BIP-110's 185-Block Cliff: Bitcoin's First Non-Consensus Soft Fork Is Already Here

Bitcoin has 185 blocks left before a rule change that most of the network never signalled for begins rejecting otherwise valid blocks. As of block 961,632, any node running Bitcoin Knots will refuse to extend a chain containing a block without version bit 4 set. The signal support stands at 2.62%. No 95% threshold. No lock-in. No broad community consent. We didn't see this coming because everyone was watching ETF flows instead of version bits. But the countdown is real, and the split isn't hypothetical — it's already encoded in the software. BIP-110 is a soft fork proposal aimed at shrinking block data and improving SPV verification efficiency. The concept isn't new; it inherits a line of thinking from the 2015–2017 blocksize war SPV proposals. What's new is the activation mechanism. Standard BIP-9 activation requires 95% of hashrate signalling over a 2,016-block period. BIP-110 lowers that to 55% — 1,109 blocks out of 2,016 — but then adds a twist no prior deployment had: the forced signalling period starts at block 961,632 regardless of whether the threshold has been reached. If the fork doesn't lock in, the data-reduction rules are still planned to activate at block 965,664. In other words, this is not a miner referendum. It's a unilateral implementation timeline. Bitcoin's governance evolution has never been clean, but it has usually respected a divide: miners signal, nodes enforce, and if a change lacks consensus, it doesn't activate. BIP-110 breaks that pattern. Bitcoin Knots, the alternative node implementation, will enforce BIP-110. Bitcoin Core, the dominant implementation, will not. The Core PR for BIP-110 was closed unmerged on March 26, 2025, and Core contributor Antoine Poinsot said on June 4, speaking personally, that Core does not implement the proposal. So after block 961,632, two sets of nodes will use different rules to judge the same block. That is not a soft fork in any traditional sense. It is an implementation-level schism. Let me be precise about what happens mechanically. After 961,632, a miner produces a block without bit 4 set. Bitcoin Core sees a block that satisfies all existing consensus rules and accepts it. Bitcoin Knots sees an invalid block and refuses to extend that chain. Knots nodes either wait for a compliant block or, if none arrives, stall. If OCEAN's miners — the pool that switched its default endpoint to a BIP-110-compliant one on July 15 — keep producing bit-4 blocks, Knots can continue building its own chain. That chain would be transparent to Core nodes, and Core nodes would not acknowledge it as part of the main chain if it lacks sufficient work. The result is not necessarily a persistent fork with two marketable tokens. It can be a temporary divergence, a stalled node population, or a quiet reorg. Based on my years auditing exchange node infrastructure, I can tell you where the real damage lands: not in the elegant consensus layer, but in the ugly state handover. The Knots team itself warned on August 7 that old non-enforcing software, including Bitcoin Core, may not fully validate BIP-110 rules and could leave an unsafe chain state. BlockSlop's regtest reproduction found something more worrying than a classic fork. When you switch from a Knots node that enforces BIP-110 back to a non-enforcing Knots or Core software, the data directory still contains blocks that were accepted under the old rules. On a normal start, the node doesn't immediately reconnect the full inherited history, so it briefly sits in a rule-inconsistent state. The Knots team has since merged a safety measure that scans inherited block headers for forced-signal violations and invalidates offending blocks. But that scan only sees headers. Transaction-level or script-level violations that aren't visible in the header still require reconnecting validation, and possibly a full reindex. That's the kind of bug that takes down an indexer, delays a withdrawal, or creates a false confirmation in a block explorer — not a 'Bitcoin is dead' narrative, but a slow bleed of trust. The market's read on this has been casual. With 2.62% signal support, most participants assume BIP-110 is irrational noise from a fringe implementation. That's partly true. But the lazy narrative that 'a fork just means free coins' misses the structural point. This isn't BCH 2017, where a large miner- and exchange-backed faction split intentionally. This is a minority node implementation drawing a line in the sand and forcing the network to react. If the divergence persists, Bitcoin's core assumption — one chain, one consensus rule — becomes conditional. Every exchange, every custody provider, every Lightning node operator will have to ask a question they've never had to ask before: which implementation's view do I trust? That question, once asked, doesn't go away. It becomes a checklist item, a compliance inquiry, a basis for risk off. Then there is the precedent problem. BIP-110 may go down as the first 'non-consensus soft fork' that succeeded in fragmenting rule enforcement without winning hashrate support. The potential damage to Bitcoin's decentralization story is not a chain split; it's the demonstration that a small but determined implementation can create enough ambiguity to harm users. If any state actor with a regulatory axe to grind is looking for evidence that Bitcoin's infrastructure isn't robust, this event writes the script for them. I won't put a probability on that scenario; the market impact is more direct through infrastructure failures. So forget the price chart for a moment. The next blocks to watch are 961,632 and 965,664. If OCEAN mines a bit-4 block immediately after the forced period, you'll see a live test of Bitcoin's multi-implementation philosophy. If no compliant block appears, watch how many Knots nodes go quiet. And if you run a service that touches Bitcoin data, now is the time to audit your node stack and understand which rule set it trusts on startup. The rhetorical question I keep asking myself: can a protocol that prides itself on adversarial tolerance survive the one thing it never designed for — an adversarial node implementation?