Four thousand words crossed my desk last week. Nine analytical dimensions. Thirty-one tables. A risk matrix with six rows, a token supply schedule with four tranches, an ecosystem diagram in three layers. Every single field returned N/A. Not one number survived the page.
The document was not lazy. That is what unsettled me. It was disciplined — it refused to invent a supply figure, refused to guess a TPS, refused to run a Howey test against a legal entity nobody had bothered to name. Around the fourth page I stopped reading it as a memo and started reading it as a specimen. A genre had appeared while I was looking elsewhere: the null report, an artifact engineered to carry the exact silhouette of diligence while containing none of its inputs.
Nineteen years of reading these things, and I had never encountered one that was simultaneously more honest and less useful than everything else in my inbox.
Crypto research has a short and not especially flattering archaeology. 2017 produced the Telegram thread: six paragraphs, one chart, an accusation. 2020 produced the Dune dashboard — permissionless, reproducible, and instantly gamed by whoever chose the window that flattered the thesis. 2021 produced the research desk, a staffing pattern lifted wholesale from equity research, where two or three analysts generate PDFs that justify positions the PM took on a hunch. Post-2024 produced the template.
Each generation traded verifiability for throughput. The 2017 thread was roughly ninety percent checkable — you could open Etherscan and confirm the claim in under a minute. The 2026 template is zero percent checkable and one hundred percent well-formatted, and it lands in the inbox in four seconds because a model rendered it against a schema. Volume rose by three orders of magnitude. Independent confirmation fell to nothing.
The schema itself is not stupid. It descends directly from TradFi venture diligence: technical, tokenomic, market, ecosystem, regulatory, team, risk, narrative, supply-chain transmission. Nine boxes. Someone observed that every good memo contains all nine and inferred that filling all nine produces a good memo. It is the same category error as assuming that because every successful DeFi protocol ships a governance token, bolting on a governance token produces a successful protocol.
I have committed that error. In 2021 I shipped a 15,000-word NFT investigation — twelve founder interviews, 50,000 secondary transactions — and I know precisely which 2,000 words did the intellectual work and which 13,000 existed to make those 2,000 legible to an investment committee. The scaffolding is not research. It is a delivery mechanism, and delivery mechanisms get mistaken for substance whenever the reader is short on time.
Here is the structural point, and it is the one that matters. A template that can be filled entirely with N/A has no falsification pressure built into it. That describes the schema, not the analyst who executed it.
What the null report actually got right is that its input sat three derivatives deep. A news article is derived from a project announcement, which is derived from a protocol change, which is derived from an incentive program that was itself designed to generate announcements. Every derivation strips verifiable artifacts and substitutes claims. By the time you reach the article layer, the only remaining substance is narrative — and narrative is precisely the layer where N/A is the correct, honest answer to a technical question.
My own method runs the opposite direction, and it started by accident. In early 2017 I spent six weeks reverse-engineering early ERC-20 implementations during the ICO frenzy and found a reentrancy flaw in a fundraising contract that had already absorbed $4.2 million in ETH. Reentrancy was not news even then. What made the thread land was that I could point at a specific line in a deployed contract. One artifact. The whole thing was 340 words and it outperformed everything my fund published that quarter.
So I run every research product, mine included, through a single number: verification density — the ratio of independently checkable artifacts to total assertions. That 340-word thread scored close to 1. The four-thousand-word null report scores exactly 0. A confident, well-sourced report that is wrong scores negative, because disproving it consumes reader hours and the error propagates straight into position sizing before anyone catches it.
Negative density is the disease. N/A is the immune response.
The next layer is where the template fails structurally. The nine dimensions do not cost the same to fill, and a schema that pretends they do cannot tell you where to spend your next hour. Market data is nearly free — funding, open interest, stablecoin flows, all public and timestamped. The risk matrix is cheap to write and expensive to verify, which is exactly why every crypto risk matrix in circulation is a taxonomy of categories rather than a register of findings. Team and governance is the expensive one: proving a DAO is not a three-address oligarchy means pulling the top-100 holder distribution, checking delegate overlap, and reading the last forty proposals for who actually authored them. That is a week of work, and it is the only week that changes the conclusion.
An all-N/A report is scale-free. It performs the shape of diligence without discriminating between cheap unknowns and load-bearing ones. A good memo spends seventy percent of its length on the two dimensions that are both material and currently unknowable, and dismisses the rest in a paragraph.
I have one comparison worth keeping. When I mapped LUNA's narrative decay across 500+ community channels in 2022, the deliverable was not "sentiment deteriorated." It was a date — the specific week when decentralization rhetoric stopped tracking on-chain behavior. One timestamp pulled from a mountain of noise, reproducible by anyone who re-ran the scrape. Verification density near 1.
In a market going sideways, that distinction stops being academic. When there is no trend to borrow conviction from, positioning is the entire product, and positioning requires knowing which questions are open versus which are closed. The null report at least tells you the questions are open. What it cannot tell you is which open question is worth answering.
The reflexive read is that the null report is a failure of research. I think it is the only artifact in the stack with a defensible edge, because it prices the thing everyone downstream is pretending not to know.
The mechanism is simple. A token with no verifiable supply schedule, no independently attested reserves, no revenue line you can reconcile against a deployed address does not trade on fundamentals. It trades on the liquidity and half-life of its narrative wrapper. Those are real, measurable properties — but they are different properties, with different sizing, different invalidation levels, different exit conditions. Desks blow up on this constantly, because they apply fundamental logic to a wrapper trade: they buy a story at fundamental size, then get stopped out by a narrative that was never load-bearing in the first place.
The hunt for alpha in the noise of the herd has always meant, operationally, one thing — knowing which questions cannot be answered yet. That is not the failure mode of research. That is its conclusion.
There is a second-order effect I find stranger. Crypto's information asymmetry has inverted. In 2017 the scarce resource was access. In 2026 the scarce resource is a signature — someone willing to attach their name to a claim and be wrong in public. The null report is perfectly rational for the analyst and ruinous for the reader, because every N/A is a risk the author declined to hold. Templates are, among other things, liability shields with good typography. The story behind the template, not just the ticker, is that nobody signed it.
So my forecast for the next cycle is not that AI research agents will fill these schemas faster. It is that verifiability itself becomes the marketing surface. Protocols that publish signed attestations, enforce unlock schedules in contract rather than in a Medium post, and expose revenue you can reconcile against an address will out-narrate protocols that publish decks — not because the market suddenly loves rigor, but because the wrapper has become expensive to trust and nearly free to check.
The story behind the token, not just the ticker, began as advice about reading team backgrounds. It is now a technical instruction: locate the one artifact that cannot be fabricated, and read the rest of the project through what it says.
Which brings back the question I could not shake. If you deleted every number from your last research report, would any conclusion change? If the answer is no, you did not produce analysis. You produced a null report with better formatting.