A research brief landed in my inbox last Tuesday. Nine sections. Technical architecture, token economics, market structure, ecosystem position, regulatory exposure, team and governance, risk matrix, narrative expectation, transmission effects. Every heading present, every subheading formatted, every comparative table drawn.
Roughly sixty percent of the cells read: N/A.
By any reasonable standard, that is a failure of output. I read it twice. Then I saved it, because it was the most truthful piece of crypto research I had received that month.
Here is the uncomfortable corollary. Every other report I received that week — the dense ones, the confident ones, the ones stamped with four-out-of-five ratings and a constructive outlook — was built on inputs of comparable thinness. The difference is that those documents had learned to fill the blanks with prose. Formatting is cheap. Sentences are cheap. The blanks are the only honest part.
The nine-dimension template did not emerge from crypto. It was transplanted. Sell-side equity research has carried mandatory sections for decades — revenue models, competitive positioning, management quality — and those sections function in equities for one specific reason: the underlying disclosures are legally enforced. A public company that omits its supply chain gets a letter from a regulator. A protocol that omits its token unlock schedule gets a governance forum post with eleven replies and a Discord emoji.
So the container migrated. The enforcement stayed behind.
Between 2021 and 2023, research became an industry rather than a craft. Funds hired analysts to produce deal memos. DAOs hired contributors to produce diligence. Telegram groups sold deep dives as a subscription product. The template was the natural output, because a template is what you produce when you are paid by the unit and the input is thin. It signals completeness. It fills a page count. It survives review from someone who is also paid by the unit.
Then search formalized it. The 2026 ranking regime rewards information gain — new information the reader could not get elsewhere — but it measures that reward through engagement signals, and structure drives engagement faster than substance. A nine-section report with tables gets skimmed longer than a three-paragraph note that says I do not know yet.
In a sideways tape, the effect compounds. When there is no price action to narrate, research volume migrates from market calls to structure. Nobody can write the token is going up in a chop, so they write about the token's architecture instead. Templates are what you get when there is nothing to say and a word count to hit.
The core problem is measurable, and it belongs to information theory, not to finance.
Claude Shannon formalized the point in 1948: a message that is identical regardless of its input carries zero information. It has entropy, but no signal. Apply that test to any research framework and the verdict is immediate. If the nine-dimension template produces the same document for a forty-million-dollar L2 and for a blank file, then the template is not measuring the L2. It is decorating the blank. The bit count is identical. The reader cannot see it, because the blank got dressed in headers.
I have watched this failure operate in the other direction too, and that direction is more damaging.
In 2017 I ran security review for a bridge team during the ICO cycle. The automated scanners returned clean. Every box ticked, every severity level at zero, a report that looked finished. I spent the following week inside the contracts line by line and found three reentrancy vulnerabilities that the tooling had no vocabulary for, because they lived in the interaction between two functions that were individually correct. Trust is not a feature, it is a failed audit — and a clean scan is not evidence of safety. It is evidence that nobody has looked yet with intent.
The clean scan and the N/A report are the same pathology wearing different clothes. Both substitute structural completeness for knowledge. The difference is that the N/A report admits it.
Now map the mechanism onto the metrics that dominate this industry.
Liquidity mining APY is not a return. It is a subsidy expressed as a percentage. It measures the intensity of an incentive, not the health of a system, and it converts into TVL the instant you point a dashboard at it. Liquidity flows like water, but greed builds dams — and the dam is legible, quantifiable, and reportable in a table, which is exactly why the table keeps getting produced. When I spent the summer of 2020 tracking front-running bots on Uniswap rather than celebrating TVL growth, the two numbers were moving in opposite directions. Total value locked was climbing. The value actually accruing to the people the protocol claimed to serve was being extracted in the mempool before their transactions landed. A template with a market structure section would have captured the first number and never seen the second.
Governance is worse, because the template has a checkbox for it. On-chain voter turnout across major DAOs has sat below five percent for years. A proposal that passes with 4.1 percent participation is structurally indistinguishable from no proposal at all — the quorum exists, the signature exists, the execution exists, and none of it constitutes a decision. The framework asks whether a protocol has governance. Yes. Checked. What it cannot ask is whether the outcome would differ if the largest three wallets abstained. Transparency reveals the cracks that opacity hides, but only if someone reads the ledger instead of the summary of the ledger.
The same reflex appeared in the NFT cycle, where I spent weeks clustering wallets rather than admiring floor prices. Roughly eighty percent of trading volume in the major collections traced back to a small set of addresses trading with themselves. Every market-structure section written that year reported a volume figure the market had generated about itself. The number was real. The meaning was circular.
This is the part practitioners resist, so let me state it plainly. The reason templates persist is not laziness. It is that templates scale and reading does not. The three vulnerabilities I found in 2017 were found by attention, one function at a time, and that method does not parallelize. A framework can be handed to a junior analyst on Monday and returned on Friday. Deep reading cannot. Templates are a scaling technology for attention — and the industry mistook them for a scaling technology for knowledge.
Of the four thousand words in a typical brief, the section carrying actual information is the one where the analyst writes something you could not have derived from the project's own documentation. In practice, that is perhaps two hundred words. The other three thousand eight hundred exist to make those two hundred feel expensive.
And the most sophisticated version of this failure is not the empty report at all. It is the fully populated one. Give me the same template with every cell filled by an analyst who read carefully, and I will still find the same structural blind spots, because the blind spots are in the categories. No section in the standard nine asks what would have to be true for this to fail. Risk matrices rank likelihood and impact, which is a taxonomy of the known. They do not model the unknown, and in this asset class the unknown is where the money is made and lost.
Here is where the conventional reading gets it backwards.
The empty report is not the failure. It is the correct output of a correct process applied to an empty input. The failure is upstream — in a market that pays by the page and in readers trained to read formatting as rigor.
Consider what a language model does when you hand it a template and an empty source. It invents the token supply schedule. It fills the team section with plausible names. It rates the technology four out of five. We call that hallucination when a machine does it. When a junior analyst does it under a Friday deadline, we call it a first draft. The mechanism is identical. The only difference is that the human feels something about it, and the feeling rarely makes it into the deliverable.
There is a second-order effect worth naming. Institutional allocators have begun demanding these templates as a condition of engagement — compliance committees want a nine-section memo in the file regardless of what is in it. The template becomes a procurement requirement, and once it is a procurement requirement, its information content is permanently irrelevant. You are no longer buying analysis. You are buying a document that lets someone else sign off.
So the report with sixty percent N/A is the anomaly worth protecting. It is the only artifact in the stack that would survive an actual audit, because it makes no claims that could fail one.
Sideways markets are where positioning happens. That makes this precisely the moment to reduce section count rather than increase it — to prize the analyst who writes I have not verified the unlock schedule over the one who writes a confident paragraph about the unlock schedule they also have not verified.
So here is the test I now apply to everything that crosses my desk. Delete every sentence that could have been written before the file was opened. Delete the architecture overview, the market-context paragraph, the closing optimism. What remains?
If it is under two hundred words, nobody wrote research. Someone built a container, and the container was the product. Volatility is the price of admission to the future — but you only pay it once you know which positions you actually hold.
