Last Tuesday, a report crossed my desk with nine empty dimensions. It was supposed to be a deep analysis of a blockchain article โ a second-stage output mapping a source text across technology, tokenomics, market structure, ecosystem position, regulatory risk, team composition, risk, narrative, and industrial-chain transmission. Every single cell said the same thing: N/A โ information insufficient. No project name. No protocol. No data. No conclusion. Just a repeated, quiet refusal to guess.
At first, the document reads like a systems failure. A crawler somewhere had retrieved nothing. An upstream extraction model had returned zero information points. And the downstream stage โ the one designed to produce judgment โ chose to output nothing rather than fabricate something. It even labeled itself: "Input invalid, analysis terminated."
I sat with that document for a long time.
Over the past seven days, another protocol lost forty percent of its liquidity providers. Funding rates flipped negative across major venues. The market is sideways โ chopping, waiting for a direction nobody can honestly provide. In all that noise, the most radical signal I received was a sixteen-page report that said, in effect: I do not know.
My code was the covenant, not just the contract. That empty report was a covenant. It refused to lie.
The system that permits silence
I should explain what this document actually is, because the news here is not the N/A. The news is the existence of a system that permits N/A at all.
Two-stage analysis pipelines have become standard infrastructure among crypto research desks, DAO treasuries, and quantitative funds. Stage one takes an article from the wild โ a Medium post, a governance forum thread, a Telegram announcement, a news story โ and extracts what the framework calls "information points": the smallest citable facts, the atomic units of analysis. Stage two takes those points and maps them onto a nine-dimensional grid. The framework carries an explicit execution constraint: when information is insufficient, state it clearly rather than speculate.
The report I received is the product of that constraint. Its opening section is a quality-control rejection. It lists what upstream failed to provide: article title, source, type, and domain tags all unclassified. The core viewpoint and the author's position unsigned. The information-point list empty โ no citable facts, no attributable claims, no identifiable projects. The report then walks through all nine dimensions and marks each one N/A, with tables whose rows exist but whose cells contain only those three letters.
Then comes the production-environment note, which is the real content. "Empty stage-one output is an abnormal state," it warns. It could indicate an upstream pipeline failure: an article crawl that returned nothing, a paywall or anti-bot wall intercepting the fetch, a source text that is pure image or video, or parameters never passed to the first stage. The recommendation is blunt: add a hard gate before the analysis stage โ if the number of information points is below a threshold, fail the task explicitly, trigger a re-crawl, and preserve metadata like capture time, parser version, and token usage so the failure can be traced. And above all: it is strictly forbidden for the model to fabricate analysis content when there is no information.
I read that sentence several times. It is not an engineering specification. It is a governance principle. The document even closes with a disclaimer โ that it is not investment advice, that crypto assets carry extreme risk, that you should do your own research. A report with zero data, telling you not to trust it. I found that oddly beautiful.
In the summer of 2017, I was a sophomore computer science student in Singapore, and I spent my entire break analyzing fifteen initial coin offerings. I wrote a twenty-page critique titled "Tokenomics as Social Contract," arguing that most of those projects had no genuine community value โ they were speculative shells wearing the costume of protocol. The speculators ignored it. A small Discord of like-minded builders did not. That experience taught me that truth resonates with those seeking meaning, not profit. This empty report has taught me the same lesson again: sometimes the most meaningful thing a system can do is say nothing at all.
The hallucination economy
Every large language model, when handed an unrecognizable project, will produce about five hundred words of confident nonsense. I have tested this. You feed it a ticker that does not exist, a whitepaper that is a PDF of blank pages, a roadmap that is a screenshot of a cat, and the model will solemnly explain the token's deflationary mechanics, the founding team's background, and its strategic partnership with a bank you have never heard of. The output is grammatically perfect and factually dead.
The crypto information ecosystem runs on exactly such output. For every genuine analysis, there are ten thousand generated reports โ each one filling empty cells with plausible fiction. The pipeline that produced my N/A document is a corrective: a machine that refuses to do the thing machines are now taught to do everywhere, which is to pretend.
