, quite. On August 5th, three days before the fork and while the "temporary, nothing at risk" pitch was still running, Luke Dashjr had already posted that if BIP-110 didn't pan out, "the only option is a PoW change." Developer Chris Guida had rebased Dashjr's 2017-era proof-of-work fork code onto Knots in the first days of the month. The nuclear option wasn't a reaction to defeat. It was loaded before the campaign finished.

And two weeks before the split, Dashjr had posted that "BIP110 cannot cause a chain split." It caused one on schedule, at the block height its own specification named, by the mechanism that specification described.

The check: find out what plan B was, and when it was written. A proposal marketed as low-risk whose fallback is the highest-risk change available to it was never a low-risk proposal. It was a low-risk framing.

Trying to look at all of this fairly

It would be easy to read all of that as "the anti-spam people are idiots." That's not what I think.

The underlying grievance is legitimate. In October 2025, Bitcoin Core v30 raised the default OP_RETURN data limit from 83 bytes to 100,000 and allowed multiple data outputs per transaction. That's relay policy rather than a consensus rule, a distinction that matters enormously and gets muddled constantly. But it was a real decision, made by a small group, that made it materially easier to use Bitcoin's ledger as a file host. If you think Bitcoin is money and only money, you can be sincerely angry about that. Node operators carry the cost of storing other people's JPEGs. Blockspace spent on collectibles isn't spent on payments. Knots went from a rounding error to a peak above a quarter of all nodes (roughly a fifth today) because a lot of people feel this, and they are not stupid.

Here's what complicates it, and it's the strongest thing the other side has. The data was already going in. Miners took large-data transactions directly, through private channels like Marathon's Slipstream and relay networks built to carry what ordinary nodes refused to pass along. Those transactions got mined either way. What the 83-byte default accomplished was hiding them from you. Your node wouldn't relay them, so they never entered your mempool, so your node painted a picture of a network that didn't exist. You'd watch your mempool, then watch a block land full of things your mempool never saw. Gloria Zhao of Chaincode Labs made exactly this case for the change, arguing the old cap pushed users toward "more harmful, unprunable alternatives or direct miner submission."

So raising the default didn't invite the data in. It stopped your node from lying to you about data that was already there. That's why plenty of people who genuinely hate inscriptions still backed the change: a filter that only blinds you isn't protecting anything.

Wherever you land on that, the grievance and the campaign are separate questions. You can hold the first and still notice the four claims above are false. In fact you should, and here's why it matters more to you than to the people you're arguing with. A real grievance argued dishonestly gets buried along with the dishonesty. Every unfalsifiable node count, every retroactive reframe, every "cannot cause a chain split" that causes a chain split spends down the credibility of the concern itself. The spam argument deserved a better campaign than it got.

By August 13th the argument had moved to the character and masculinity of the developers on the other side. That's usually the tell. When a technical dispute becomes a dispute about the moral fiber of the people who disagree, the technical case has generally been exhausted.

Menger's ghost, and why none of this could have worked

There's an older idea underneath this whole episode, and it belongs to Carl Menger, the founder of the Austrian school and the man both Mises and Rothbard built on.

In 1892, Menger asked where money comes from. The intuitive answer, then and now, is that some authority declared it. A king, a council, a treasury. Menger's answer was that nobody declared anything. Money emerged, unplanned, from an enormous number of individuals each independently choosing the most saleable good available to them, until a convergence appeared that no one designed and no one could have imposed. Certain institutions, he argued, are constituted by decentralized acceptance, which means they can't be installed by decree. Try, and you don't get the institution. You get a decree people ignore.

Bitcoin's consensus rules are that kind of object. They aren't legislation. They're the residue of what everyone independently chose to run, which is why they're so absurdly hard to change and why that difficulty is the product rather than a defect in it. Same reason Bitcoin never had an insider round: no pre-mine, no VC seed at a penny, no accredited-investor wall. Nobody was ever positioned to hand it down, so nobody can take it back.

Mandatory signaling was an attempt to shortcut Menger. Declare the rule, refuse to recognize anyone who hasn't adopted it, and hope the refusal manufactures the adoption. It produced exactly what he'd predict: a decree, and a chain with nobody on it, sitting at block 961,633 with a difficulty adjustment 6.3 years away.

What the stuck chain actually proves

If this is the first you've heard of any of it, the takeaway is not "Bitcoin nearly broke."

Nobody could change Bitcoin's rules. Not a developer with fifteen years of commit history and a BIP editor seat (he lost the seat on August 10th, two days after the fork stalled). Not seventeen thousand nodes. Not a mining pool. Not a YouTube channel with a large and devoted audience. The rules held, without a vote, a court, a regulator, or anyone being in charge, because changing them requires nearly everyone to agree and nearly everyone didn't.

The stalled chain at 961,633 isn't a tragedy. It's a receipt.

The trade is honest, though: the same property that stopped this change would stop a good one. Bitcoin is hard to change on purpose, and you don't get to choose which changes it's hard for.

Not your node, not your rules

There's one point both sides agreed on completely, and it's the one worth taking with you.

If you held your own keys, this was a spectator sport. Your private keys sign transactions on any chain that inherits Bitcoin's history, which meant you owned coins on both sides of the split and could safely do absolutely nothing. Sit still. Don't transact. Wait. Total cost to you: zero.

If your coins sat on an exchange, you found out what you actually owned, which was an IOU whose terms somebody else got to interpret. Whether you'd receive anything on the forked chain, and which side you'd be credited with, was a policy decision made in a room you weren't in.

The fork was an argument about who controls Bitcoin's rules. Custody is that same argument, scaled down to you. Running a node is how you enforce the rules you actually consent to. Holding your keys is how you make sure the answer to "what do I own?" doesn't depend on somebody else's weekend.

Not your keys, not your coins. Not your node, not your rules.

At Bitcoin Well, buying Bitcoin sends it straight to your wallet, because the only fork that ever really threatened you is the one where somebody else is holding the coins.

A note on sourcing

BIP-110's specification (the 55% threshold, bit 4, the mandatory signaling window, the one-year expiry, the exemption for pre-activation UTXOs) comes from the BIP text in the bitcoin/bips repository. The block 961,632 timestamp is Blockstream's explorer. Fork timeline, signaling counts, pool attribution, and the difficulty-trap figures are from CoinDesk, Bitcoin Magazine, ForkLog, and Bitcoin.com News, August 8th to 14th, 2026; OCEAN's hashrate figures are from its own web portal. Michael Saylor's hashpower figure is his post of August 9th. Luke Dashjr's "only option is a PoW change" is August 5th, via TFTC; "BIP110 cannot cause a chain split" is his own account, July 26th. His removal as BIP editor followed a motion by Mark "Murch" Erhardt, confirmed by Bryan Bishop on August 10th. The BLAKE2b selection and September 1st target are from Odaily and CryptoPotato; "fire the miners" is the BIP author's post of August 9th. Core v30's datacarrier changes and the relay-versus-consensus distinction are from OAK Research; Gloria Zhao's rationale via The Block; the out-of-band routes (Slipstream, Peter Todd's Libre Relay) from Bitcoin Magazine's technical coverage; Knots node share from coin.dance. Quotes attributed to Bitcoin University are verbatim from that channel's published transcripts, used here for criticism and commentary. Where the channel was reading a third party's written statement aloud, I've said so.

Carl Menger's argument on the spontaneous emergence of money is from On the Origins of Money (1892).