Did BIP-110 actually cause a real chain split?

Yes. On August 8, 2026, at block 961,632, the minority of miners running the activation client rejected the first non-signaling block and split off from the main network. It's exactly the isolation mechanism we'd already described as possible, at minimal scale.

Is there a new coin now?

Not yet, functionally. A contingency plan to change the mining algorithm and effectively spin off a separate coin exists, with September 1, 2026 floated as a target in press coverage, but nothing has shipped in any client. No exchange lists the BIP-110 chain; it has no market and no ticker.

Do I now have 'double' the bitcoin I had before?

Technically, any bitcoin you held before August 8 exists as an identical balance on the minority BIP-110 chain too, same history up to the split. That isn't a bonus. That second chain has no market and carries a real technical risk if you try to move it, which is what this piece is actually about.

What exactly is the replay-attack risk?

Until BIP-110's data restrictions actually activate (block 965,664, which that chain won't reach at its current pace), a signed transaction is valid on both chains simultaneously. Move the 'forked' coins and the same signed transaction can be rebroadcast on the real chain too, moving your real bitcoin without separate authorization.

What should I actually do?

If you don't know how to isolate your coins between the two chains at the UTXO level, the safe move is to not move anything at all. No signed transaction means nothing to replay. This doesn't apply to exchange or ETF holders, and it doesn't require any urgency for most self-custody users either.

A golden 3D Bitcoin coin right at a fork point on a dark circuit board: one path continues as a bright, solid orange beam of light, the other fades into scattered fragments just steps from the split — Original illustration generated with AI (Adobe Firefly) — not representative of real data.

BIP-110's Fork Was Real. Here's the Replay-Attack Risk Self-Custody Users Actually Face

NodeWitness Team August 13, 2026 5 min read

On August 8, 2026, at block 961,632, a minority of miners rejected the first non-signaling block and genuinely split off from Bitcoin's main chain. That chain is already stuck at 2 blocks. The part worth your time isn't the headline, it's what a live replay-attack risk actually means for anyone holding their own keys.

We called this isolation risk "close to nil" five days ago. We got the mechanism right, and underweighted the probability: it only took one large miner staying on the activation client for it to happen, even at trivial scale. Coverage of the headline event has been extensive since; what's been thinner is the practical part.

The correction, stated plainly

Where things stand, verified against the numbers that matter

The BIP-110 chain has been stuck at block 961,633 since August 9, now more than 240 blocks behind mainnet and falling further behind every ten minutes. The arithmetic problem is structural: that chain inherited Bitcoin's full mining difficulty, calibrated for the entire network's hashpower, while running on a sliver of it. The mining effort behind BIP-110 signaling collapsed from roughly 36 EH/s to about 1.25 EH/s once the split occurred, and its next difficulty adjustment (which requires 2,016 of its own blocks) sits years out at the current pace.

Roughnecks briefly paused mining on August 9, calling it "an escalation, not a defeat," and said it would resume the next day. As of this writing, no third block has appeared.

Diagram of the split at block 961,632: real Bitcoin keeps advancing, the BIP-110 chain stalls at block 961,633 with hashrate collapsing from 36 to 1.25 EH/s, and the replay-attack window that stays open until BIP-110's own rules activate — Original diagram built from the verified numbers: the exact split point, the hashrate collapse, and the replay-risk window.

Update, September 5, 2026: the algorithm switch already happened

The contingency plan this section originally described as "a target, not shipped" stopped being that. Unable to compete for blocks against the full network's inherited difficulty, mining operation Roughnecks restarted the stalled chain on August 10 with a "Knots-RDTS" build and switched the mining algorithm. The new algorithm, BLAKE2b, was chosen on August 11 via a deterministic random process over Testnet4 blocks (designed specifically to prevent premining), and the first block under the new rules was mined on August 30, 2026, at block 961,640. The chain, now simply known as "BLAKE2b," has since produced over 800 blocks. BLAKE2b favors CPU and GPU mining, which wipes out the SHA-256 ASIC advantage that mattered on the rest of the network.

For scale, Bitcoin Gold made this exact move (a PoW change after a 2017 fork) and trades around a dollar today. Not a prediction about this chain, just the only real comparison available. For the full breakdown against the unrelated chain that surfaced in parallel (eCash, from Paul Sztorc, no connection to BIP-110 whatsoever), see our dedicated writeup.

The replay-attack mechanics, precisely

This is the part that actually matters if you self-custody, and it's mechanical, not speculative.

BIP-110's data restrictions, the entire point of the proposal, don't activate until block 965,664, which the stalled chain has no realistic path to reaching. Correction (September 5, 2026): replay protection on the BLAKE2b chain exists, but it's opt-in, not automatic. It runs through a signature extension called SIGHASH_UNIFIED (a wallet that implements it, like the dedicated Sparrow build shipped August 31, explicitly marks a send as protected); a signature that doesn't set that flag stays valid on both chains at once.

The concrete risk shows up if you try to separate or sell the "fork" balance: sign and broadcast a transaction to move those coins, and that same signed transaction is simultaneously valid on the real chain. Anyone who obtains it can rebroadcast it there, moving your actual bitcoin without your having authorized that. It's the identical failure mode that hit the 2017 Bitcoin Cash split before dedicated replay protection was added.

That 2017 precedent is worth a beat, because it shows what actually closes this gap and why BIP-110's minority chain doesn't have it yet. Bitcoin Cash's fix was SIGHASH_FORKID, a change to how a transaction's signature hash is computed so a signature valid on one chain becomes mathematically invalid on the other, by construction, with no coordination required from the person signing. BIP-110's current client has shipped no equivalent marker. Until the split chain either adopts something similar or reaches the block where its own data restrictions activate, the two chains keep validating each other's old-format signatures as if nothing had happened, which is exactly the condition a replay attack needs.

The safest instruction, credited to developer Kevin Loaec: if you don't know how to split your coins between chains at the UTXO level, don't move anything. No signed transaction, nothing to replay. Doing this safely requires isolating and spending outputs that exist on only one chain, an operation most wallets don't support yet.

What doesn't require action

If your bitcoin sits on an exchange or in an ETF, none of this touches you, those platforms run on the real chain with the network's full liquidity and hashpower behind it. And if you simply never touch your keys, there's nothing to do either: the replay risk only activates the moment you sign and broadcast a transaction, not from the mere fact that a balance also exists on the minority chain. That's also why there's no clock running on this: a stalled chain with no market and 1.25 EH/s of security doesn't get more dangerous the longer it sits untouched, it only matters the moment someone decides to interact with it.

Would rather check the real chain's state yourself than take a headline's word for it? Network and mining publishes our own node's live status, following the real chain with zero intervention from us, exactly as the overwhelming majority of the network does.

Last updated: September 5, 2026