Introduction
Nature holds within it fundamental laws that we humans, no matter how hard we try, cannot ignore. Sometimes we invent technologies that are in harmony with such an order, and sometimes we invent technologies, or make changes to existing ones, that violate that order. Often we do this unknowingly, only later discovering the consequence of our mistakes. In turn, those mistakes deepen our appreciation for the fabric of reality we exist within.
Bitcoin, this entropy engine, is not exempt from this pattern of human action. Bitcoin is an incredibly complex and beautiful invention, exhibiting properties that align human incentives and behavior in ways that orient us toward a more efficient utilization, movement, storage, and transmission of energy and information. Yet, we have made changes to it in the past that may have contained an oversight, one whose consequences are only now becoming apparent. In witnessing them, we gain a clearer understanding of what Bitcoin truly is and how we ought to steward it.
A conversation is unfolding across the Bitcoin community over the network’s use for non-financial data. Images, text, and arbitrary files are inscribed onto the timechain through inscriptions and ordinals, burdening node runners with data unrelated to the movement of money. Most discourse treats this as a matter of personal preference, spam, or node policy. This article argues that the issue is more fundamental: a mispricing of the true thermodynamic cost of permanent memory, made possible by a decision nearly a decade ago. We will call this a Landauer attack.
Before applying this idea to Bitcoin, let us review the relevant physics.
The first law of thermodynamics holds that energy is never created or destroyed, only transformed. The second holds that entropy in an isolated system never decreases. Entropy is a measure of the number of configurations a system can occupy; more particles mean more possible arrangements, and therefore more entropy. The universe trends toward its most probable state: maximal entropy. This is why heat flows from hot to cold, why a dropped glass never reassembles, and why every engine ever built loses energy to friction and waste heat. This loss is not an engineering flaw. It is structural to the universe.
Less appreciated is the fact that information and energy are not separate domains, but two expressions of one reality. Claude Shannon formalized information as a measurable quantity using a formulation nearly identical to the entropy equations of statistical thermodynamics1. Information obeys the same conservation laws that govern heat and motion, including when it is remembered or forgotten.
In 1961, Landauer introduced a concept fundamental to the physics of computation: erasing a bit of information is not free2. It requires a minimum amount of energy to be dissipated as heat. This is Landauer’s principle. Any operation that destroys information, collapsing two possibilities into one, must shed entropy somewhere, and that shedding has an energetic floor that no engineering can escape. It is woven into the second law itself.
For decades, the principle remained largely theoretical, confirmed only at small scales. In 2025, physicists led by Stefan Aimet and Mohammadamin Tajik experimentally probed Landauer’s principle in the quantum many-body regime, publishing their findings in Nature Physics. Using a quantum field simulator of ultracold Bose gases, they tracked a quantum field through a mass quench and measured the entropy production tied to its information changes3. Their results confirmed, for the first time in a complex quantum system, what Landauer proposed over sixty years earlier: the link between information and energy is not a metaphor. It is real, measurable, and binding.
Why This Matters for Bitcoin
This gives us a powerful lens through which to reconsider what Bitcoin is. Every byte written to the chain was purchased with irreversible computation: trillions of discarded hashes, each one information destroyed, each destruction paid for in dissipated heat. Once written, that byte imposes a standing obligation, replicated and defended against entropy by every node, forever. The chain is not merely a record: it is a physical commitment, paid for in energy at its creation and maintained by energy for as long as it exists.
It is tempting to view mining as merely the process of finding a winning hash, but it is better understood as a competitive process for inscribing information into permanent memory, a record that will persist far into the future.
There is only one Bitcoin ledger. Just as the supply of satoshis is finite, so too is the block space available at any moment. Every node must store this data forever. There is no opt-out, no deletion, no changing the rules, no altering history, no turning back the clock. Permanence is the entire point of the invention. Like human time, and like entropy itself, the timechain flows in one direction only. Every bit committed to it is therefore not merely data, but a permanent energetic burden borne on every node operator: disk space, bandwidth, and the electricity required to maintain it.
This is why mining functions, in part, as a bidding process. A transaction fee is an economic signal. It is the expression that a piece of data is worth its permanent inscription because someone chose to spend their precious sats to compensate the network for the burden it imposes. In a well-functioning fee market, that signal is honest, reflecting the true cost each byte imposes on every node that carries it forward. Because every byte shares an identical thermodynamic burden, every byte of a transaction should therefore be priced identically.
What happens, then, when that signal is distorted because one class of bits is different from another? This is exactly what occurred with Segregated Witness.
A Brief History: The Block Size Wars and SegWit
To understand how we arrived here, recall the block size wars of 2015–2017, the pivotal conflict over Bitcoin’s scaling. The debate was whether to raise the 1 MB block limit via a hard fork or preserve the status quo in the interest of decentralization. Large miners and businesses favored bigger blocks, lower fees, higher throughput. Node operators, developers, and users favored smaller blocks so anyone could keep running a full node and keep the network decentralized.
