Taproot Was the Last Fork: Bitcoin’s Upgrade Crisis in 2026

On 14 November 2021, at block 709,632, Bitcoin switched on Taproot with almost no drama. The upgrade bundled Schnorr signatures, a smarter way to hide unused spending conditions, and a more flexible scripting language, and it did so with the broadest miner agreement the network had ever recorded. Nearly five years later, Taproot holds a stranger distinction: it is the last time Bitcoin changed its own consensus rules and made the change stick.

That should be uncomfortable for anyone who thinks of Bitcoin as living software. Since Taproot locked in, a queue of finished or nearly finished proposals has piled up behind it: a package of security bug fixes, a family of covenant opcodes that would finally give Bitcoin real programmability, and an entire post-quantum migration plan. None of them has activated. In August 2026 a contested soft fork called BIP-110 tried to force the issue and split off a ghost chain that mined two blocks before stalling. The real story of Bitcoin this year is not what developers want to build on top of Taproot. It is whether the network can still agree to change itself at all.

The Upgrade That Stuck, and Everything That Hasn’t

Taproot was not one change but three, shipped together as BIP 340 (Schnorr signatures), BIP 341 (the Taproot output type and its Merkle-tree script commitments), and BIP 342 (Tapscript, the upgraded scripting language). Together they made complex spending conditions cheaper, more private, and harder to distinguish from an ordinary payment. Almost everything interesting that has happened on Bitcoin since, from Ordinals and Runes to the covenant experiments and the new generation of layer-two protocols, runs on foundations that Taproot laid.

And yet the foundation is where the building stopped. As of mid-September 2026, with Bitcoin trading near $78,000, roughly 38 percent below its October 2025 record of $126,080 according to CoinGecko, the network has not enforced a single new consensus rule in almost five years. This is not for lack of ideas. The proposals exist, the code exists, and in several cases the specifications are finished. What is missing is the one thing Bitcoin has never had a reliable formula for: a way to say yes. This November marks Taproot’s fifth anniversary, and the most honest way to celebrate it is to admit that nobody is sure the network could repeat the feat today.

There is an irony worth sitting with. The most visible use of Taproot has not been the private, efficient multisig its designers imagined, but Ordinals and Runes, which exploited Tapscript’s roomier witness data and Bitcoin’s witness discount to stamp images and tokens onto the chain. Taproot briefly became a large share of all Bitcoin transactions in 2024 on that inscription wave, then settled back as the speculation cooled. The upgrade delivered, just not in the way anyone expected, which is a fitting overture for a network that can no longer predict or direct its own evolution.

How Taproot Actually Got Activated

Taproot activated through a mechanism called Speedy Trial, and the name was the whole point. Rather than reopen the trench warfare of the 2017 scaling fight, developers set a short, low-stakes signaling period. Miners were given a roughly three-month window to signal readiness in their blocks. If 90 percent of blocks in a single 2,016-block difficulty period signaled, the change would lock in and then activate on a fixed delay. If they did not, the attempt would simply expire with no lasting damage.

It worked on the first try. Signaling crossed the threshold, Taproot locked in at block 687,284 on 12 June 2021, and it activated on schedule at block 709,632 that November, as CoinDesk reported at the time. The quiet was possible because Taproot was genuinely uncontroversial: it made Bitcoin better without taking anything away from anyone, and no meaningful faction opposed it. Speedy Trial did not resolve the deeper question of who gets to approve a change to Bitcoin. It simply found a case where the question did not need answering. The proposals waiting behind Taproot do not have that luxury.

Even Speedy Trial was born from deadlock. Through 2020 and early 2021, developers split bitterly over whether Taproot should activate with the mandatory LOT=true setting, which forces activation on a timeout, or the softer LOT=false. Neither camp could convince the other, and the stalemate threatened to strand a change almost nobody actually opposed. Speedy Trial was the escape hatch: a deliberately minimal, low-commitment test that sidestepped the argument entirely by making failure cheap. It succeeded because it lowered the stakes, not because it settled the underlying question of who holds the authority to say yes. That question was merely postponed, and it is now overdue.

