The Null Response: On Hallucination, Silence, and the Ethics of Knowing Nothing

Hook

At 2:14 in the morning, Manila time, a document landed in my inbox that contained almost nothing at all — and it was the most honest thing I had read in months.

It was a nine-part analysis of a crypto project. Nine dimensions: technology, tokenomics, market structure, ecosystem position, regulatory posture, team and governance, risk, narrative, and supply-chain transmission. Every heading was present. Every table was drawn. Every row was labeled. And every single cell, without exception, read the same four words: information insufficient.

A colleague had run it through an automated research pipeline. The pipeline had failed upstream — the source material never arrived, the parser returned a null object, and the model downstream was handed an empty box. What it did next is the reason I am writing this essay.

It refused to invent. It returned a skeleton of a report, and then, rather than dressing that skeleton in the flesh of plausible numbers, it said so. It named the failure. It flagged the pipeline, the parser, the possibility of an OCR fault, the possibility that the article had simply never existed. It wrote, in effect: I would rather tell you clearly that the information is insufficient than fabricate conclusions that look professional on empty input.

I have spent twelve years in and around this industry. I have read thousands of research notes, thread essays, dashboards, and so-called deep dives. I can count on one hand the number of times a crypto analyst has told me, plainly and without hedging, that they do not know.

So I sat in the dark of my apartment, the fan turning slowly overhead, and I thought about what that empty document had actually delivered. It had not delivered a single fact. It had delivered something rarer: a demonstration of integrity under pressure. And it made me realize that the central crisis of the 2025–2026 bear market is not liquidity, not regulation, not the slow institutional creep of ETFs into a space that once belonged to dreamers.

The central crisis is that we have built machines — and cultures — that would rather hallucinate than admit they are blind.

Context

Let me set the scene, because the empty document did not appear in a vacuum. It appeared in a specific moment, in a specific industry, and it was answering a question nobody had asked out loud.

We are, as I write this, deep in a bear market. Not the theatrical bear of a twenty-percent drawdown, but the grinding, soul-testing bear that strips projects down to their bones. The kind of market where the question "will this still be here in eighteen months?" is not rhetorical. Readers do not want alpha right now. They want to know whether their assets are safe. They want to know which protocols are bleeding liquidity, which treasuries are quietly insolvent, which teams have stopped shipping while their community managers post motivational threads into the void.

This is the market in which analysis became an industry unto itself. When prices fall, the demand for explanation rises. Everyone wants to know why. Everyone wants a framework, a model, a dashboard, a signal. And so an entire apparatus grew up to feed that demand — half of it human, half of it increasingly automated.

Here is the thing about that apparatus: it was optimized for output, not for truth.

I have watched this happen from the inside. In 2021, when I launched my own community for women and marginalized creators in the NFT space, I watched newcomers ask simple questions — how do I know if this is safe? — and get flooded with confident answers that turned out to be wrong. I sat through twelve virtual workshops guiding fifty women through wallet setup and minting on Ethereum, and I learned that the most dangerous moment in a beginner's journey is not the technical step. It is the moment someone with a confident voice tells them something that is not true. In 2022, when my own portfolio drew down by 85% and I retreated into six months of studying algorithmic stablecoin collapses, I learned to distinguish between the analysts who understood the machinery and the analysts who understood the vibes. The vibes crowd was louder. It always is.

And now, in 2025 and 2026, the vibes crowd has been automated.

Large language models can generate a thousand words of confident crypto analysis in seconds. They can imitate the cadence of a research note so precisely that a casual reader cannot tell the difference between a report grounded in on-chain data and a report grounded in nothing at all. They can invent a TVL figure, a funding rate, a governance proposal, a founder's background, and wrap it all in the reassuring grammar of expertise. They can do this at a scale no human newsroom could ever match, and they can do it for free, and they can do it about projects that do not exist.

This is the phenomenon the industry has started to call, half-jokingly, hallucination — the tendency of a system trained on language to produce language that is fluent, plausible, and false. But I want to be careful with that word, because it lets us off the hook. Hallucination sounds like a glitch, an accident, a bug in the machine. What we are actually dealing with is something older and more human: the compulsion to fill silence with noise, to answer questions we have no business answering, to perform certainty because certainty is what sells.

The empty document that arrived at 2:14 a.m. was remarkable precisely because it did the opposite. It performed nothing. It sat in the silence and let the silence be.

