A Bitcoin fork never destroys your coins on its own. What costs people money is what they do with those coins while the network is split. This guide covers the real difference between a soft fork and a hard fork, how a replay attack empties a wallet, and the single rule that protects a non-technical holder.
Key Takeaways
- A soft fork tightens the rules, a hard fork creates new ones that old nodes reject outright
- Without replay protection, selling coins on one chain can spend your bitcoin on the other
- Coins that never move cannot be replayed, because no signed transaction exists to copy
Soft Fork or Hard Fork: The Only Distinction That Matters
A fork is a change to the network’s validation rules. Everything else follows from one question: do nodes that skipped the upgrade still accept blocks produced by nodes that took it?
In a soft fork, the answer is yes. The new rules are stricter than the old ones, so a block that passes for upgraded nodes still passes for everyone else. Backward compatibility holds, and the network can adopt the rule gradually without splitting.
In a hard fork, the answer is no. The new rules allow blocks that old nodes reject. From that block onward two chains coexist, sharing one history up to the split point and nothing after it. A hard fork mints a new asset, a soft fork does not.
The distinction matters because a poorly supported soft fork can still produce a split. If a minority of nodes enforces a rule the mining majority ignores, those nodes start rejecting the main chain’s blocks. Technically it remains a soft fork. Practically, the user is looking at two networks.
That is precisely the setup created by the BIP-110 proposal as published in the official BIP repository in August 2026, and the reason this deserves a guide rather than a news item.
What BIP-110 Actually Changes in the Consensus Rules
BIP-110 carries the official title Reduced Data Temporary Softfork. It is a consensus-layer proposal authored by Dathon Ohm, designed to temporarily restrict how much non-payment data can be embedded in a Bitcoin transaction.
Seven rules make up the package. New scriptPubKeys are capped at 34 bytes, with 83 tolerated for OP_RETURN. Data pushes and witness items are limited to 256 bytes. Spending an undefined witness version becomes invalid, Taproot annexes are prohibited, control blocks are capped at 257 bytes, and two opcode families (OP_SUCCESS along with OP_IF and OP_NOTIF) are barred from executing in Tapscript.
One clause settles the biggest worry immediately. UTXOs created before activation are explicitly exempt from every restriction. No existing balance gets frozen by the rule itself. The limits apply to what gets written afterwards, not to what already sits in the chain.
Activation runs on a modified BIP9 deployment with a 55% threshold, meaning 1,109 signalling blocks inside a 2,016-block period. The clock started on December 1, 2025. Mandatory signalling covers blocks 961,632 through 963,647, and the maximum activation height is 965,664, roughly September 1, 2026. Once active the rule runs for 52,416 blocks, about a year, then expires on its own.
The trouble sits elsewhere. Miner support has never cleared roughly 1% since signalling began in spring 2026, and sat near 2.6% heading into the deadline. Bitcoin Core has not merged the proposal. Live signalling rates are published on the monitor built for the BIP-110 deployment. A 55% threshold against 2.6% support is not a gap that closes in days, especially with hashrate operators busy elsewhere, between margin pressure and a growing share of the fleet pivoting toward AI workloads.
Also on Cryptonomic:
- Bybit Sues North Korea Over Its $1.5B Hack
- MetaMask Opens Its Wallet to Autonomous AI Agents
- Ondo Founder’s Mother Sues to Take Back Control
Replay Attacks Are Where Holders Actually Lose Money
Here is the mechanism that costs real money. At the moment of a split, every holder owns the same balance on both chains, because both share the history up to the separation point. Two identical balances, one private key.
A transaction signed to spend coins on one chain is just a file. Nothing stops anyone from copying it and broadcasting it on the other chain, where it is equally valid because the signature rules are identical. One signature spends both balances. That is a replay attack.
In practice, a holder who sells forked coins on a venue can watch the buyer, or any observer, replay that transaction on the main network and take the real bitcoin too. Selling a secondary asset drains the primary wallet. The attacker needs no exploit and no privileged access, only a copy of a transaction that was already broadcast in public.