A Short, Contentious History of Bitcoin Soft Forks

Bitcoin has changed its rules many times, and the way it does so has evolved with every scare. The very first soft fork, in November 2009, was a hardcoded activation at block 31,000, decided by Satoshi Nakamoto alone. By 2012, Pay-to-Script-Hash (BIP 16) needed a crude miner poll and still caused a brief chain fork before activating on 1 April that year. Then came IsSuperMajority, which watched block version numbers and flipped enforcement on at 95 percent, per the history maintained by Bitcoin Optech.

That approach nearly caused a disaster. In July 2015 the strict-signature soft fork BIP 66 reached its 95 percent threshold at block 363,725, but a miner that had signaled readiness without actually validating blocks produced an invalid chain six blocks long, briefly giving lightweight wallets false confirmations and costing miners tens of thousands of dollars. The lesson was that miner signaling measures intent, not competence. Each mechanism since has been an attempt to buy safety with more structure.

The throughline across two decades is a slow transfer of trust away from any single decider. Satoshi could set a flag day alone; by 2016, BIP 9 tried to make miners the sensor for readiness without making them the authority; by 2017, users demonstrated they could overrule miners outright. Each step added safety rails after a near-miss, and each added friction. The mechanisms grew more elaborate precisely because the community grew less willing to let any one group, whether developers, miners, or a subset of users, unilaterally decide what Bitcoin becomes.

The 2017 Ghost Still Runs the Room

No account of Bitcoin governance is complete without the SegWit fight, because every activation argument since has been fought in its shadow. Segregated Witness deployed with BIP 9 parameters in 2016 and stalled well below the 95 percent miner threshold, as large pools withheld support to extract concessions on block size. The deadlock broke through raw user pressure. A group of node operators adopted BIP 148, a User-Activated Soft Fork that promised to reject any block not signaling SegWit after a fixed date, forcing miners to choose between compliance and a chain split.

Miners blinked. A compromise called BIP 91 lowered the effective threshold to 80 percent over compressed 336-block windows, SegWit locked in, and the mandatory user enforcement expired without a split. Bitcoin Optech notes drily that the exact contribution of each maneuver to the outcome remains a subject of debate. What is not debated is the precedent: users, not miners, hold the ultimate veto, and they proved willing to use it. That memory is why miner signaling today is treated as a convenience rather than a source of authority, and why any proposal that looks like it is trying to ram a change through, or to block one, gets measured against 2017.

The fight did not end cleanly, either. The same period produced the New York Agreement, a closed-door deal among companies to pair SegWit with a later block-size increase called SegWit2x, and it produced a permanent schism: on 1 August 2017 a faction split off to create Bitcoin Cash, a separate chain with bigger blocks. SegWit2x itself was called off that November for lack of consensus. The lessons still shape 2026 instincts: back-room agreements among businesses draw suspicion, hard forks that split the chain are treated as failures rather than options, and anyone proposing a change learns to expect a fight even when they are technically right.

BIP-110 and the Chain Split That Fizzled

Which brings us to August 2026 and the most recent attempt to change Bitcoin, which failed so completely that it became an argument all by itself. BIP-110 was a temporary soft fork designed to restrict arbitrary data, the kind that Ordinals and Runes pack into transactions. Its backers borrowed the UASF playbook but set the mandatory signaling threshold at just 55 percent. When the signaling window opened at block 961,632 on 8 August, support peaked at 2.53 percent of blocks, according to The Cryptonomist. A minority chain enforcing the new rules did split off, mined two blocks, and then stalled with a rounding error of the network’s hashpower before the main chain rolled on as if nothing had happened.

Jameson Lopp, co-founder of the custody firm Casa, had warned exactly this in a widely-read essay, calling the design a misguided, risky overreach and noting that its low threshold greatly increases the chances of a chain split. His broader point outlasted the specific proposal: “Bitcoin is not a democracy whereby a simple majority can overwhelm the minority,” he wrote, arguing that changes “need broad buy-in, not node-counting games.” Blockstream chief executive Adam Back was blunter, telling Bitcoin.com News that the proposal “doesn’t have tech nor ecosystem consensus” and that “this is not the way.” BIP-110 proved a network can refuse a change decisively. It said nothing about how the network could ever accept one.