That, I think, is the beginning of a real analysis. And it is the thing the next cycle will be built on, or it is the thing the next cycle will bury. I do not yet know which. But I know that the question is now on the table, and once a question like that is on the table, it does not come off.

Core

Here is where I want to slow down, because the philosophical point I am making has a technical spine, and if I do not show you the spine, I am guilty of the same vagueness I am criticizing.

Let me talk about what data actually means in crypto, and why the temptation to fabricate it is not just an ethical failure but a technical one.

The Softness of Hard Numbers

Start with the most basic unit: the number on a dashboard. A user sees "TVL: $412 million" and feels something — reassurance, or alarm, depending on the trend line. But that number is not a fact about the world. It is the output of a formula, and that formula contains assumptions that are almost never shown.

Consider lending protocols. The interest rate a borrower pays on Aave or Compound is not discovered by the market the way a stock price is discovered. It is administered by a curve — a piecewise function that says: if utilization is this high, the rate is that high. The curve was chosen by humans, calibrated to historical conditions, and then frozen into the protocol. It has a certain internal logic. But it has no necessary relationship to the actual supply and demand for credit in the real economy. When I read a dashboard that reports "the market rate for USDC is 4.7%," I want to ask: which market? Because there is no market there. There is a curve, and the curve is a decision, and the decision was made by someone who is not you.

I want to be precise about this, because it is the technical heart of my ethical argument. An interest rate model is a kind of narration. It tells a story about what borrowing should cost, and it tells that story in the language of mathematics, which makes it feel discovered rather than invented. But it is invented. The slope of the curve, the location of the kink, the choice of the base rate — these are aesthetic and political choices dressed as engineering. The precision is cosmetic. The number is a narrative.

This matters for our topic because it shows that even the hardest number in DeFi is already a soft thing dressed in hard clothes. If the underlying reality is already a constructed story, then a model that hallucinates a number is not doing something categorically different from what the protocol is doing. It is doing the same thing faster, and without the courtesy of a whitepaper.

The Blob and the Story

Now layer automation on top of that. An AI model that has been trained on years of crypto dashboards learns the shape of these numbers — the plausible ranges, the typical relationships, the reassuring roundness of a figure like $412 million. When you ask it to analyze a project it has no data on, it does not return an error. It returns a number that looks like the numbers it has seen. The fluency is the danger. A bad analyst produces obviously bad analysis; you can smell it. A hallucinating model produces analysis that is indistinguishable, at the level of surface texture, from the real thing. It is wrong in the way a counterfeit bill is wrong — only detectable under ultraviolet light, only if you know to look.

I have made this mistake myself, and I want to name it honestly, because my credibility depends on it. In early 2025, I was preparing a community explainer on Layer 2 economics. I had a model help me summarize the fee dynamics after Dencun. It produced a clean, confident paragraph about how blob space had driven rollup costs down and how fees would remain structurally low. It sounded right. It was the consensus narrative. But when I went to verify it against actual blob utilization data — based on my own habit of pulling the raw numbers before I repeat them — I found the trend I had been handed was not merely imprecise. It was pointed in a direction that ignored the mechanism entirely.

Blob space is a finite resource. Dencun expanded the supply, which collapsed the price of posting data, which collapsed rollup fees. But the supply is fixed by the protocol, and demand is growing, and there is no mechanism that guarantees the two stay in balance. What I actually believe — what I have come to believe after watching the utilization curve climb month over month — is that post-Dencun blob data will be saturated within roughly two years, and that when it is, the price of blob space will rise, and rollup gas fees will double again. Not because of a conspiracy, not because of a decision, but because that is what happens when a subsidized resource meets unconstrained demand.

The model could not tell me that, because the model had learned the story of cheap L2 fees, not the mechanism of blob scarcity. It had learned to hallucinate the consensus, which is a subtler sin than inventing a number from nothing, but it is the same sin. The consensus is a story. The mechanism is a fact. And the two diverge exactly at the moment when it matters most — the moment when the subsidy ends and the bill comes due.

The Anatomy of a Null

I want to dwell on the anatomy of that empty document, because the way it failed is instructive.

It failed in layers. The first layer was the data pipeline — the machinery that was supposed to fetch the source article and hand it to the analysis engine. Something in that chain broke. The article may have never been scraped, or the scrape may have returned an error page, or the parser may have choked on a format it did not recognize. This is the most common failure in crypto research, and the least discussed: not that the analysis is wrong, but that the input never arrived, and nobody noticed.

