The Empty Table: Why Crypto Research in This Bull Market Is Ninety Percent Scaffolding

Last month, a founder I respect sent me a research report about his own protocol. Forty-one pages. Nine sections: technical architecture, token economics, market structure, ecosystem position, regulatory posture, team and governance, risk matrix, narrative durability, and supply-chain transmission. Beautiful typography. Every table had a header row, a footer row, and a source column.

And most of the cells said some version of not available.

I read it twice. The first time I laughed, because it was absurd — nine analytical dimensions, forty-one pages, and the actual analytical content was a rounding error. The second time I stopped laughing, because I recognized the shape of the thing. I have written that report. Not literally that one, but the same architecture: the confidence of the structure standing in for the confidence of the conclusion.

In August 2020 I put my entire personal savings — fifteen thousand Australian dollars — into a yield farm that had been live for eleven days and audited by nobody. The reason wasn't that I was reckless. The reason was that I had a framework, and I filled the framework in with hope. I built a spreadsheet with columns for APY, for lockup period, for "team doxxed," for "audit pending." I filled every cell. The cells were the problem. They made a guess look like a due diligence memo.

We didn't lose that money to a bug. We lost it to a table.

Six years later, the table has gotten bigger, and the bull market has made it load-bearing.

Where the nine-dimension template came from

The framework has a lineage, and it's worth knowing it. Between 2018 and 2020, crypto venture funds borrowed their diligence templates from growth-equity and public-markets research. That borrowing made sense at the time: allocators wanted a document structured like the documents they already trusted. Columns for market size. Columns for competitive positioning. Columns for team pedigree. The template migrated from internal memo to public "research report," and by 2021 it had become a genre — a genre with its own SEO conventions, its own Twitter distribution strategy, its own tier of writers paid per word.

I participated in that migration, and I'm not proud of all of it. In 2021 I was running a small NFT education platform and writing a weekly market map for a paying audience. Every week I filled in a grid: project name, chain, category, funding, "narrative strength." I filled in narrative strength with a word I had invented and could not defend. Nobody ever asked me to. The grid was the product. The grid got shared. The grid was, in retrospect, an advertisement for my own confidence rather than a description of reality.

Then three things happened in sequence. In 2022, the bear market purged the weakest analysts, but the template survived because templates don't have opinions. In 2023, points programs and airdrop farming created a sudden, enormous demand for pre-launch "coverage" — readers didn't want analysis, they wanted a reason to farm. And in 2024, with spot Bitcoin ETFs approved and institutional capital formally entering the asset class, allocators began demanding research coverage as a condition of allocation. Not understanding. Coverage.

That last substitution is the whole story. Coverage is a document. Understanding is a position. The industry now produces the first at industrial scale and treats it as evidence of the second.

Here's the asymmetry that keeps it running: publishing a report is cheap, and being wrong is expensive but delayed. In a bull market, the delay is long enough to be invisible. Every thesis appears validated because the output variable — price — went up for reasons that have almost nothing to do with the thesis. When the tide goes out, the analyst who wrote the empty table is not the one who pays. The reader is.

I saw this from the other side in 2024, when I was producing a podcast series putting Wall Street analysts in the same room as DeFi developers. The analysts asked about TAM, about revenue multiples, about comparable-company sets. The developers answered in good faith. Almost nobody in the room asked who controlled the upgrade key, or what happened to user funds if the sequencer went dark. Those questions didn't fit the format. Formats are very good at excluding the questions that would break them.

So let me spend the rest of this piece inside the cells that get filled in most confidently, and least truthfully. Not because the framework is evil. Because it's seductive, and because I know exactly how seductive it is.

Cell one: "decentralized governance"

Every template has this row. The scoring is almost always the same: token exists, governance forum exists, snapshot votes happen, therefore decentralized.

I have read the contracts of roughly thirty protocols that describe themselves as governed by their communities. In twenty-six of them, the upgrade path terminates at a multisig holding between three and seven signers. That is not a scandal. It is the normal case, and it is usually disclosed, somewhere, in a footnote of a forum post that nobody reads after launch week.

Here's the technical picture, because the picture matters more than the verdict. Most of these systems use a transparent proxy pattern: users interact with a proxy contract, the proxy delegates to an implementation contract, and the implementation can be swapped. The swap is authorized by an admin address. That address is usually a ProxyAdmin contract, and the owner of that contract is — nearly always — a multisig. Some teams also deploy a TimelockController with a 24- to 72-hour delay between a governance vote passing and execution.

That's where the table stops. Here's where the analysis should begin.

The timelock has a proposer role. Who holds it? In most deployments, the same multisig. The timelock also frequently has a canceller role, which can halt a queued proposal instantly — also the same multisig. And the protocol almost always has an emergency pause function on the core contracts, callable without a timelock, because a 72-hour delay on a pause button during an exploit is a liability. That is a defensible engineering decision. It also means the effective governance latency of the protocol is not 72 hours. It is the time it takes three people to open a group chat.

