Last Tuesday, at two in the morning Chengdu time, I opened a document someone had sent me to review. It was a nine-dimension research report on a crypto protocol. Forty-one tables. A Howey test breakdown. A risk matrix with probability and impact columns. A token unlock schedule. A narrative-sustainability index. A five-star rating system for "information value." Eleven thousand words long.
Every cell said the same thing: N/A โ insufficient information.
I have read a few thousand of these documents. I have written a few. I had never seen one that admitted, in every field, that it knew nothing โ and then still ran to forty pages. The tea beside my laptop went cold. Outside, Chengdu was doing what cities do at two in the morning, moving without anyone watching. I sat with the blinking cursor and felt something rare in this industry: respect.
Then I felt the fear, because I know exactly how that document came to exist.
Back in 2017, when I was thirty-three and drafting a forty-page whitepaper on tokenized equity for Polymath, the research culture looked nothing like this. There were no templates. There were no nine-dimension frameworks. There were eleven people in a Telegram group arguing about whether a security token could be a form of digital citizenship, and most of us were wrong about most things. The uncertainty was visible because nobody had yet invented a machine to hide it.
That machine exists now. It is the analyst template โ a product, not a method. Nine dimensions, forty-one tables, a verdict at the end. Research desks sell it. KOLs sell a compressed version of it. It has become the standard unit of credibility in a market where credibility is the scarcest asset of all.
The template spread because the demand for it is real. It is 2026, and we are deep enough into this bear market that readers no longer want philosophy. They want to know whether their collateral is safe, which liquidity pools are bleeding, which protocols will still have a treasury in eighteen months. In a bear market, an unanswered question feels like a betrayal. And so the industry learned to answer every question, whether or not it had the material to answer it.
Which is why the document in front of me mattered. It had been handed a blank input โ no title, no source, no information points, no identified protocol โ and it had refused to invent one. It preserved the template. It kept the tables. And in every cell where a conclusion belonged, it wrote: insufficient information.
I want to explain why that refusal is technically correct, and why it is almost never what happens.
A missing price is not a price of zero. This is the first thing anyone learns when they build oracle infrastructure, and it is the principle the entire research industry quietly violates. When a Chainlink aggregator cannot reach consensus on a feed, it does not return zero; it reverts. It refuses. The refusal is the signal. Protocols that treat a stale or empty feed as a valid answer โ that read silence as a number โ get liquidated in the dark, and their users find out at 4 a.m. from a liquidation bot rather than a dashboard.
In 2020, I led a governance working group at MakerDAO and read more than five hundred voting proposals in a single stretch of DeFi Summer. The proposals that frightened me were never the ones with bad numbers. Those were easy; you can argue with a number. The frightening ones were the proposals with no numbers and confident prose. An author who has not measured anything will still write "risk is low" โ because the template has a field for risk, and a field with nothing in it looks like a mistake. The empty field is the most dangerous field, because it invites the writer to fill it with the average opinion of the last twelve months.
I have started calling this hallucinated analysis. In machine learning, a model given an empty context does not output nothing; it outputs the base rate of its training distribution. Ask it about a protocol it has never seen and it will describe a protocol that does not exist, fluently, in the correct format. The crypto research industry has exactly the same failure mode, and no context window to blame. Given a blank input, the analyst writes the median narrative of the cycle. Given a missing team, they write "anonymous but experienced." Given no tokenomics, they write "unlock schedule to be confirmed."
The nine-dimension report I was reading did something rarer. It kept the structure and emptied the claims. It even scored itself: technical value, one star out of five. Investment value, one star. Reference value, one star. And then, in the risk section โ where every other report in my inbox lists market risk, regulatory risk, competitive risk โ it listed a single high-priority item: input data missing, analytical chain broken.
That is an audit trail of ignorance. I have read few more honest artifacts this quarter.