The second layer was the handoff. The pipeline passed an empty object downstream without raising a flag. In a well-designed system, an empty input should trigger an alarm — a stop, a log, a human review. Instead, the empty object slid through, and the analysis engine was asked to reason about nothing.

The third layer — and this is the one that matters — was the engine's response. It could have done what a hundred marketing bots would have done: manufactured a project, invented a thesis, produced a confident report. It had the capacity. It had the language. It had, presumably, been trained on enough crypto content to fake expertise indefinitely.

Instead, it drew a line. It said: here is what I would need, and here is why I will not proceed without it.

And then it did something I found genuinely moving. It explained why the discipline mattered. It wrote that fabricating analysis on empty input would violate the core constraint of the framework — that every dimension must be grounded in source material, and that where the material is absent, the correct output is not a guess but a marker. It wrote that it would rather say insufficient than hallucinate, and that this distinction is the boundary between rigorous analysis and content hallucination.

I have spent a career trying to articulate that boundary, and this document articulated it in a footnote.

So let me be precise about what the document is. It is not a failure of analysis. It is a demonstration of a meta-skill that the industry has almost entirely neglected: the skill of knowing when you do not know, and having the discipline to say so. In a bull market, that skill looks like weakness. In a bear market, it is the only thing standing between a reader and a catastrophic decision.

The Culture That Rewards the Mirror

I want to broaden this out, because I have been talking about AI systems, and it would be a mistake to let human analysts off the hook. The empty document was honest. But the reason it was remarkable is that the culture around it is so dishonest.

Let me trace the culture. Crypto, at its best, was a promise about the relationship between truth and verification. The whole point of a blockchain is that you do not have to trust a narrator — you can check the ledger. The whitepapers that captivated me in 2017, when I was a nineteen-year-old finance student in Manila reading about decentralized compute and trustless exchange, were not promises of profit. They were promises of verifiability. The dream was that we could build systems where the question is this true had a mechanical answer. I sat in a hackathon that year as the only woman among fifty male engineers, and what held me there was not the money. It was the idea that code could be a form of law, and that law could be checked by anyone.

Somewhere along the way, we replaced verifiability with vibes.

The substitution happened gradually, and it happened for understandable reasons. Verifiability is hard. Checking a wallet is tedious. Reading a contract is technical. Reading a whitepaper is slower than reading a thread. And so the market rewarded the people who narrated the ledger instead of the people who read it. The narrators grew loud, then louder, and then the machines learned to narrate too.

By the time institutional capital arrived — through Bitcoin ETFs, through the slow normalization of crypto in traditional portfolios — the culture had already been hollowed out. The institutions came expecting rigorous analysis, and they found a hall of mirrors. And here is the irony: the institutions brought their own analysts, their own models, their own AI tools, trained on the same polluted corpus, and the pollution propagated. The mirror reflected the mirror.

This is what I mean when I say hallucination is a cultural problem. It is not that the machines are broken. It is that we built an information economy in which the appearance of knowledge was worth more than knowledge itself, and then we automated the appearance.

And this is why the empty document felt like a small act of resistance. It was a machine that refused to participate in the economy of appearance. It was a null response in a world that cannot tolerate nulls.

A null result is not nothing. It is a mirror.

Contrarian

Now I want to turn the knife on my own argument, because the position I have been building is satisfying, and satisfying positions are usually wrong somewhere.

Here is the contrarian angle: the demand for I don't know is not actually a demand for truth. It is a demand for a different kind of reassurance.

Think about it. When I say information insufficient, cannot evaluate, the reader does not feel informed. They feel abandoned. They came to me because they were scared, because their portfolio was down, because they needed someone to tell them whether to hold or fold. And I handed them a null. From the reader's perspective, the honest analyst and the useless analyst look exactly the same: both give them nothing.

This is the trap. Rigor, practiced for its own sake, becomes a form of cowardice. It is easy to say I don't know — it protects you from being wrong. It is much harder to say here is what I believe, here is my confidence level, here is what would change my mind. The empty document I admired so much is, in one reading, a beautiful refusal. In another reading, it is an abdication.

I have felt this tension in my own work. After my 85% drawdown in 2022, I became — for a while — the analyst who only wrote about what could go wrong. Every essay was a risk analysis. Every conclusion was a caveat. And my readers, the ones who had stayed with me through the crash, started to drift. Not because I was wrong. Because I was unhelpful. They did not need a list of ways to lose money. They needed a judgment, however provisional.