The real governance question is not "is there a token." It's: who can change the code, how many humans is that, are they identifiable, are they legally separable from the foundation, and how long does it actually take them — measured, not documented. That is five questions. It fits in one cell. It almost never appears in one.

I understand the counterargument: upgradeability is a feature, contracts are never finished, and a genuinely immutable protocol cannot respond to a novel exploit. Fine. But then say so. Put a number in the cell: three signers, all employed by the foundation, 90-day median upgrade latency over the last two years. That's honest. That's analyzable. "Decentralized: yes" is not.

There's a version of this I find almost funny. In the twenty-six deployments where upgrade authority sits with a small multisig, the governance token still does something real: it determines who gets paid. Fee switches, treasury grants, emissions schedules — those genuinely pass through votes, and the votes are genuinely contested. What doesn't pass through votes is the code. Which means the community has real authority over the money and no authority over the machine that counts the money. That's not decentralization with caveats. That's a different thing entirely, and the template can't express it.

Cell two: "decentralized sequencing"

I want to be fair here, because this is the cell where the industry has been most disciplined about lying to itself in public.

Since about 2022, most rollups have published a roadmap with a line item called "decentralized sequencer." It is typically three phases, with the third phase described as "permissionless sequencing with a leader election mechanism," and no date. For two years, that was the PowerPoint. It remains, in most cases, the PowerPoint.

Back in 2022, when I retreated into modular blockchain research during the worst of the bear market, I spent four months reading Celestia's early data-availability papers and writing long explainers about separating consensus from execution. It was the most intellectually alive period of my career. Here's what that research gave me that the price cycle never did: a vocabulary for asking where the trust actually lives in a system, rather than where the marketing diagram says it lives. Sequencing is the clearest example of that gap.

What actually exists on mainnet today is a single sequencer — one node, operated by the team, that orders transactions, posts batches to the L1, and occasionally proposes state roots. Everything about the user experience in a modern rollup is downstream of that one machine being alive and honest.

Escape hatches exist. Most rollups have a force-inclusion path: you send a transaction to an L1 contract, and the sequencer is obligated to include it within some window. This is a real mechanism and it has saved real users. But look at the dependency graph. If the sequencer is offline, the force path still requires someone to eventually process the queue — and in several deployments, the party responsible for processing that queue is the same entity that runs the sequencer, using different infrastructure that sits in the same cloud region with the same operator keys.

Then there's the data availability question underneath it. If your rollup posts state diffs to an external DA layer, good — that's a genuine decentralization claim and it deserves credit. If it posts to a committee, ask who's on the committee. I have yet to see a template row for that. I have seen the same five company names on a lot of committees.

The question I ask every rollup team, and which I have now asked of about ten: if your company shut off the servers on a Friday night, how long until a user can exit a position using only L1 contracts, and have you ever tested that on mainnet with real money? Two teams had rehearsed the drill. One had done it with a live position. The rest had a document.

Truth in blockchain isn't a property of the ledger. It's a property of who can afford to verify it, and how fast. A rollup with an untested escape hatch does not have an escape hatch. It has a paragraph.

The genuinely interesting work right now is in the opposite direction: based rollups and shared sequencing, where ordering is pushed to the L1 or to a competitive set of proposers. That movement is where the intellectual honesty lives. It's also where the tradeoffs live, and nobody wants to say them out loud: based sequencing gives up cheap fee capture and low latency, which were the two reasons rollups won in the first place. If your thesis is "rollups are great," based sequencing complicates it. If your thesis is "decentralization is real," based sequencing is the only honest way to say it.

Cell three: "emerging-market payments TAM"

This cell is the most reliably inflated in the entire template, because it asks for a total addressable market and every analyst answers with a population figure.

The template reasoning goes: country has X million people, Y percent unbanked, therefore a blockchain payments opportunity of size Z. It's the same reasoning as a 1999 dot-com deck, and it has the same relationship to reality.

Here is what I've learned from people who actually run remittance corridors and crypto exchanges in Lagos, Buenos Aires, Istanbul, and Nairobi — because the founders building there don't write whitepapers, they answer messages at 2 a.m. People are not adopting dollar stablecoins because they believe in decentralization. They are adopting them because the local currency lost a double-digit percentage of its value last year and will probably do it again. The stablecoin is a savings vehicle that happens to be denominated in dollars, and the blockchain is a delivery mechanism, and if the blockchain disappeared tomorrow they would find another delivery mechanism within a month.