The regulatory section of these templates is where the practice becomes actively dangerous. A Howey test with four rows โ money invested, common enterprise, expectation of profit, efforts of others โ is a genuine analytical instrument. Run it on a protocol you have actually read, and it will tell you something. Run it on a blank input, and you get four rows of the word "unclear," followed by a composite judgment of "cannot assess" that a hurried reader will skim as "not a security." I have watched this happen in real deal rooms. The Tornado Cash precedent made the stakes plain: when a court can treat the publication of code as the operation of a money-transmitting business, the difference between a careful legal read and a formatted one stops being academic. Precedent built on unexamined assumptions is precedent that outlives everyone who assumed.
There is a governance parallel here, and it is not decorative. In a DAO, the question of what to do with abstention is genuinely hard. Most contracts treat a non-vote as a no โ which is fine until turnout collapses and a quorum of apathy quietly ratifies a treasury transfer. When I designed the governance structure for CivicChain last year, a DAO handling municipal data sovereignty, we spent six weeks on precisely this problem: how does a system record the state of not knowing? The answer we settled on was a null state that halts execution rather than defaulting to either yes or no. The same principle governs the data itself. For a municipal record, an absent entry must never resolve to "no such record exists." Absence and negation are different states, and conflating them is how systems lose their legitimacy.
Regulators understand this better than most crypto natives, incidentally. The entire structure of disclosure law is built around the difference between a statement and an omission. An omission is not a lie, but a system that converts omissions into statements is a lying system. That is what a template does when it forces a verdict into every cell.
And here is the part that unsettles me most. I do not think the authors of these reports are dishonest. I think they are trapped. The deliverable is the unit of payment. Nobody pays for "I could not determine this." The incentive gradient runs one way: toward a filled page. So the blank input arrives, and the honest response โ stop, request source material, refuse to publish โ costs the analyst the engagement, while the dishonest response costs nothing except the reader's money.
Which brings me to the bear market, where the cost stops being abstract. In a bull market, hallucinated analysis is a victimless crime; everything goes up, everyone is a genius, and a confident paragraph is worth as much as a correct one. In a bear market, a reader makes a survival decision on the basis of a filled-in table. They stay in a pool because a report rated its risk two stars out of five. They hold a token because someone wrote "team: experienced" about people who had already left.
In the past eighteen months I have made it a habit to open every research report twice: once to read the argument, and once to count how many sentences contain a number that could be independently verified. The ratio is usually below one in ten. In a market where a single mispriced collateral ratio can cascade through every lending market on three chains, that ratio is not a stylistic preference. It is the difference between analysis and decoration.
So the counter-intuitive claim: the empty report is more valuable than the full one. Not because emptiness is virtuous, but because it preserves the only thing that lets a reader act โ the boundary between what is known and what is assumed.
But I want to test that claim rather than preach it, because there is a blind spot in my own enthusiasm. Refusal can be abdication. "DYOR" is the crypto industry's most reliable way of converting its own ignorance into the reader's responsibility. A report that says insufficient information on every line is, functionally, a report that says: figure it out yourself. That is defensible when the analyst genuinely had no source. It is cowardice when the analyst had a source and did not want to read it.
The deeper problem is the template itself. A framework that can run to forty pages on zero input is not measuring the world; it is measuring its own shape. Nine dimensions, forty-one tables, a star rating โ these are not instruments of discovery. They are instruments of formatting. The conclusion is predetermined by the skeleton, and only the adjectives vary. That is the definition of a derivative clone: identical in structure, hollow at the center. I have spent five years arguing that on-chain artifacts deserve to be treated as documents rather than commodities, and the same standard applies here. A report is a document. It should carry the fingerprints of the person who made it, including the fingerprints of their uncertainty.
What would it change if "insufficient information" were a first-class output โ a respected result rather than a failure to deliver? I think we would discover that most of the certainty in this market was never earned. We would lose some threads. We would lose some engagement. And we would finally be able to tell the difference between an analyst who has looked and an analyst who has merely formatted.
I have kept the document. It sits in a folder I have labeled, without irony, honest work, next to the dissenting essay I published at MakerDAO in 2020 โ the one arguing that algorithmic neutrality often masks systemic bias. They are the same document, written six years apart.
The next cycle will not be won by whoever writes the most confident table. It will be won by whoever can tolerate the void long enough to see what is actually in it. Curating the soul in a world of derivative clones begins, always, with the refusal to clone.