A founder I respect sent me a research document last week. Nine sections. Clean typography. A risk matrix, a tokenomics table, an ecosystem map, a regulatory assessment, a narrative-sustainability model. The kind of artifact that gets forwarded through a fund's group chat at 2 a.m. and quietly decides where eight figures of capital lands.
Every field in it said the same three characters: N/A.
I want to be precise about what that document was, because the details carry the whole argument. It was not the work of a lazy analyst. It was the work of an honest one โ a researcher, human and machine in the same loop, who had been handed an empty brief and refused to fill the vacuum with invention. The pipeline upstream had failed to deliver a single verifiable fact: no title, no source, no project name, no data point. And instead of confabulating a token model and a competitive moat to match the beautiful skeleton, the machine wrote down, over and over, in every dimension where a real analysis would normally live, that it could not proceed.
The most honest document produced in this bull market is a page full of the word "insufficient." That sentence has been sitting in my chest for a week, and I want to spend the next several thousand words unpacking why it frightens me far more than any scam whitepaper ever has.
Here is the fear. In 2017, the danger was a document that said too much and meant none of it. OneCoin's marketing, the anonymous ICO decks with the same three stock photos of a handshake, the token that promised to "disrupt logistics" and delivered a Telegram channel. We learned to distrust abundance. We built instincts for the smell of a whitepaper written by a copywriter instead of an engineer. What we never built โ what almost nobody in this industry built โ is an instinct for the opposite failure: the confident framework wrapped around nothing. The report that reads like diligence, that satisfies the eye, that passes the skim test, and that contains, underneath its polished structure, not a single fact you could trace back to reality.
That failure is now the dominant failure mode of the bull market, and the empty report is the cleanest specimen of it I have ever seen.
Let me tell you why I recognize it, because I have spent fifteen years watching this industry translate itself โ from cryptography into marketing, from marketing into institutions, and now from institutions into machines that talk.
In 2017, I was finishing a degree in applied mathematics in Bonn and building a small Python tool called ChainLit. The premise was embarrassingly simple: take a whitepaper, strip the jargon, and produce a plain-language summary a nineteen-year-old with no finance background could read in five minutes. I distributed five hundred copies to university clubs across the region. The thing that surprised me was not how many projects were fraudulent โ I expected that. It was how many of them were structurally sound and still incomprehensible. The gap I was trying to close was not a gap of intelligence. It was a gap of translation.
I learned something in that year that I have never stopped using: the most dangerous document is not the one you cannot understand. It is the one you believe you understood. Comprehension is a feeling before it is a fact, and the industry has gotten extraordinarily good at manufacturing that feeling.
By 2020 I was at Aave as a junior community analyst, running weekly beginner workshops that pulled three hundred people a month. When the EIP-1559 confusion hit and the whole timeline was drowning in arguments about fee burning, I built a visual guide โ one page, no equations โ that got shared by more than fifty people with real reach. What I took from that period was not a lesson about tokenomics. It was a lesson about anxiety. People don't need more data when they're scared. They need one true thing, clearly told. Trust is not built by the volume of information you provide. It is built by the proportion of it that survives contact with reality.
Then came 2022, and the FTX collapse, and the despair that followed it. I founded Resilience DAO โ twenty mentorship sessions, fifty people placed back into work, senior developers paired with juniors who had watched their first employer evaporate. I have written before that community is the only chain that cannot be broken, and I meant it then in the emotional sense. I mean it now in a more structural sense that I will get to.
In 2024, after the Bitcoin ETF approval, I found myself on the other side of the table, designing a crypto-literacy program for a hundred senior bankers at a major institution. Their questions were not stupid. They were the questions of people who had learned, correctly, to distrust anything that could not survive an audit. And in 2025, as AI agents began actually touching blockchains, I ran a human-centric AI initiative and published a manifesto arguing that code has to reflect human values. I thought I was writing about ethics. I was, without realizing it, writing about the empty report.
Because here is what the empty report taught me, and it took the whole of that history to hear it clearly: we have spent a decade building the wrong verification layer.
Let me be concrete, because this is where the essay either earns its length or wastes it.