I spent three hundred hours in 2020 auditing Uniswap V2's smart contracts โ not for security vulnerabilities, but to understand its fair-launch philosophy, how immutable code enforces equality. What I learned is that an audit is only as honest as its object. You cannot audit a contract that does not exist. You cannot extract information points from an empty input. The discipline is identical at every layer of the stack: say what you see, and when you see nothing, say that.
Every broken token taught me how to hold value. I used to think that sentence was about markets โ about 2022, about watching projects I believed in unwind. I understand it more precisely now. The report's pages of N/A are the most valuable token I have held in months. They hold the value of honesty.
The information point as the atomic unit of trust
There is a deeper symmetry here that deserves attention. In blockchain, we talk about the smallest units of verified state โ the hash, the receipt, the log โ as the atoms of trust. The information point plays the same role in this analysis pipeline. It is a fact that can be traced, attributed, and re-verified against the original source. The entire nine-dimensional grid is built from these atoms. No atoms, no analysis.
The report's insistence on metadata โ capture time, parser version, token usage โ mirrors the provenance model of a block header. It treats an analysis as a chain of custody. Who fetched the article? When? With which tool? What did it cost? These are not bureaucratic details. They are the difference between a claim you can audit and a claim you must take on faith. The crypto industry is supposed to be built on the former. Most of the time, it is built on the latter.
I have watched investors make decisions from AI-generated project reviews that contained not a single verifiable information point. The review was a hallucination, but it was a well-formatted hallucination, so it moved capital. The N/A report is the counter-example: it moves nothing, because it contains nothing. And in a world where fabricated information moves everything, the refusal to move anything is the first honest movement I have seen from automated analysis in years.
The subsidized TVL, dressed as analysis
The empty report reminds me of a DeFi protocol I evaluated earlier this year. Its quarterly report led with a liquidity mining APY of 212 percent. Strip out the protocol's own emissions โ the freshly minted token paid directly to farmers โ and the real return was negative. The total value locked was rented. The "users" were mercenaries with no loyalty beyond the incentive schedule. When the emissions stopped, which they always do, the liquidity vanished within four days.
This is the same hallucination problem, expressed in numbers. We have built pipelines that produce metrics, and we have trained the market to treat those metrics as information points. But the underlying input is empty. It is N/A wearing a yield curve.
The uncomfortable parallel: the empty report has the decency to say N/A. Most protocols do not. They keep the emissions running, keep the TVL chart inflated, keep the narrative alive, and call it growth. They are the language model that, given an empty context, chooses to invent a roadmap rather than admit the context is empty.
Liquidity mining APY is a project subsidizing its TVL number; stop the incentives and the real users vanish. I have said this so often it has become a mantra. The sideways market we are in is exactly the season when this becomes visible. As incentives taper and the chop grinds on, the projects that never built real retention are now returning empty extractions. The pipeline that outputs N/A for them is functioning correctly. It is identifying absence.
In the silence of the bear, we heard the truth. We are in a quieter silence now โ a sideways drift, a funding-rate ground fog. The market is waiting for direction, and most of the direction signals out there are noise: outputs from systems that never learned to say "I do not know." The discipline of N/A โ refusing to pay for liquidity that will not stay, refusing to report users who will not return โ is the only discipline that positions you properly in a consolidation.
The data-availability layer that no one needs
Another of my recurring positions connects to the report in an unexpected way. The dedicated data-availability layer โ the blob-carrying network, the modular data bus, the DA chain โ is overhyped. Ninety-nine percent of rollups do not generate enough transaction data to justify a dedicated DA layer. I have told this to founders who looked at me as if I had insulted their children. They have raised thirty million dollars. They have a token. They have a testnet dashboard with beautiful block explorers. What they do not have is data volume.