I've watched surveys of stablecoin users in inflation-stressed economies come back with the same result for four years running: the top reasons are savings protection, dollar access, and remittance cost. Ideology shows up near the bottom, if it shows up at all. One founder put it to me more bluntly than I could: "Nobody in my country uses USDT because they're a cypherpunk. They use it because their paycheck buys less every month."

This flips three cells in the table. First, retention: it's high, higher than almost any consumer crypto product, because people don't churn out of a savings hedge. Second, revenue: it's thin. Holders hold; they don't pay fat fees, and any business model that depends on taking 1% of a survival hedge is a business model that will get legislated, competed, or shamed out of existence. Third — and this is the cell that matters most — the chokepoint is not the chain. It's the fiat on-ramp, the bank partner, the local licensing regime, and the KYC stack.

A payments thesis that doesn't name the banking partner in the target country is a payments thesis that has not started. And notice what that does to the "blockchain" column: it shrinks it. The chain is a commodity input. The relationship with a local bank is the moat. We didn't need a new consensus mechanism to learn that. We needed someone to ask who holds the account.

Cell four: "real revenue"

The last cell I want to open is the one where the template invites fabrication most politely: treasury and tokenomics.

The standard column is "TVL" or "annualized revenue," and the standard source is a dashboard. The problem is that TVL is not a measurement. It's an accounting convention. If a protocol's treasury emits tokens to depositors, and depositors deposit those tokens back into the protocol, the number goes up. Nothing was created. A number was arranged.

Here's a specific method, from work I did last quarter while preparing a piece on incentive design. For one mid-cap lending protocol, I wanted a single number: what share of TVL came from wallets that had received tokens from the protocol's own treasury in the prior 180 days? The query took about forty minutes to write. The answer was seventy-one percent.

That is one query. It is not a "dimension" in any template I've seen. And it changes the meaning of every other cell in the report, because it tells you whether you're looking at a business or a round-trip.

The disciplined version of this analysis is not complicated. Separate fee-paying users from emissions-funded depositors. Track the ratio over time. Ask what happens to the number when emissions go to zero. Ask who is on the other side of the trade. None of this requires a nine-part framework. It requires one uncomfortable question and the willingness to publish the answer — including when the answer indicts a project you like.

The contrarian part: the empty table is the most honest document in the industry

Here is where I want to push back on my own argument, because it would be easy — and wrong — to read everything above as a complaint about bad analysts.

The truth is that the forty-one-page report with the blank cells was the most intellectually honest thing I read that month. Its author had done something almost nobody in this industry does: he had built a structure that displayed its own ignorance. He hadn't written "decentralized: yes." He had written "N/A," and he had known exactly what that meant.

The industry punished him for it. Two people told him the report looked unfinished. Someone asked if he'd "considered adding a bull case." Nobody asked what was blocking the blank cells, or offered to help fill them.

So let me be contrarian about the diagnosis. The problem is not that we lack data. We have more data than any financial system in history, and it's public, and it's queryable, and it's free. The problem is that data without a question is noise, and a nine-dimension framework is a machine for generating questions that can be answered without ever asking the one that matters. The template isn't a failure of rigor. It's a defense mechanism — a way of producing something that looks like understanding while never being forced to say I don't know who controls this key.

The blank cell is a confession. The filled cell is a claim. And an industry with a lot of capital and no memory will always reward the claim.

There's a second contrarian point, and it's the one that keeps me up. In a bull market, the catastrophic risk is not the obvious scam. Scams get caught, usually loudly, usually within a cycle. The risk is the legitimate project — real team, real code, real revenue, real audits — where you understood sixty percent of the mechanics and quietly assumed the other forty. The forty percent is where the sequencer is a single node. The forty percent is where the timelock's proposer role is the foundation multisig. The forty percent is where the seventy-one percent came from.

You will never see that on a risk matrix, because a risk matrix only contains risks somebody thought to name. And the naming is the hard part. Naming requires admitting that you didn't know, which is the one thing the format was built to avoid.

What I'd put in the table instead

Five questions. I would genuinely rather read honest one-word answers to these than another nine-dimensional report.

Who can change the code, and how many humans is that? If the sequencer stops, how long is my exit, and has it been tested with real money? Where does the revenue come from when emissions are zero? Who is the banking or licensing counterparty in the jurisdiction you claim to serve? And what would have to be true for me to be wrong?

That last one is the whole discipline. Not "what is the bull case." Not "what is the bear case." What specific, observable thing would falsify my position, and when will I know?

I don't think a framework can save us from this, and I've stopped pretending otherwise. The next cycle's most valuable research product will not be a document with more columns. It will be a short, published list of what the author does not know — dated, specific, and embarrassing enough that someone will try to prove it wrong. That product will lose money in the bull market and be the only thing anybody reads in the bear market. Someone is going to build it, and when they do, they'll be competing against a hundred firms with identical data and no idea which cells are empty.

The table isn't the problem. Believing it is.