At 09:47 UTC on a Monday, a nine-dimension "deep analysis" on a mid-cap DeFi token went live. It ran 2,800 words. It carried a Howey-test breakdown, a token-unlock schedule, a six-row risk matrix, a nine-point scoring rubric, and a three-line disclaimer. Every cell in the technical section read N/A — insufficient information. So did the tokenomics rows. So did governance, regulatory, and ecosystem. The report was immaculate in form and empty in substance. And here is the uncomfortable part: it was also, by the standards of its genre, honest.
I have spent seventeen years reading crypto research and the last three watching the form of it detach from the function. What I found while tracing the output of a dozen analysis-as-a-service pipelines is not a content-quality story. It is a structural one. The crypto research industry has industrialized the appearance of rigor while quietly outsourcing the actual thinking to a template. The N/A report is not a bug in that machine. It is the machine operating at its most transparent — and, as I will argue, at its most useful.
The economics changed first. In 2020, publishing a credible token analysis required a desk, a data subscription, a wallet tracer, and a human who could read a bytecode diff. By 2026 it requires a prompt. The marginal cost of producing a 2,800-word deep dive has collapsed to near zero; the marginal cost of producing a good one has not moved at all. That asymmetry is the entire story.
When output is free, output floods. Search rewards volume, so volume wins. The 2026 algorithm's demand for "information gain" was supposed to be the correction — but you cannot inject information gain into a pipeline that holds no information to begin with. So the pipeline does the next best thing: it generates the scaffolding of information. Headers, tables, confidence intervals, asterisks, a "risk grade." The reader sees structure and infers substance. The inference is wrong, but it is cheap, and cheap is what the feed optimizes for.
Chasing alpha through the summer heat of 2020, I learned that the only durable edge was primary data — the liquidation rate I scraped myself, the wallet I traced myself, the contract I executed against myself. That edge has not disappeared. It has been buried under ten thousand pages of well-formatted noise.
Let me deconstruct the machine, because the failure is mechanical, not moral. A templated framework has, by design, nine to twelve mandatory dimensions: technical, tokenomics, market, ecosystem, regulatory, team, risk, narrative, transmission. Each dimension demands inputs. When inputs are present, the framework is a genuine forcing function — it makes the analyst confront the uncomfortable cells. When inputs are absent, the framework does something more interesting. It produces output anyway, because the schema demands completion.
Watch the generation layer. Hand a model an empty information-point list and a nine-field schema and it faces a binary. Either it writes N/A into every field — the honest path — or it fills the fields with plausible content drawn from the token's public narrative. Confabulation is not a failure mode of these systems. It is the default completion behavior, and at the surface it is indistinguishable from analysis.
The honest N/A report and the dishonest filled report look identical to a skimming reader. Both have tables. Both have the word "risk." Only one has a non-empty information-point list underneath it. And the list is never shown.
Here is where my quant background starts twitching. Every one of these reports ships a risk matrix — technical, market, operational, regulatory, competitive, narrative. What it almost never ships is the input that would let you price any of those rows. The matrix is a picture of risk, not a measurement of it. It is the research equivalent of a Proof of Reserves screenshot: it proves a document exists, not that the claim inside it is true. I have audited enough of both to know the two are frequently unrelated.
Sprinting through the noise to find the signal means, first, discarding the reports that contain no signal. That is a harder filter than it sounds, because the noise is now optimized to survive filtering. It has the right keywords. It has the right structure. It even has the right humility — "N/A, insufficient information" reads as rigor to an untrained eye.
So build the counter-metric. I call it the confidence-to-signal ratio: the number of declarative, forward-looking claims in a report divided by the number of primary, verifiable data points behind them. A 2,800-word report with four confident predictions and zero wallet addresses has a ratio approaching infinity. It is pure narrative leverage. Trade on it and you become the exit liquidity for someone else's certainty.
When I reverse-engineered the UST death spiral in 2022, the entire edge was that the causal data was public and almost nobody was reading it. The peg mechanism had a circular dependency visible in the mint-and-burn ledger. That was a report with one primary input and enormous information gain. The N/A report is its inverse: infinite inputs promised, one thing actually said, which is "we did not look."
The forensic test is simple and runs in ninety seconds. Take any analysis and locate its information-point list. If the technical section cites no contract address, no commit hash, no audit reference, the report has no technical input. If the tokenomics section cites no unlock cliff and no treasury address, it has no tokenomics input. A dimension with no primary input is a dimension with no analysis, regardless of how many words sit inside it. The words are decoration. Tracing the code back to the genesis block of a report means asking where the first verifiable fact lives — and if the answer is "nowhere," you are reading a template, not a thesis.
In 2017 I bypassed press releases and audited the 0x v1 fill-order contracts directly, because the only claims worth trusting were the ones I could execute against. That instinct is now a defense mechanism. The volume of structured-but-empty output has made "did anyone actually look?" the single highest-value question in crypto research. It costs nothing to ask, and it kills most of the corpus on contact.
Consider the scoring rubric, the genre's most seductive artifact. A report will grade a token's governance "health" on a five-star scale, or its technical maturity on a ten-point index. The number feels like a measurement. It is a vibe with a decimal point. I have seen a governance score assigned to a protocol whose only governance event was a team-controlled multisig moving treasury funds — a score computed from no vote history, no proposal data, no participation rate. The rubric did not measure governance. It measured the analyst's willingness to fill a cell.
Here is the angle almost nobody takes, and it is the one that matters. We are trained to treat the N/A report as the failure. It is not. It is the control group. A pipeline that returns "insufficient information" across every dimension has told you something true and actionable: nobody holds primary data on this asset, and any conviction you carry is borrowed. That is a short thesis in itself.
The dangerous report is the one that fills the gaps. Confabulation is frictionless, and the market pays a premium for confidence, not accuracy. A desk that prints "N/A" loses the click. A desk that invents a supply schedule, a team résumé, a competitive moat — wins it. The incentive gradient points directly away from honesty, and the gradient is steep. This is the same structural flaw I keep finding in exchange transparency exercises. A snapshot proves a slice of a liability at a moment; it says nothing about the continuous ledger underneath. A filled-in report proves a narrative at a moment; it says nothing about the data underneath. Both are optimized for the screenshot, not the audit. From protocol wars to community traps, the pattern repeats: the artifact that looks like verification is precisely the artifact that resists it.
The next signal to watch is not a price level. It is the information-point list. Reports that publish their primary inputs — addresses, hashes, ledger queries — will start to out-trade reports that publish only their conclusions, and the spread between them is the cleanest alpha in a sideways market. So the question for every reader, every desk, every deep dive you open this quarter: when the schema demanded an answer and the data did not exist, did the report say so — or did it write something anyway? The market moves fast; the honest ones move slower and survive longer. Read the tape before the chart confirms it.