The N/A report is the mirror of that phenomenon. It is infrastructure built for inputs that never arrive. A DA layer for a rollup processing three transactions a day is an analysis pipeline producing nine dimensions of N/A. The architecture is elegant; the substrate is empty. We have become remarkably good at building systems that would function magnificently if the data existed. The crypto ecosystem is now, in large part, a collection of solutions awaiting problems โ pipelines awaiting articles, DA layers awaiting blobs, settlement layers awaiting users.
The report's production note contains the correct response to this pathology. It does not recommend building more elaborate downstream machinery to process nothing. It recommends failing. Re-crawling. Fixing the upstream. It is the same advice I would give to every rollup founder: stop building data availability for a chain that does not generate data, or accept the emptiness honestly. The N/A is not a bug to be routed around. It is a diagnosis.
The regulator who fills the blanks anyway
The report's fifth dimension runs the Howey test for security classification: money invested, common enterprise, expectation of profits, efforts of others. Every row is N/A. The combined verdict: cannot be evaluated. That is the correct answer. And it is an answer almost no regulator in the world is willing to give.
Hong Kong has spent the past two years constructing a virtual asset licensing regime. The Western press celebrates it as innovation-friendly regulation. I read it differently: it is a geopolitical maneuver dressed as consumer protection. The objective is to claim the position Singapore currently holds as Asia's financial hub. Every license issued, every rule drafted, every consultation paper published is an output generated with almost no regard for the facts of the projects being regulated. The inputs are empty; the outputs are voluminous.
I live in Singapore, so I watch this rivalry closely. The N/A report is a useful lens. A compliant regulator, faced with a project whose attributes are unknown, unproven, unclassifiable, has two options. It can do what the pipeline did: mark the cells N/A, refuse to fabricate, demand better inputs. Or it can do what regulators actually do: fill the blanks with assumptions, call the assumptions findings, and regulate the imagination. Assumption, when the facts are missing, is institutional hallucination. The report's risk flags โ unaudited code, centralized sequencers, excessive admin authority, no peer review โ all appear unchecked, each prefaced with "cannot be confirmed." Regulators never write "cannot be confirmed." They write the license anyway.
This is where my 2025 work comes in. I co-authored a whitepaper called "Algorithmic Stewardship," proposing a framework in which DAOs govern AI models by encoding human values into smart contracts. The feedback was mixed โ some called it premature, others called it naive โ and yet the empty report is now the clearest demonstration of what algorithmic stewardship looks like in production. It is a system that encoded the first and most important human value: humility. The machine chose abstention over invention.
The architecture of refusal
Beneath the philosophy, the report is a piece of production engineering, and it is worth examining the mechanics. Stage-one extraction is the most fragile link in the chain. It depends on a crawler, a parser, a scraping bypass, a language model with temperature settings, a prompt template, a schema. Any one of these can fail silently. A paywall returns a page of cookie-consent text. A JavaScript-rendered article returns an empty body. An image-only protocol captures zero information points. And if the pipeline has no gate โ no threshold check, no validation that the information-point count equals at least one, no hard failure on empty โ the emptiness flows downstream and becomes fake analysis.
That is the real failure mode. Not the N/A report. The report is the system working as designed. The failure would be the absence of the gate. And the temptation to remove the gate is enormous, because empty outputs are operationally inconvenient. They trigger alerts. They burn budget on re-crawls. They require human review. A model allowed to "reason creatively" from nothing would produce output that invoices, that fills dashboards, that justifies the pipeline's existence. The report's recommendation is the opposite: fail loudly. Preserve metadata. Do not let the model be creative with nothing.
I have built enough community systems to know how rare this design choice is. When I founded The Commons in 2024 โ a community platform for ethical Web3 builders โ we grew to two thousand active members by refusing to publish content that did not meet a quality bar. We hosted twelve virtual roundtables on "Technology for Human Flourishing." We rejected more proposals than we accepted. Rejection is the upstream gate; without it, the downstream decays. The same is true of DAOs, where treasuries are drained by governance proposals that contain no information points, dressed in the language of legitimacy. Most governance frameworks lack an N/A state โ a legitimate, respected output meaning "we do not know yet, and we will not pretend otherwise." The empty report offers a model one.