BIP-110 ships without replay protection, and that is not an oversight. Replay protection is what you build when you intend to separate two chains cleanly. BIP-110 was never designed to split Bitcoin, so nothing was provisioned for a split happening anyway. Developer Kevin Loaec flagged the risk publicly on August 6, 2026.
The survival rule fits in one sentence. Coins that do not move cannot be replayed, because no signed transaction exists to copy. For a non-technical holder, doing nothing at all is the safest posture during a potential split. The same discipline applies to custody itself, a trade-off covered in our comparison of cold wallets versus hot wallets in 2026.
What 2017 Taught the Network and What Is Still Open
Bitcoin has run this exercise before. On August 1, 2017, Bitcoin Cash separated from the main network by lifting the block size from 1MB to 8MB, as the scaling fight came to a head. The split was clean, because Bitcoin Cash implemented replay protection.
The counterexample arrived the same year. SegWit2x backers refused to add that protection despite repeated calls from the ecosystem. The project was dropped days before its scheduled hard fork, partly because exchanges and miners would not carry the replay risk on behalf of their users.
Both episodes point at the same lesson. Replay protection is not a technical nicety argued over by developers, it is the line between a split that costs users nothing and one that empties wallets. A fork that ships without it is telling holders, whether it means to or not, that the safe move is to sit still until the situation resolves.
A second August 2026 deadline sits alongside BIP-110 and has nothing to do with it. Paul Sztorc’s eCash hard fork targets block 964,000, around August 21, and would credit every holder an equivalent balance on the new chain. It introduces the drivechains described in BIP-300 and BIP-301, with seven sidechains already in development.
That project carries its own controversy. It proposes reassigning the coins matching Satoshi Nakamoto’s 1.1 million bitcoin to reward early investors, which a large part of the community calls theft outright. The precedent worries people more than the amount.
For BIP-110, the likeliest outcome remains no lasting fork: either enforcing nodes revert, or a minority chain survives with thin hashpower and uncertain market infrastructure. Replay protection only kicks in from block 965,664 in early September. The risk window sits between those two dates, which is exactly when a holder is better off leaving coins untouched, a lesson the Coldcard flaw and its $70M drain in 41 minutes delivered on different ground.
Frequently Asked Questions
Will BIP-110 actually fork Bitcoin?
Nothing is settled. The rule requires 55% of blocks signalling, meaning 1,109 out of 2,016, and miner support tops out near 2.6%. Bitcoin Core has not merged the proposal. A lasting fork would require enforcing nodes to hold their position against a majority chain that ignores them, which at best produces a minority chain with very little hashpower.
Can BIP-110 freeze my bitcoin?
No. The proposal explicitly exempts every UTXO created before activation. The restrictions apply to new data embedding, not to funds already recorded in the chain. No existing balance becomes unspendable because of the rule itself.
What should I actually do if the chain splits?
Sign nothing. As long as no transaction is broadcast from your addresses, no signature exists to be replayed on the other chain. Waiting for exchanges to publish their policy and for replay protection to take effect costs a few days of inactivity, which beats a permanent loss.
Do I automatically receive the forked coins?
It depends on the fork type and on how you hold. On a hard fork such as eCash, a self-custodying holder controls the equivalent balance on the new chain. On funds deposited with a platform, the operator decides whether to credit the asset, and that decision gets announced case by case.
Test Yourself
Can a soft fork still split the chain?
Show answer
Yes. If a minority of nodes enforces a rule the mining majority ignores, those nodes reject the main chain’s blocks and end up on a separate chain, despite the soft fork’s theoretical backward compatibility.
What signalling threshold does BIP-110 require?
Show answer
55%, meaning 1,109 signalling blocks within a 2,016-block period. Observed miner support sat around 2.6% heading into the mandatory signalling window.
Why are unmoved coins safe from replay?
Show answer
Because a replay attack works by copying an already-signed transaction onto the other chain. With no signed transaction, there is nothing to copy and nothing to replay.
Follow the story on Cryptonomic.