The case for smaller blocks was simple: bigger blocks mean greater storage and bandwidth costs, and because blocks are produced at roughly ten-minute intervals with a fixed size limit, that growth is predictable. Larger blocks burden every node, and this burden falls hardest on those with the least access to hardware and bandwidth, concentrating influence in fewer, wealthier hands.
Node operators won this war, but not without cost. The compromise was SegWit, activated in 2017, which separated signature data, or witness data, from base transaction data. The 1 MB base limit remained, but witness data gained roughly three additional megabytes of allowance, raising the practical block size limit to around 4 MB.
The justification seemed reasonable at the time: signatures validate a transaction once but are not strictly needed to verify the UTXO set thereafter, so witness data could theoretically be pruned more aggressively. On this presumption, witness data was weighted at one quarter the fee cost of base data. As a result, a witness byte costs roughly 75% less than a base byte, despite occupying the same physical storage on every full node.
This upgrade also solved transaction malleability, the ability to alter a signature without changing a transaction’s validity. By segregating signatures from the transaction identifier, SegWit enabled commitment transactions whose validity could be guaranteed in advance, the foundational requirement for the Lightning Network. In solving malleability, SegWit unlocked a genuinely transformative scaling solution.
It is worth being clear regarding two separate features of Segwit. The malleability fix, which made Lightning possible, follows from segregating signature data from the transaction identifier. The fee discount applied to that data was a separate decision.
Defining the Landauer Attack
Prices, properly understood, are not arbitrary numbers attached to goods by sellers. They are signals, condensed from the voluntary exchange of countless individuals acting on knowledge no central authority could ever fully gather. In a free market, a price tells us, in a single figure, how scarce a resource is relative to demand, directing it toward its most valued use without requiring anyone to grasp the system as a whole.
A Landauer attack is what occurs when that calculation breaks. It is the gap between the price charged for permanently inscribing a bit and the true thermodynamic cost that bit imposes on every node forever. It requires no malicious intent, no more than a shortage under a price ceiling requires malice from a shopper who buys more simply because the goods are so cheap. It requires only mispricing. Rational actors faced with an artificially cheap resource will use more of it. This is not a flaw in human behavior; it is the correct, predictable response to a false signal, which is precisely why integrity of that signal matters.
It is worth being precise about the actual attack surface. A discount alone is not catastrophic, no more than a single subsidized good collapses an economy. The danger lies in what subsidies always produce over time: a misallocation that compounds the longer it persists. The fee is paid once. The storage cost is permanent and recurring. When a one-time price fails to capture an ongoing cost, the market is not pricing the good at all; it is pricing only the moment of sale, while the true cost is quietly shifted onto a third party with no say in the transaction: the node operator, who bears it long after the fee is spent. This is the signature of every subsidy: benefit concentrated on the chooser, cost diffused onto everyone who must live with it.
This differs meaningfully from the centralization concern of the block size wars, which was a question of affordability: could the average person still afford to run a node as blocks grew, a question technology could, in principle, ameliorate? The Landauer attack is a question technology cannot solve because it does not originate in technology. Strip away hardware cost, Moore’s Law, the falling price of storage, and one fact remains: a bit is physical, and physics does not distinguish a witness bit from a base bit. Landauer’s principle establishes that information carries real energetic cost, and that cost attaches to the bit itself, not to the category a protocol assigns it. Hardware improvement lowers the burden for every byte equally. It cannot make one class of physically identical bytes cheaper than another. Only a pricing rule can do that.
Discounting witness data relative to base data, despite an identical thermodynamic burden, is a subsidy of one category over another. Subsidies do not simply make things cheaper; they distort market signals so that economic actors do not make the decisions that properly reflect the healthy allocation of resources. Capital, here in the form of block space, no longer flows to its most valued use; it flows to whatever sits on the subsidized side of the ledger. This is not a moral failing of those who exploit the discount. It is the same rational response that drives any actor toward an underpriced good, whether housing, credit, or a permanent inscription on the most secure ledger in history. The fault lies with the incentive, not the actor responding to it. An incentive structure that rewards witness inscription over base inscription, despite identical cost to the network, is not a neutral feature of the protocol. It is a standing invitation for capital to misallocate itself until the price signal is repaired.
This is not hypothetical. It is the proximate cause of the debate now roiling the node-running community.
In 2023, a developer discovered a method for inscribing arbitrary data, including images, directly into Bitcoin transactions, giving rise to Ordinals: Bitcoin-native NFTs. The technique relies on witness space introduced by SegWit and made further accessible by Taproot.
