Last week a fund handed me a research deliverable. Nine analysis dimensions. Fourteen tables. A color-coded completeness scorecard, and at the top of it a red badge reading: information integrity — severely insufficient.
Every cell in the technical column: N/A. Every cell in the tokenomics column: N/A. Team and governance: N/A. Market and float: N/A. The document closed with four sentences that amounted to a refusal — no valuation framework, no risk rating, no recommendation, because there was no primary source to rate.
That is the most honest document I have read this cycle. It is also, statistically, the rarest.
Context
The 2026 bull market has industrialized research. A nine-figure seed round now ships with a forty-page due diligence framework attached as a marketing asset. The framework has nine dimensions because nine sounds thorough; the dimensions have names like "ecosystem positioning" and "narrative-to-fundamentals realization," which are not measurable quantities but do fill a page. Nobody asks what the dimensions are for. They are for producing a document that contains dimensions.
I spent 2017 doing the opposite. I traced 4,200 lines of the Geth transaction pool hunting memory leaks, found three that would have degraded the client under load, and submitted patches. The community gave me nothing — no grant, no acknowledgment, no reply. What I got instead was a working definition of evidence that has governed everything I have written since: a claim is worth exactly as much as the artifact behind it. You didn't "improve client stability." You changed a map's eviction policy and linked the commit.
That definition is now unfashionable. The current unit of research output is the slide. The current unit of credibility is the logo on the cover. And the current unit of accountability is nothing at all, because nobody signs the numbers.
So when a document arrives with every field marked N/A and a warning that N/A does not mean low risk, I read it twice. It is making an argument that the market is structurally unable to hear.
Core
The first thing to understand about the template is that it is a fabrication machine. A form with empty cells wants to be filled. That is not a moral failing of the analyst; it is a property of the artifact. Give a junior researcher nine dimensions and a deadline and the path of least resistance runs through interpolation — a comparable here, an assumed unlock schedule there, a "team appears experienced" that survives review because nobody can disprove it either.
The blank-field document resists this. It says: the token supply distribution is unknown, therefore the float is unknown, therefore the dilution schedule is unknown, therefore no valuation multiple can be constructed. Each "therefore" is load-bearing. Remove one and the whole chain of inference collapses into vibes.
N/A is not a risk rating. It is a statement about the analyst's posterior distribution — and in a market with adverse selection, the posterior should be grim. Base rates matter here. Projects that publish no supply schedule are not a random sample of projects. Projects that decline to name a jurisdiction are not a random sample. The act of withholding is itself informative, and the information points one direction.
I learned that arithmetic in 2022, mapping the Terra de-pegging. The causal chain ran back through Anchor's yield subsidy to a single large liquidity withdrawal, and $40 billion of market value evaporated in a sequence that was visible in the parameters months before it was visible in the price. The missing field in every Terra model I reviewed was the same one: no circuit breaker. Not a wrong assumption — an absent control. The analysts had built elaborate spreadsheets around a protocol whose failure mode was documented in its own code, and the code lost. Logic doesn't care about your roadmap.
The second property of the template is that it hides which blanks are dangerous. Some N/A fields are cosmetic; nobody's thesis dies because the GitHub contributor graph is unavailable. Others are terminal. I use a crude triage now, borrowed from stress-testing: for each missing field, ask what happens to the position if that field is at its worst plausible value rather than its median. Compound's interest model looked arithmetically clean in 2020 until I ran 10,000 leverage scenarios through the compounding logic in Python and found a rounding behavior that could, under high volatility, produce a yield path nobody had priced. The field wasn't blank — it was filled with a number that rounded to something harmless. That is worse. A blank is a question. A rounded zero is a lie with a decimal point.
The third property is scale. In 2026 the analyst is increasingly a model, and the model has no concept of provenance. I spent part of last year testing an AI trading agent wired into Chainlink feeds; the agent's decision layer was sound and its execution layer was sound, and its inputs came from a node that had been compromised. It traded confidently on corrupted data for eleven hours. Nobody had filled in the field labeled "feed provenance," because the template did not have that field. The agent inherited the analyst's blind spot and executed it at machine speed. AI does not create new failure modes in this industry; it amortizes the old ones across more capital.
The fourth property is that only one control actually works, and it is boring. A citation column. Every factual claim carries a pointer to an artifact — a commit hash, a contract address, an on-chain transfer, a signed disclosure, a timestamped governance post. In a proper pipeline those pointers get IDs: P1, P2, P3. A conclusion that cites P7 and P12 is auditable. A conclusion that cites nothing is a mood.
I rebuilt my own workflow around this after the Axie bridge work in 2021. I had reverse-engineered the interaction paths and found a gas optimization in the bridge contract that opened a reentrancy window during high-traffic periods. I disclosed responsibly. The team ignored it until I published a minimal reproducible proof of concept. The patch took two weeks. The exploit wasn't clever. It was unexamined — and it stayed unexamined because due diligence in that ecosystem consisted of reading the audit badge, not the contract.
Which brings the argument home. The nine-dimension template, filled in or not, is a document about a document. The only artifacts that matter are the ones that exist independently of the team's willingness to describe them: bytecode, verified source, admin key topology, timelock parameters, unlock cliffs encoded in the vesting contract, treasury outflows, the actual concentration of the governance token across the top twenty addresses. Every one of those is queryable. Every one of them is more informative than a paragraph about "vision."
So the correct reading of a report full of N/A is not that the research failed. It is that the research succeeded at the only task available: it refused to convert absence into confidence. That refusal has a price — I have watched honest memos get filtered out of allocation meetings because they lacked a number to anchor on — and the price is why the market produces so few of them.
Contrarian
The bulls have a point, and it is not the one they usually make. Their real argument is that research is a product, and products that say "insufficient data" do not sell into a market paying for conviction. That is factually correct. Demand for nulls is near zero, so supply of nulls is near zero, so the honest analyst who publishes nothing gets paid nothing and eventually publishes something. Greed is the feature; the bug is just the trigger.
The blind spot is assuming that because nulls don't sell, they aren't valuable. They are — they're just mispriced. A null is a tradeable signal precisely because the crowd cannot see it. When a project's disclosure is empty on exactly the fields that determine solvency — float, unlock schedule, admin key custody, treasury runway — the absence is a leading indicator that front-runs the price by months, and it is nearly free to acquire, because nobody is bidding for it. The information asymmetry in this industry is not that some people know more. It is that most people cannot tolerate not knowing, and so they buy a story to fill the hole.
That asymmetry is also the reason I don't publish confident verdicts on projects whose artifacts I cannot reach. You didn't get an opinion from me. You got the shape of the hole.
Takeaway
The next time a deck lands on your desk, count two numbers: how many fields are sourced, and how many are asserted. The ratio is the only due diligence metric that has ever predicted anything for me. If the sourced count is zero, you are not reading research. You are reading a résumé for a position that does not exist — and the question worth asking is not whether the project deserves your capital, but who, exactly, signed the page you're about to trust.