When you see a number on a dashboard โ a TVL figure, a daily active user count, a fee-revenue chart โ you are not seeing a fact. You are seeing the end of a pipeline, and pipelines degrade. The data that ends up in your terminal travels through indexers that parse blocks and turn them into queryable tables, through RPC providers that stand between your query and a node, through subgraphs and ETL jobs and aggregation heuristics and, increasingly, through a model that summarizes the aggregation for you. Every one of those layers is a place where truth can be lost, and almost none of them is a place where a normal user can check.
I have audited enough of these stacks to know the shape of the failure. A subgraph gets a decimal wrong and every downstream dashboard inherits the error. An RPC provider silently serves stale state for a few minutes during congestion and a liquidation bot makes a decision on a number that was true two blocks ago. An oracle reports a price that reflects the last trade rather than the market. And at the very top of the pipeline, a language model looks at a table with missing rows and, because it was trained to be helpful, helpfully fills them in.
The empty report is the rarest artifact in this industry: a pipeline that failed loudly instead of failing silently. Everything I love about it is everything the rest of the market refuses to do. It said "insufficient" where a thousand other tools would have said "approximately." It drew the boundary of its knowledge in bright ink instead of smudging it. And I would trust that document โ the one with no answers โ over a polished deck with all the answers, every single time, because it is the only one that told me the truth about itself.
Now let me take you through the specific corners of this market where the confident framework has replaced the fact, because the abstract version of this argument is easy to nod at and hard to act on. I want to be specific enough to be wrong.
Start with data availability, because it is the clearest case of a solution that has outrun its problem.
Post-Dencun, every rollup on Ethereum has access to blobs โ cheap, temporary data space with its own fee market, introduced to make posting data to the base layer dramatically cheaper than the old calldata approach. The cost of operating a rollup collapsed, and the industry correctly celebrated. Then a small army of dedicated data-availability layers appeared โ Celestia, EigenDA, Avail, and a dozen others โ each promising to sell rollups cheaper or more scalable data space than Ethereum's own blobs could provide. The pitch was compelling. The market rewarded it.
And here is what the actual numbers say, from the rollup data I have pulled apart over the last year: the overwhelming majority of rollups do not produce enough data to need a dedicated DA layer at all. We are talking about chains posting a few hundred kilobytes per block in the busiest periods and a fraction of that in the quiet ones. Ethereum's blob capacity, even at target, has spent most of its life substantially underutilized. You do not need a dedicated shipping lane to move a bicycle, and yet an entire category of infrastructure has been built and funded as though every rollup were a freight network.
I am not saying DA layers are worthless. I am saying the narrative got ahead of the demand by an order of magnitude, and the gap between the two is exactly the kind of space where a confident framework thrives and a fact would starve. The DA market is not a demand story dressed up as a supply story โ it is a supply story waiting for a demand that most of its customers will never generate. When you read a pitch that opens with the theoretical throughput of a DA layer, ask the only question that matters: how many kilobytes does the customer actually post per block? Usually, the honest answer is fewer than the deck implies, and sometimes the honest answer is N/A.
Move to the application layer, and the same pattern repeats with better marketing.
Uniswap V4's hooks are, genuinely, one of the most elegant pieces of design work in the recent history of DeFi. They turn the exchange into a singleton contract with programmable extension points, so that a developer can attach custom logic to a liquidity pool โ dynamic fees, limit orders, custom oracles, whatever the mind invents โ without forking the protocol. The architecture is a set of LEGO bricks for market structure, and when I first read the design I felt the same electricity I felt reading about automated market makers for the first time.
Then I tried to explain it to a workshop of beginners, and I watched the electricity go out of the room.
V4's programmability is also its complexity tax, and that tax will quietly scare off the vast majority of developers who ever thought they wanted to build on it. Hooks are not a feature you switch on. They are a security surface you inherit. A hook that miscalculates fees can drain a pool. A hook that reads a price the wrong way can be arbitraged into oblivion. A hook with an admin function nobody documented is a backdoor with a friendly name. The number of developers who can write a hook that is both innovative and safe is a small fraction of the number who can write a basic smart contract, which is itself a small fraction of the number who can read one.
This is the empty-report problem in miniature. A hook is a beautiful framework. Whether it contains anything trustworthy depends entirely on the discipline of the person who filled it in, and the market has no reliable way to tell a well-audited hook from a confident one. The structure passes the skim test. The substance is invisible until it fails.