The details of the failure were almost comic. Essentially all of BIP-110’s meager signaling came from a single pool, Ocean, which had switched its default to signal before reverting once the fork stalled; no other major pool joined. A messier hard-fork cousin, an opt-in chain split that would have reassigned long-dormant coins, loomed a couple of weeks later and drew even less support. Michael Saylor, executive chairman of Strategy, read the episode as vindication rather than crisis, posting that “Bitcoin worked exactly as designed” because “BIP-110 was free to fork, and the network was free not to follow,” and adding that “consensus is earned, not declared,” in remarks reported by Crypto Times. He is right that the immune system worked. The open question is whether the same immune system now rejects healthy tissue.

The Queue Behind Taproot

Strip away the fights and a simple fact remains: Bitcoin has a backlog. Three distinct bodies of work are ready or close to ready, and each is blocked less by its own merits than by the absence of a way to activate anything. The first is defensive, a set of narrow bug fixes called Consensus Cleanup. The second is expansive, the covenant proposals that would let Bitcoin script constrain how coins are spent in the future. The third is existential, the post-quantum address types meant to protect coins before a future quantum computer can forge Schnorr signatures.

These are not equivalent in urgency or in controversy, and lumping them together is itself part of the problem. But viewed as a queue, they show how much has accumulated in one activation-free half-decade.

The quantum row deserves a closer look, because it is the item where doing nothing is most obviously dangerous. Taproot addresses reveal their public key on the chain the moment they are funded, which means that if a large enough quantum computer ever arrives, every bc1p output is exposed from birth rather than only at the moment it is spent. Research disclosed by Google’s quantum team in 2026 lowered the estimated resources such an attack would need, and a large share of all bitcoin already sits at addresses that have exposed their public key. The proposed answers, a quantum-safe address type and a phased freeze of legacy coins described at bip360.org, are among the most complex changes ever contemplated for Bitcoin, and they cannot even begin until the network can agree to activate a far simpler one first.

Consensus Cleanup: The Least Controversial Fork That Still Can’t Ship

If any proposal should be easy, it is Consensus Cleanup. First floated by developer Matt Corallo in 2019 as the Great Consensus Cleanup, revived in late 2023, and now formalized as BIP-54, it adds no features and asks for no new capabilities. It fixes four longstanding weaknesses, described in detail by Bitcoin Optech: the timewarp bug, which lets a majority-hashrate attacker manipulate difficulty by lying about timestamps; blocks that are pathologically slow to validate, addressed by capping legacy signature operations; a Merkle-tree quirk that lets a 64-byte transaction be confused with an internal tree node and fool lightweight wallets; and a duplicate-transaction edge case, fixed by binding coinbase transactions to their block height.

These are the kind of latent flaws that stay theoretical until they do not. The specification and a reference implementation now live in the Bitcoin repository. Antoine Poinsot of Chaincode Labs, the proposal’s lead author, described the work plainly on X: he submitted “a Bitcoin Core pull request implementing the BIP 54 soft fork (Consensus Cleanup), without mainnet activation.” That last clause is the entire drama of 2026 compressed into three words. The code is ready. The switch is deliberately left disconnected, because nobody can agree on how to throw it.

None of these bugs is a five-alarm fire today. The timewarp attack requires a hostile majority of hashrate, the slow-validation blocks require an attacker willing to burn money to make the point, and the duplicate-transaction case is already defended against in practice. That is exactly what makes Consensus Cleanup such a clean test of Bitcoin’s machinery: it is the boring, responsible maintenance that a healthy engineering culture ships without drama. The proposal has been through years of review, carries testnet activation parameters, and enjoys about as much developer goodwill as anything in the queue. If the network cannot activate even this, the problem is not the merits of any single change. It is the process itself.

