Who decides the rules

Bitcoin has no chief executive, no board, no shareholders, and no mechanism by which anyone can be made to accept a change. There is no vote that binds anybody. And yet the rules have changed several times, and other attempts to change them have failed decisively.

Understanding how that actually resolves is not political trivia. It is the reason the 21 million limit is credible, and it is the clearest argument for running your own node.

1The rules are whatever nodes enforce

Start from the mechanism rather than the politics. A node holds a set of validity rules and rejects anything violating them. It does not consult anyone. It does not care what a majority thinks. It applies its own rules to every block it receives.

So "the rules of bitcoin" is not a document. It is the overlap between what everyone's software independently enforces. Where that overlap is total, there is one chain. Where it stops being total, there are two — and everyone follows the one their own node accepts.

Which gives the honest answer to who decides: everyone, individually, by choosing what to run — and nobody, collectively, because there is no body to hold the decision. It is less an election than a question of what people will accept.

2Two kinds of change

The distinction matters enormously, because it determines what happens to people who do not upgrade.

Soft forks tighten the rules

A soft fork makes previously valid things invalid. Blocks following the new stricter rules still satisfy the old ones, so nodes that never upgrade continue accepting the chain — they simply do not understand the new features and do not need to.

This backwards compatibility is why almost every successful upgrade has been a soft fork. SegWit in 2017 and Taproot in 2021 both worked this way. Nobody was forced to do anything on a deadline.

Hard forks loosen them

A hard fork makes previously invalid things valid — larger blocks, a different supply, a new opcode requiring relaxed limits. Old nodes reject the new blocks outright, because from where they stand the rules are being broken.

The consequence is stark: everybody must upgrade, or the network splits into two chains. There is no gentle version. This is why hard forks are treated with such suspicion — not because change is forbidden, but because the coordination requirement is absolute.

3The block size war

Between roughly 2015 and 2017 bitcoin had its defining fight, and it is worth knowing because it established the precedent everything since has followed.

The dispute was over the one-megabyte block limit. One side argued for raising it — more transactions per block, lower fees, more usable payments. The other argued that larger blocks make running a node more expensive, and that if ordinary people cannot afford to validate, the property that makes bitcoin work is quietly lost.

Proposal after proposal was made to raise the limit: BIP 101 (Bitcoin XT), BIP 109 (Bitcoin Classic), and several others, all now closed. Later came SegWit2x, an agreement among many large companies and most of the mining industry to activate SegWit and then hard fork to two-megabyte blocks.

By any conventional reading of power, the larger-block side should have won. They had most of the hashpower and most of the major businesses.

They lost, and the reason is the whole lesson.

Miners produce blocks. They do not decide which blocks count. A block nobody will accept is worthless no matter how much electricity produced it.

Node operators, exchanges and users largely declined to run the new software. A movement formed around a user-activated soft fork — BIP 148 — in which nodes would simply begin rejecting blocks that did not signal for SegWit, regardless of what miners preferred. Faced with producing blocks the economy would refuse, the miners moved. SegWit activated. SegWit2x was abandoned weeks before it was due. The larger-block chains that did split away exist, and trade at a small fraction of bitcoin.

4What happened in August 2026

If that history feels distant, a much more recent episode demonstrated the same mechanics in a matter of hours.

The dispute this time was about arbitrary data. Since 2022, inscriptions and token schemes had been using bitcoin's transaction structure to store non-financial data on-chain, which one group regarded as an abuse imposing costs on every node operator and another regarded as a legitimate use of paid-for block space.

BIP 110, the Reduced Data Temporary Softfork, proposed doing something about it. For one year it would have capped most new outputs at 34 bytes, limited OP_RETURN to 83, restricted data pushes to 256 bytes, and disabled several Taproot features used to carry large payloads. Coins created before activation were exempt.

One design choice mattered more than all the technical content: it set an activation threshold of 55% of blocks signalling, where soft forks conventionally use 95%. That difference is not a detail. A 95% threshold is a way of confirming that essentially nobody objects. A 55% threshold is a decision to proceed while a large minority disagrees — which is a recipe for exactly one outcome.

The result was not close:

• Only about 2.5% of blocks signalled support during the window — far below even the reduced threshold the proposal set itself.
• The chain split at block 961,632 on 8 August 2026, when nodes enforcing the new rules rejected a block that did not comply.
• The minority chain produced two blocks in eight hours and then stalled, while the main chain continued at its usual pace and pulled far ahead.
• Roughly 99.85% of hashpower stayed with the original rules. The BIP was marked closed the following day.

• The proposal shipped without replay protection, meaning a transaction broadcast on one chain could be valid on the other. Holders who transacted during the split risked moving coins on both chains unintentionally.
• The standing advice during any contentious split is simple: do not transact until it resolves. Coins sitting still are unaffected by a fork; coins in motion during one are where the accidents happen.
• Your keys cover both chains automatically. There is nothing to claim, migrate, or rescue — which is another argument for holding keys rather than an exchange balance.

Whatever one thinks of the underlying grievance — and it is a real disagreement held sincerely on both sides — the episode is an unusually clean demonstration. A change with a low threshold, insufficient support, and no broad consensus did not force anything. It produced a two-block chain that stopped.

A railway junction photographed from above where one line continues into the distance while a second peels off and ends abruptly in overgrown gravel a short way along.

Image to come

5How successful changes actually happen

Set against those failures, the pattern behind the changes that did succeed is fairly consistent.

• A proposal is published as a BIP and argued over publicly, often for years. Taproot was discussed for roughly four before activating.
• It is made a soft fork wherever possible, so nobody is forced to act on a deadline.
• Objections are addressed rather than outvoted. Since there is no vote, an unresolved technical objection simply remains unresolved, and that is usually fatal.
• Activation is deliberately conservative. High thresholds, long windows, and mechanisms designed to fail quietly rather than split the chain.
• Adoption is gradual afterwards. Years after Taproot, plenty of wallets still default to older address types, and nothing is broken by that.

The system is strongly biased against change, and that is not a defect. For money whose main claim is that its rules cannot be altered to suit whoever currently holds power, difficulty in changing the rules is the product.

6Where you fit

"Running a node is voting" is a slogan that overstates things, so here is the precise version.

Your node does not cast a ballot. It enforces the rules you chose when you decided which software to run, and it independently refuses anything violating them. Multiply that by everyone doing the same and you get the only thing that has ever actually decided a bitcoin rule dispute: what the economy will accept.

If you hold coins through an exchange, your position in that is held by the exchange. If you hold your own keys and validate with your own node, it is held by you. That is the entire mechanism, and it is not a metaphor.

The short version

There is no authority. Rules are whatever nodes independently enforce, so change requires persuading people to run different software rather than winning a vote. Soft forks tighten rules and stay compatible; hard forks loosen them and split the network unless everyone moves. The block size war established that hashpower does not decide, and BIP 110 demonstrated the same thing in 2026 in under a day.