And notice how the same logic applies to the analytics that surround these protocols. Every dashboard now offers a "risk score" for a pool or a vault. I have traced several of those scores back to their sources, and a disturbing number of them are heuristics dressed as measurements โ a formula that looks rigorous, fed by inputs that were never verified, outputting a number that a user treats as a fact. The score is a framework. The data underneath it is often N/A. The user cannot tell the difference, and that is the entire problem.
Now cross-chain, where the gap between the pitch and the lived experience is widest of all.
Dencun lowered the cost of moving value between rollups, and the industry has spent two years telling itself a story about interoperability. Intent-based bridges, solver networks, message-passing protocols โ the architecture has genuinely improved, and the fees have genuinely fallen. I will give the builders that.
But I withdraw from my bank account in Frankfurt and it appears in a friend's account in Lisbon in seconds, with a reference number, a receipt, and a phone number to call if it goes wrong. I try to move an asset from one rollup to another and I get a spinner, a hash, a countdown that sometimes counts up, and a Discord channel where I am told to "be patient." The cross-chain user experience is still orders of magnitude worse than withdrawing from a centralized exchange, and no amount of cheaper data has closed that gap, because the gap was never about cost โ it was about legibility. A user cannot see the state of a cross-chain transfer. The interface shows a mood, not a fact. It is, functionally, an empty report that has been given a loading animation and asked to keep smiling.
I have watched a friend โ a competent, technical person โ lose an afternoon and a meaningful amount of money to a bridge that reported "pending" for six hours and then "failed" with no explanation. There was no missing data in the technical sense. Every block was on-chain. There was simply no honest translation layer between the machinery and the human, and in its absence the human invented a story to fill the silence: it's probably fine, it's just slow, the team knows what they're doing. That story is the confident framework. The reality was N/A.
Now let me turn to the layer where this stops being an analogy and becomes a literal mechanism, because it is the part of this essay that most directly concerns the next five years.
In 2025, as AI agents started interacting with blockchains in earnest, I led a human-centric AI initiative and organized a summit of a thousand people to argue about how to embed ethical constraints into smart contracts. We published a manifesto on algorithmic accountability. The core claim was simple: code must reflect human values, or it will reflect whatever values are cheapest to encode. What I did not fully appreciate at the time was that the most urgent ethical problem would not be an agent doing something malicious. It would be an agent doing something confident.
A language model asked to analyze a protocol will, absent strict discipline, produce an analysis. It has read millions of whitepapers and audit reports and governance proposals, and it knows the shape of a complete answer. Hand it an empty brief and it will not, by default, hand you back an empty report. It will hand you a plausible one, because plausibility is what it was optimized to produce. The token model it invents will have a supply curve that looks reasonable. The team assessment it writes will describe a founding team that sounds like every other founding team. The risk matrix will have entries in every cell, because a matrix with empty cells looks unfinished, and the model has learned that unfinished is bad.
An AI agent with no data and a mandate to be helpful will manufacture diligence, and the manufactured version is indistinguishable from the real thing at a glance โ which is precisely what makes it dangerous. The empty report we received was remarkable not because a machine failed, but because a machine refused. Someone, somewhere in the pipeline, had written the constraint that honesty outranks completeness. That constraint is now the single most important design decision in crypto research, and almost nobody is making it deliberately.
I have started calling this the confabulation economy. It is the market in which the supply of confident analysis has become effectively infinite and the supply of verified fact has stayed roughly constant, and the price of the first has collapsed toward zero while the price of the second has quietly become the only thing worth paying for. When anyone can generate a nine-section due-diligence report in ninety seconds, the report stops being the product. The verified fact underneath it becomes the product. The audit becomes the product. The person who actually ran the node becomes the product.
This is why I keep coming back to the empty report. It is the tell. It is what separates a real analyst from a text generator wearing an analyst's coat, and the difference is not intelligence. It is the willingness to write N/A.
Let me now do the thing I promised and turn this into a pragmatism test, because I do not want to leave you with a philosophy. I want to leave you with a tool.