The F2Pool Objection

Even a bug fix has opponents, and the most consequential one is a mining pool. Wang Chun, co-founder of F2Pool, has publicly opposed forcing through both BIP-110 and BIP-54. As summarized by KuCoin, he likened packaging four unrelated changes into one soft fork to the bundled-legislation tactic that politicians use to pass unpopular measures, and argued that most of the risks are too remote to justify a coordinated rule change. The timewarp attack, he noted, offers miners no real benefit, and validation speed has already improved through better software and hardware.

Crucially, Wang Chun has said F2Pool would update its nodes if BIP-54 reached the required majority through a standard signaling process, but that the pool will not signal for it beforehand, per Bitcoin.com News. On its face this is reasonable. In practice, when a single large pool refuses to be the first mover, a BIP-9 style process that needs a supermajority of signaling blocks can stall indefinitely, because everyone waits for everyone else. The F2Pool position is a live demonstration that Bitcoin’s activation problem is not really technical. It is a coordination game in which any large player can sit on its hands and call it prudence.

It is worth being fair to the miners’ position, because it is not pure obstruction. Pools answer to customers and to their own risk teams, and signaling for a change they consider unnecessary carries operational cost and reputational exposure if it goes wrong. The deeper backdrop is a node-software culture that has fractured over unrelated fights, most visibly the running argument over how much arbitrary data belongs in blocks, which pushed a meaningful minority of operators onto the alternative Bitcoin Knots client. When the people who run the network cannot agree on what Bitcoin is for, expecting them to agree on how to change it starts to look optimistic.

The Covenant Deadlock

The covenant proposals are where the stakes get higher and the disagreement gets real. Covenants let a Bitcoin output restrict how it can be spent next, which sounds abstract until you list what it enables: vaults that force a time delay and a second approval before coins can move, congestion-control constructions, better payment pools, and cheaper, more scalable layer-two protocols. A covenant-enforced vault can even freeze an attempted theft mid-flight, the kind of last line of defense that was missing in 2026’s largest key heists. The same primitives would let a hardware signer enforce spending policy in Bitcoin’s own consensus rules rather than in fragile wallet software.

The developer frontrunner, per a survey of proposals by Spark, is the pairing of OP_CTV (template covenants) with OP_CSFS (signature checks over arbitrary data), sometimes bundled as LNHANCE alongside a helper opcode. An activation client for OP_CTV exists and specifies a Speedy Trial that times out in March 2027, but miner signaling has hovered near zero all year. A rival camp, including Blockstream’s Andrew Poelstra, prefers the broader OP_CAT, which would enable a wider design space at the cost of recursive covenants that some fear are too powerful. The result is a standoff: the safe option cannot get signaling, the powerful option cannot get consensus, and the tools that would let non-custodial desks lend against bitcoin more safely stay just out of reach.

To see why this is more than an academic loss, picture the vault that covenants would make possible. A saver could lock coins so that any withdrawal must first sit in a public waiting room for a day, during which a thief’s attempted spend is visible and can be canceled with a pre-authorized recovery key. Covenants would also let the newer scaling protocols, the shared-UTXO systems that let thousands of users transact off-chain, run their rounds without every participant needing to be online to co-sign. Those systems exist today, but they lean on interactivity and trust assumptions that a single opcode like OP_CTV would relax. The programmability is designed and demonstrated on test networks. It waits on the same locked door as everything else.

The Mechanisms on the Table

Underneath every one of these fights is the same unresolved question: what mechanism should decide? Bitcoin has a menu of options, each with a different answer to the question of who holds the final say, and none commands universal trust. The choice is not merely procedural. Picking a mechanism is picking who wins when the network disagrees, which is why the mechanism debate is often more heated than the proposals themselves.

Speedy Trial worked for Taproot precisely because there was no objection to overrule. For a contested change, it collapses back into the older dilemma: either let miners hold a veto through BIP-9, or hand enforcement to users through a UASF and accept the split risk that BIP-110 just demonstrated in miniature. There is no neutral choice, and the search for one is a large part of why nothing moves.

