I've been heads-down working on shipping Miniscript in Nunchuk. I look up three months later, and people are STILL debating mempool filters. 😀 Since we have a moment before our next major release, I’ll jot down my thoughts on the matter. Disclaimer: I have no horse in this race. I don't think keeping the current OP_RETURN limit or raising it will have a material impact. My only interest is the truth and what's best for Bitcoin. Interestingly, I disagree with both sides of the debate (the so-called "spammers" and the "filterers") for reasons I'll explain below. Let’s first establish some ground truths: 1. Proof-of-Work is more special than you think. History is filled with rulers who debased their currencies—the original "Proof-of-Stake" systems. Yet, historical PoW money (gold) always prevailed. People in ancient Rome and Confucian China alike knew gold was valuable. This is distilled wisdom, passed down through diverse cultures. Our ancestors have an intuitive understanding of PoW, despite not knowing PoW is (unforgeable costliness). It is the most objective truth we have, independent of language or belief system. Everything else is fake in comparison. The JPEG peddlers don't fully grasp PoW's economic implications, and the filterers gravely underestimate its power as the ultimate filter. 2. Monetary transactions have a higher economic density than data transactions. A monetary transaction moves value (V) for a fee (F). A data transaction moves value (V) plus some perceived data premium (D1​) for a fee (F) plus an extra fee for the data payload (D2​). Thus, the fee rate for a monetary transaction is F / V. The fee rate for a data transaction is (F+D2​) / (V+D1​). In the long run, the on-chain data premium (D1​) trends toward zero, except for a few special truly non-fungible, rare sats (not artificially rare). As this happens, monetary transactions become vastly more economically efficient and will inevitably outcompete data transactions for block space. 3. Relying on data alone to prove a positive is a logical fallacy. Data can help formulate a hypothesis, but that hypothesis is fragile. It only takes one contradictory data point to destroy it. For instance, I've seen the argument: "Our OP_RETURN filter must be working, otherwise why would there be significantly more transactions with under 80 bytes of data than over 80 bytes?" This logic is trivially debunked. For 16 years, we rarely saw transactions paying less than 1 sat/vB, a long-standing mempool policy. One could have concluded the policy was an iron law. Yet, in just a few months, 0.1 sat/vB transactions have become common despite 99% of nodes still enforcing the old policies, proving the policy was a transient heuristic, not a fundamental truth. It works until it doesn't. Here’s another analogy: A society that lives next to a volcano for a thousand years without an eruption might theorize that fire can never fall from the sky. That theory, based on a millennium of "data," will one day get them killed. Data is insufficient without being coupled with first-principles analysis. Now, for my opinions: I strongly believe that any non-PoW-based method to filter spam, which must rely on subjectivity and "rough social consensus," is doomed to fail. Worse, it has a centralizing effect, mirroring the dynamics of Proof-of-Stake. This "social consensus" filtering is just PoS by another name. We've seen this cat-and-mouse movie too many times, especially during the internet's evolution (see: email, SMS, DNS, social media). Email is the perfect case study. It demonstrates two things: (a) The cat-and-mouse game of subjective filtering inevitably leads to extreme centralization. (b) Bitcoin is luckier than email because it has a built-in spam-mitigation tool: transaction fees. The cost-benefit analysis for email spam is that it's nearly costless to send, so a 0.001% success rate is a win. All costs are externalized to the network and its users. The cost-benefit for on-chain JPEGs is the opposite: one must pay a higher fee rate per unit of economic value transferred. This cost is internalized by the "spammer" (with some long-term storage costs borne by the network). Therefore, on-chain JPEG spam is inherently not sustainable. Email spam is. Furthermore, concerns about short-term consequences or chain "bloat" are sufficiently mitigated by the existing blocksize limit. The worst-case scenario is a linear chain growth of 100-200 GB per year. In the grand scheme, this is perfectly acceptable, as the falling cost of storage continues to follow Moore's Law, making this a manageable and decreasing burden over time. Ultimately, this all comes down to one thing: PoW is the only objective, incorruptible mechanism for separating signal from noise in the digital realm. Chasing a non-PoW filtering method is like chasing a perpetual motion machine, a utopia that defies physics. PoW is about understanding the world and its constraints through the lens of physics. It's the highest signal of truth because it's literally built on the undisputed currency of the universe: energy. Everything that isn't grounded in this way is a pale imitation. Bitcoin's PoW mechanism is a beautiful, profound emulation of the cosmic process that forges gold from energy (neutron star collision). It connects digital truth to physical reality. To truly appreciate Bitcoin is to appreciate this fundamental connection. It's a special, almost sacred, principle, and I believe it's the one that matters most. What's ironic is that while we debate, the economic reality I described is already playing out. All past and current attempts at selling on-chain JPEGs have fizzled out or are in the process of doing so as people wise up to their true worth. At this rate, people will spend more time talking about filtering JPEGs than the JPEGs themselves will remain relevant. Spending time debating JPEGs when the mempool is near empty, fee rates at historical low, and people flocking to paper Bitcoin is an inefficient allocation of time and resources. Believe in PoW. Let's focus on the real fight: making self-custody safe and accessible for everyone. That means working on things like Miniscript, MuSig2, and FROST, more hardware signers and form factors/UX, and education. We must dispel the myth that self-custody is impossible for the average user. Turning as many people as possible into sovereign individuals. That is the real fight.