Chop is for positioning
The market context belongs in this analysis, because the timing of this report is not incidental. We are in a sideways market. Funding rates have hovered near zero and flipped negative. Spot volume has thinned. Retail attention has wandered. This is the season when the industry's true inventory becomes visible. In a bull market, everything is information: every tweet, every partnership, every fork, every heavily botted protocol. In a chop, the information points contract sharply, and the N/A cells multiply.
That contraction is clarifying. The protocols with real usage โ actual revenue, retained users, genuine information content โ still produce data in a dead market. The ones that were always subsidized TVL and narrative now return empty extractions. A lateral market is best spent positioning for the next cycle, and the most reliable predictor of which projects will survive that cycle is which ones still have substance in this one. The N/A report is a screening tool. It identifies empty input now, so we are not fooled by its apparent abundance later, when the bull returns and every empty shell is suddenly "information-rich" again.
I spent the 2022 crash in my Singapore apartment, offline, writing twenty essays for a private newsletter I called The Quiet Chain. The most-read essay was about silence โ about the value of withdrawal from the noise. The empty report is that essay, written by a machine. It is a meditation on abstinence dressed as an error log.
The expensive, irrational honesty
And yet โ here is the contrarian turn โ the market will punish this honesty. I have to say that plainly.
An all-N/A report has no alpha. No ticker, no price target, no catalyst. A human analyst who publishes "I do not know" loses followers, loses the algorithmic distribution that rewards confidence and volume. And a pipeline that returns N/A will be "fixed" โ usually not by improving the crawler, but by loosening the prompt, raising the temperature, instructing the model to infer reasonably from context. The institutional pressure to fill empty cells is enormous. Hallucination is cheap: five hundred words of plausible nonsense costs less than a re-crawl, less than a human review, less than the discomfort of telling the client that their article cannot be analyzed.
In production systems, empty output is treated as failure, and failure triggers visibility, costs money, demands maintenance. False output does none of those things. False output becomes a deliverable. A dashboard. A meeting with ten charts. The incentives all point one direction: toward fabrication. This is true across the whole industry. The protocol whose real yield is N/A keeps the emissions running, because stopping means admitting the truth. The rollup with no data keeps its DA layer, because the DA layer is the story. The licensing authority keeps issuing decisions, because decisions are the product regardless of the facts beneath them. We have built an industry that prefers the warm glow of a filled cell to the cold clarity of an empty one.
So the N/A report is not merely rare. It is economically irrational. It is an artifact that exists against the incentives of its own creator. Whoever configured this pipeline chose the expensive option: honesty over billing.
Which is precisely why it is valuable. In an age of industrial fabrication, the system that refuses to fabricate becomes the rarest signal in the market. Every third-party analytics dashboard, every AI-written research note, every paid promotional article is cheap and abundant. Verified silence is the only scarce output left. The moral floor I proposed in Algorithmic Stewardship โ do not invent; if the input is missing, say so โ has now been demonstrated in a live production environment. It is a collector's item. An output that moves no markets, and is more trustworthy than any of the outputs that do.
The covenant of silence
I keep the report on my desk, printed, beside the Uniswap V2 contracts and the 2017 tokenomics critique that nobody read. They are the same object: artifacts of refusal.
The technology was never the covenant. The code was the covenant, not just the contract โ not the token, not the TVL, not the licensing regime, but the willingness to be quiet when there is nothing true to say. In the silence of the bear, we heard the truth; in the silence of the empty report, we hear it again. Every broken token taught me how to hold value; every empty cell teaches me how to build systems worth trusting.
We will know we have finally built a healthy industry when N/A is a respectable answer. When a protocol can report that its metrics are unknown, and the market rewards it for not inventing them. When a DAO can abstain. When a regulator can say "we do not know" and mean it.
The next cycle will not be built by the loudest generators. It will be built by the quiet ones who refused to lie when the cells were empty. That is the direction I am watching for. Chop is for positioning โ and honesty is the position.