The result has been a measurable acceleration in chain growth, driven by data unrelated to monetary flow. Following the introduction of Ordinals, average block sizes increased. Node operators now absorb a burden that is being mispriced in the fee market by 75%.
The Search for a Solution
The community’s response has centered on node software. Many have migrated from Bitcoin Core to Bitcoin Knots, which maintains stricter relay filters and greater configurability. This is an understandable choice, but it does not address the underlying mispricing. Filtering what a node relays does not change the user’s incentive to pay a miner to include a transaction that uses cheap witness data. Node-level filtering patches a problem whose root cause lies upstream in the protocol’s pricing structure.
More recently, this tension gave rise to BIP-110, a proposed fork that restricts certain methods of embedding non-monetary data through limits on specific Taproot and witness script structures. Its mechanics and activation method merit their own discussion elsewhere. What matters here is that BIP-110, whatever its merits, is another downstream patch. It restricts particular exploitation methods without correcting the discount itself. Furthermore, it is a temporary rule change, and is not designed to be a permanent solution. As demonstrated by a user who embedded an image within BIP-110’s own constraints, restricting specific techniques tends to produce a technical game of whack-a-mole rather than a lasting solution, because the economic incentive to exploit discounted witness space remains fully intact.
If this diagnosis is correct, the right frame is not spam, personal preference, or even decentralization narrowly construed. It is conservation.
Bitcoin’s original protocol enforced a clean symmetry: every byte of block space competed under identical economic rules, regardless of content. 1 byte = 1 byte. This symmetry allowed the fee market to function with honest price discovery. Whatever made it onto the chain had, by definition, paid a price that the market determined was commensurate with the permanent burden it imposed on the network.
Recent work viewing Bitcoin through the lens of physics argues this symmetry is not merely good design, but is analogous to a conservation law. In their 2026 paper, Bitcoin: The Architecture of Time, researchers Jack and Nick treat Bitcoin as a system in which time itself is computable and quantized, each block representing a discrete, irreversible unit of inscription rather than a point on a smooth continuum4.
Within that frame, one could argue that the protocol’s original constants, the 21 million supply, the 1 MB block limit, and the relationship between satoshis and bits, are not arbitrary, but expressions of an underlying physical symmetry: 1 sat = 1 sat, 1 bit = 1 bit. Bitcoin’s legitimacy as a ledger of permanent memory, in their view, depends on that symmetry holding, much as a physical theory depends on its conservation laws. Whether or not one accepts the full scope of their argument, the intuition is useful: a protocol that claims to price permanence honestly cannot do so while treating two physically identical bits as different categories of cost.
SegWit, in solving malleability and unlocking Lightning, introduced an asymmetry into this otherwise clean system. It created two classes of data, identical in physical reality and thermodynamic burden, but priced as though fundamentally different in kind. This is no different than any economic subsidy: a distortion of price signals that misallocates resources.
A return to symmetric pricing would not require reverting to a strict 1 MB block, nor would it end the Lightning Network. It requires only recognizing that the thermodynamic cost of permanent inscription does not discriminate by category. To the node that must store them forever, a signature and an output are identical in cost. They should not be distinguishable in price.
When pricing reflects this honestly, the fee market resumes its function as an arbiter of value. Data reaches the chain only when its worth to the payer exceeds the true cost of permanent inscription. Spam becomes self-limiting, not through an ever-expanding set of rules that define and exclude it after the fact, but because the incentive to exploit a subsidized resource disappears once pricing is honest. Accordingly, node operators are fairly compensated for the burden they carry. The conservation law that gave Bitcoin’s original design its integrity, one bit equals one bit, just as one satoshi equals one satoshi, is restored.
Conclusion
We return to where we began. Bitcoin is an entropy engine, converting physical energy into an immutable record, bound by the same thermodynamic laws that govern every process in the observable universe. The realization that information itself carries an inescapable energetic cost, proposed by Landauer over sixty years ago and now confirmed by recent research, gives us a sharper understanding of what we have built, and what we may have unintentionally broken.
SegWit was not a mistake in conception. It solved a real problem and enabled scaling the network surely needed. However, bundling that solution with an unexamined discount on one category of data may simply reflect that we did not yet appreciate how fully Bitcoin mirrors the physical order from which it was born from. Nearly a decade later, we are watching the consequences unfold, as the community fractures over node implementations, relay filters, and increasingly aggressive soft forks, each patching a symptom while leaving the cause untouched.
If Bitcoin is truly governed by conservation laws as fundamental as those governing energy and mass, the path forward lies not in more elaborate filters or endless debates over legitimate use. It lies in asking what the physics requires and having the discipline to restore it. Bit by bit.
References
1. Shannon CE. A mathematical theory of communication. Bell Syst Tech J. 1948;27(3):379-423.
4. Jack & Nick. Bitcoin: The architecture of time. Published online February 2026. https://bitcoinlens.net/