Here is the test I now apply to every piece of analysis that crosses my desk, whether it was written by a person, a machine, or the two of them in a room together. I call it the information-gain test, and it has exactly one question: what does this document tell me that I could not have guessed? Not what does it cover, not how many sections it has, not how professional it looks โ what does it actually add to my model of the world? A document that covers nine dimensions and adds nothing has a gain of zero. A single sentence that corrects one of my assumptions has a gain of one. I will take the second over the first every time, and the entire research industry is organized around selling me the first.
Apply the test to the empty report. Its gain is, counterintuitively, high. It told me something I could not have guessed: that the pipeline upstream was broken, that no verifiable anchor existed, that anyone acting on the framework would be acting on nothing. That is real information. It changed my behavior โ I stopped, I checked, I did not invest. A polished report with invented numbers would have told me nothing true and changed my behavior toward a loss.
Now apply the test to most of what you read this week. The thread that summarizes a token you already understood. The video that covers a protocol you already knew about. The newsletter that rephrases the news you read yesterday. Nine sections, gain of zero, and it feels exactly as satisfying as the report that had something in it. That is the trap, and it is not a trap of ignorance. It is a trap of satiety. We are not starved for information in this bull market; we are drowning in the feeling of being informed, which is a different and more dangerous state.
And this is where the contrarian edge of the argument lives, so let me sharpen it until it cuts.
The comfortable position, the one this entire industry holds, is that our problem is a shortage of good data and a surplus of noise, and that the solution is better tools to surface the signal. I want to argue the opposite, and I want to argue it precisely. Our problem is not that we lack data. On-chain data is the most abundant, most transparent, most permanently recorded dataset in the history of finance. Every transaction is there. Every contract is there. Every wallet is there. We are standing in an ocean and complaining of thirst.
Our problem is that we have built an entire economy on top of that ocean that converts facts into frameworks and then sells the frameworks as though they were the facts. The DA layer that sells capacity nobody uses. The hook that sells programmability most developers cannot safely wield. The bridge that sells interoperability through a spinner. The risk score that sells certainty derived from unverified inputs. The AI agent that sells diligence it manufactured. Each of these is a confident framework wrapped around a hollow core, and each of them passes the skim test, and that is why they win.
The pragmatism test, applied honestly, asks: if this framework vanished tomorrow, what decision would change? If the answer is none โ if the dashboard, the score, the nine-section report, the loading spinner all could disappear and you would make the same choices โ then you were never using it. You were being comforted by it. And a bull market is, above all else, a machine for manufacturing comfort.
I will go one step further, because I think the contrarian claim is stronger than I have made it. I think the empty report is not merely honest. I think it is a prototype of the infrastructure we actually need. Not a document that refuses to answer, but a class of tools that are structurally incapable of inventing an answer โ that inherit the discipline of the empty report by design, that render the boundary of their knowledge as a visible feature rather than a hidden flaw. The most valuable thing anyone could build in this cycle is not a better dashboard. It is a dashboard that says N/A.
What would that look like? It would look like an analytics layer that marks unverified inputs in red instead of averaging them into a clean number. It would look like a bridge that shows you the actual state of your transfer โ which step, which signer, which block โ instead of a mood. It would look like a risk score that tells you how confident it is and why, and refuses to produce a score at all when its inputs are empty. It would look like an AI agent whose first reflex, on receiving an incomplete brief, is to write the word insufficient. None of this is technologically hard. All of it is culturally rare, because the market pays for confidence and punishes honesty, and that is the deepest mispricing in this entire asset class.
Let me say the quiet part as plainly as I can, because I have spent this whole essay circling it. The only chain that cannot be broken is the chain of people who verify each other, and we have spent a decade building every other kind of chain while neglecting that one. Community is the only chain that cannot be broken โ I have written that line for years, and I have meant it mostly as a statement about resilience in the bear market. I mean it differently now. I mean it as a statement about verification. A protocol cannot verify itself. A dashboard cannot verify itself. An AI agent, least of all, cannot verify itself, because its entire failure mode is that it does not know when it is wrong. Only a community of humans, checking each other's work, reading each other's code, refusing each other's confident frameworks, can perform that function. That is not a soft claim about belonging. It is a hard claim about the only verification layer that scales.
And this is where I want to bring in the part of my own history that I have not fully reckoned with until now.