So the contrarian lesson is this: honesty is not the same as abstention. The point of admitting what you do not know is to make your actual judgments more trustworthy, not to replace judgment with a shrug. The empty document was valuable because it was rare — because it existed in contrast to a thousand documents that lied. If every analysis were a null, the null would be worthless. The discipline only means something when it is deployed selectively, in service of the judgments you can make.

There is a second, sharper contrarian point. The industry's obsession with verification and on-chain truth has its own dark side. When we say the chain does not lie, we are telling a partial truth. The chain records transactions. It does not record meaning. It cannot tell you whether a transfer was a payment, a bribe, a wash trade, or a mistake. It cannot tell you whether a governance vote reflects conviction or a whale's whim. The ledger is true and the truth is thin. And an analyst who worships the ledger will produce exactly the kind of analysis the empty document refused to produce: technically grounded, humanly empty.

I think about this whenever the conversation turns to CBDCs and the future of money. The techno-optimist position is that blockchains will free us from surveillance. But the same ledger that makes fraud detectable makes every transaction legible to whoever holds the keys. A CBDC is not the opposite of a blockchain; it is a blockchain with a different governance layer — one where the observer is the state and the privacy is nominal. I believe these two futures — the surveillance coin and the freedom coin — cannot coexist, because they compete for the same substrate, the same legitimacy, the same claim to be the future of money. One of them wins, and the winner determines what a ledger means.

And an analyst who cannot say that — who cannot make a value judgment about which future they want — is not being rigorous. They are being neutral in the way a scale is neutral, measuring weight without asking what is being weighed. Neutrality is not a virtue when the thing you are weighing is a choice between two worlds.

So here is the synthesis, the thing I actually believe. The empty document was right to refuse to fabricate. But the reason it was right is not that fabrication is technically inaccurate. The reason is that fabrication is a form of theft — it steals the reader's ability to make a real decision. A confident lie is worse than an honest silence, but an honest silence is only valuable if it is in service of eventually saying something true.

The mature position is not I don't know as a permanent refuge. It is I don't know yet, and here is what I am doing about it. And that, in turn, means the null must be paired with a plan. The honest analyst owes the reader three things: the boundary of what is known, the boundary of what is not, and a commitment to close the gap. Anything less is just a more sophisticated way of saying nothing.

Code can be copied. Conscience cannot.

Takeaway

So let me try to say something true, knowing that I might be wrong.

The bear market will end. It always does. And when it ends, the projects that survive will not be the loudest ones, or the best-narrated ones, or the ones with the most confident dashboards. They will be the ones that were actually there — the ones whose code shipped, whose treasuries held, whose communities stayed. The null responses, in other words. The protocols that did not hallucinate a future they could not deliver.

I find myself thinking, as I often do, about the seeds we planted. From the ashes of 2022, we planted seeds for 2030. Some of those seeds were real — the researchers who kept publishing when no one was reading, the developers who kept committing when the token price was down ninety percent, the community builders who kept onboarding women and newcomers when the money had left the room. And some of those seeds were counterfeit — projects that grew by narrating a harvest they had no intention of growing.

The market is now in the business of telling them apart. And the market is not always good at it. But the market is better at it than it used to be, because the market has been burned. Burned by algorithmic stablecoins that promised stability and delivered zero. Burned by lending protocols whose risk parameters turned out to be vibes with decimal points. Burned by analysts who were certain about things they had never checked.

We plant seeds in the dark, trusting a harvest we may never see.

The lesson of the empty document is that certainty is cheap and checking is expensive, and the only way to tell them apart is to build a culture that rewards the expensive thing. Not a culture of permanent doubt — that is its own kind of cowardice. A culture of earned confidence, where every claim carries the weight of the work behind it, and where the words information insufficient are treated not as a failure but as a form of respect.

I do not know exactly what the next cycle looks like. I do not know which Layer 2 will win, or whether blob fees will double in eighteen months or thirty. I do not know how the CBDC debate resolves, or whether AI agents will hollow out the human communities that made this space worth caring about. I do not know whether the institutions now entering through ETFs will preserve the culture or dissolve it.

But I know what I will do. I will keep the nulls. I will keep the sentences that say insufficient information. I will keep the practice of checking a number before I repeat it, of reading a contract before I trust it, of distinguishing the mechanism from the story.

And I will keep writing, because the alternative — silence — is the one thing a hallucinating industry actually deserves, and the one thing it should never be given. The chain will remember what we choose to forget. The question, as always, is what we choose.