History does offer one relief valve. When a genuine emergency has struck, such as the 2010 bug that briefly let someone conjure billions of bitcoin from nothing, developers have shipped a fix within hours and the network has adopted it almost unanimously, no signaling ceremony required. Crisis manufactures the consensus that committees cannot. The uncomfortable implication is that Bitcoin may be structurally better at responding to disasters than at preventing them, which is a fine trait for a fire department and a worrying one for a monetary network staring at a slow-moving threat like quantum computing.

What Ossification Actually Costs

To one camp, this is not a crisis at all. A Bitcoin that cannot easily change is a Bitcoin that cannot easily be corrupted, captured, or inflated, and its stubbornness is exactly what makes it credible as long-term money. Ossification, in this reading, is the network growing up. The rules that matter are already in place, and the burden of proof for touching them should be close to impossible to meet. From that seat, five years without a fork is a feature, and the pressure to ship is the thing to resist.

The counterargument is that reality does not wait for consensus. Latent bugs stay unfixed. Programmability that users clearly want gets exported to other layers: much of Bitcoin’s smart-contract activity now lives one level up, in sidechains and rollups like Stacks and Rootstock, and in fraud-proof systems like BitVM that were engineered specifically to avoid needing a soft fork. That is ossification working as an escape valve, and it has costs in trust assumptions and fragmentation. The sharpest version of the argument is the quantum clock. If a future quantum computer can eventually forge signatures, then a network that cannot coordinate a migration is not conservatively preserving value; it is standing still in front of a moving train.

The escape valve is real, but it is not free, and 2026 exposed its limits. The Bitcoin layer-two sector went through a brutal contraction this year, with total value locked across Bitcoin L2s falling sharply and a string of projects shutting down in what analysts described as a dot-com-style shakeout. Routing programmability off the base layer means inheriting bridges, operators, and trust assumptions the base layer was designed to avoid, and users have learned the hard way that not every Bitcoin sidechain is as safe as Bitcoin itself. Every year the core rules stay frozen, more of Bitcoin’s real-world utility migrates to venues that carry more risk, a strange result for changes so often justified in the name of caution.

No Regulator Can Break the Tie

American readers sometimes assume that a question this consequential must ultimately land on a regulator’s desk. It does not. The Securities and Exchange Commission neither approves nor blocks a Bitcoin soft fork, because the consensus layer of a decentralized protocol is not a security, an exchange, or a registrant. There is no filing to make and no permission to grant. If every developer, miner, and node operator agreed tomorrow to activate covenants, no US agency could stop them, and if they never agree, no agency can force them.

What regulation does touch is everything downstream. Spot Bitcoin exchange-traded funds must publish how they would handle a fork, which gives large custodians a quiet stake in stability. The Internal Revenue Service has treated forked or airdropped coins as ordinary income at the moment of receipt since its 2019 guidance, so a messy chain split creates real tax questions for US holders. Moving your own coins between address types, by contrast, is not a taxable disposal. The net effect is that Washington shapes the consequences of a fork without any vote on whether it happens. The deadlock is Bitcoin’s to resolve, and Bitcoin’s alone.

That downstream influence is not trivial, though. The spot Bitcoin funds now hold a large slice of circulating supply, and their custodians and sponsors have powerful commercial reasons to prefer that nothing about the underlying asset changes unpredictably. A contested fork that produced two competing chains would force every fund issuer, exchange, and tax preparer in the country to decide which chain is the real bitcoin and to defend that choice to regulators and customers. In a market this institutionalized, the mere threat of a chain split becomes a lobbying force for the status quo, one more heavy weight on the side of never changing anything.

The Ethereum Contrast: Two Ways to Change a Blockchain

It is tempting to look at Ethereum, which ships coordinated hard forks on a roughly annual cadence, and conclude that Bitcoin is simply broken. Ethereum has a foundation, a published roadmap, a client-team culture that expects node operators to upgrade, and a social contract that treats scheduled change as normal. Bitcoin has none of these by design. It has no roadmap, no foundation with authority over consensus, and a culture that reads a contested change as a reason to wait rather than a problem to schedule around.