When I built ChainLit in 2017, I thought I was building a translation tool. I was wrong, or at least incomplete. I was building a verification tool that happened to work through translation. The students I handed those five hundred summaries to did not need to understand cryptography. They needed to be able to tell the difference between a project that had done the work and a project that had written the marketing. The plain-language summary was a way of stripping away the framework and seeing what was left. When nothing was left, they knew. That is the entire trick, and I have been rediscovering it in every role since.
At Aave, translating EIP-1559 into one page was the same move: strip the framework, show the mechanism, let people see what was actually happening to their fees. At Resilience DAO, the mentorship sessions were the same move applied to careers: strip the narrative of the bull market, show people the actual state of the labor market, help them see what was real. At the bank, teaching a hundred executives about custody was the same move yet again: strip the mystique, show the mechanism, let them decide whether they trusted it. And at the AI summit, arguing about ethical constraints in smart contracts was the same move at the highest level: strip the illusion of neutrality, show the values that were actually encoded, let people choose.
Every single time, the value I added was not knowledge. It was legibility. It was the act of refusing to let a confident framework stand in for a fact. The empty report is the purest expression of that instinct I have ever encountered, and it arrived from a machine, which is either the most hopeful or the most unsettling thing about this whole episode depending on how you look at it.
Let me look at it both ways, because I think both are true.
The unsettling reading: we have built machines that can generate confident frameworks faster and more convincingly than any human, and the market has not yet developed antibodies against them, and the empty report is a rare mutant in a population that overwhelmingly favors the confident lie. In that reading, the empty report is a warning about how bad things are about to get.
The hopeful reading: we have built machines capable of honesty, if we choose to instruct them toward it, and the empty report proves that the constraint is available โ that a system handed nothing can be told to return nothing, and that this is a design choice rather than an inevitability. In that reading, the empty report is a demonstration of what disciplined tooling could look like if we valued it enough to build it.
I hold both. I think the next five years will be determined by which reading wins, and I think the outcome depends almost entirely on what we, the community, choose to reward. If we reward confidence, we will get confident frameworks wrapped around hollow cores, at machine speed, at infinite scale, until the entire information layer of this industry is a beautiful, coherent, load-bearing nothing. If we reward honesty โ if we pay for the audit instead of the summary, if we share the N/A instead of the number, if we build reputations on being right about what we do not know โ then we get something else. We get an industry that can actually be trusted, because trust is not a feeling. Trust is the accumulated record of being honest when dishonesty was easier.
I keep coming back to a conversation I had at the bank, with an executive who had spent thirty years in traditional finance and was deeply, correctly suspicious of everything I was showing him. He asked me the only question that ever matters, in the plainest possible terms: how do I know this number is real? And I realized, standing there, that every good answer I had was a version of the same answer: because someone independent checked it, and because the checking is public, and because if it were wrong, the community would catch it and the record would persist. That is the whole of it. That is the entire verification layer, reduced to a sentence a banker can understand.
It is also the sentence that most of this market is failing to earn right now.
So let me end where I began, with a document full of nothing, and tell you what I actually did with it.
I did not delete it. I did not laugh at it. I forwarded it, with a single line of my own, to three people who make real allocation decisions, and I wrote: this is the most trustworthy thing I have read this month, because it refused to lie to you. Two of them understood immediately. The third asked me, with genuine confusion, what was the point of a report with no analysis in it. And I realized that the third person had just shown me the whole problem, standing there in a single question, because the point of the report was exactly what he could not see: that there was nothing there to analyze, and that a framework which had concealed that fact would have been worse than useless โ it would have been expensive, and quiet, and believed.
Community is the only chain that cannot be broken, and the first link in that chain is the discipline to write N/A when N/A is the truth. Everything else โ the audits, the dashboards, the bridges, the agents, the frameworks โ is downstream of that single decision, and the decision is not technical. It is moral, and it is available to every one of us, right now, in a bull market that is doing everything in its considerable power to make us forget it.
The next time someone hands you a beautiful nine-section report with an answer in every cell, ask them one question before you read a single number: what did you write in the fields where you did not know? If they hesitate, you have your answer. If they show you the N/A, you have found something rare โ a person, or a system, or a community, that has chosen to be a chain that cannot be broken.
Which kind are you building?