Each model pays for its strengths. Ethereum’s coordination buys speed but concentrates influence, and even there the promised endgame slips: its own account-abstraction roadmap keeps sliding years past its early timelines despite top-down direction. Bitcoin’s diffusion of power buys credible neutrality, the assurance that no small group can rewrite the rules, but it pays in paralysis. The two networks are not competing to solve the same problem. Bitcoin optimizes for the guarantee that it will not change against your wishes, and the price of that guarantee is that it may not change in time to meet your needs either.

The contrast is easiest to see in tempo. Ethereum shipped its Pectra upgrade in 2025 and lined up its next hard fork behind it, treating a yearly rules change as routine plumbing that node operators simply install. Bitcoin has shipped nothing in the same span and cannot say when it next will. Neither cadence is self-evidently correct. A network that changes every year can drift, accrete complexity, and concentrate the power to steer it; a network that changes never can rot quietly around its unfixed edges. Bitcoin bet its entire identity on the second failure mode being safer than the first, and 2026 is the year that bet is being tested in public.

Where the Deadlock Breaks

There are a few ways out of the current freeze, and the next twelve months should reveal which one Bitcoin is on. The cleanest is Consensus Cleanup becoming a genuine test case: if a feature-free bug fix with a named author and finished code cannot find an activation path, the community will have to admit, out loud, that it currently has no working mechanism at all, and that admission may be what finally forces agreement on one. A second path runs through the covenant camp, where the CTV plus CSFS bundle could get a real Speedy Trial attempt before its 2027 timeout if enough businesses and developers coalesce. A third, more dramatic path is a UASF revival, most plausibly if the quantum threat sharpens enough to convert a slow-moving risk into an emergency.

The likeliest near-term outcome, though, is continued stalemate, punctuated by more proposals that ship as code and wait. That is the paradox Taproot leaves behind on its fifth anniversary. The upgrade succeeded so completely, and so quietly, that it made the next one look easy, when in fact it was the exception. The question is no longer which improvement Bitcoin will choose. It is whether a network deliberately built to resist change can still summon the collective yes that Taproot got, one last easy time, in November 2021.

Frequently Asked Questions

Was Taproot Bitcoin’s last soft fork?

Yes. Taproot activated in November 2021 and remains the most recent change to Bitcoin’s consensus rules as of September 2026. Every proposal since, from the Consensus Cleanup bug fixes to covenant opcodes and post-quantum address types, is either stuck in debate or waiting for an activation path the network has not agreed on.

Why can’t Bitcoin just upgrade like Ethereum does?

Ethereum coordinates upgrades through a foundation, a published roadmap, and scheduled hard forks that node operators are expected to follow. Bitcoin has no central roadmap and treats a contested change as a reason to wait. Activation depends on rough consensus among developers, miners, businesses, and users, and any group can effectively veto a change by refusing to run the new rules.

What is BIP-54, the Consensus Cleanup?

BIP-54 is a narrow soft fork that fixes four longstanding weaknesses in Bitcoin’s rules: the timewarp difficulty attack, blocks that are slow and expensive to validate, a Merkle tree quirk that can fool lightweight wallets, and a duplicate transaction edge case. It adds no new features. Its lead author, Antoine Poinsot, has published a reference implementation that deliberately leaves out any mainnet activation logic.

What happened with BIP-110 in August 2026?

BIP-110 was a temporary soft fork meant to limit arbitrary data in Bitcoin transactions. When its mandatory signaling window opened in August 2026, only about 2.5 percent of blocks signaled support against a 55 percent threshold. A tiny minority chain split off, produced two blocks, and stalled, while almost all hashpower stayed on the main chain. The episode is now cited as a warning about minority-enforced activation.

Will covenants like OP_CTV ever activate on Bitcoin?

Possibly, but not soon. The CTV-plus-CSFS combination is the developer frontrunner, and an OP_CTV activation client exists with a timeout in March 2027, yet miner signaling has been close to zero. Without an agreed activation mechanism and broader consensus, covenants remain a 2027-or-later question, and some in the community would prefer they never activate at all.

Marcus Okafor is a senior editor at HOGE Wire covering Bitcoin protocol development and market structure.