Last month a research pipeline I was asked to review returned a document of unusual beauty. Nine analytical dimensions. Thirty-one sub-tables. A risk matrix, a supply-unlock schedule, a Howey-test breakdown, a competitive landscape grid. Consistent terminology from the first line to the last. Not a single typo. And every field, without exception, read the same three words: insufficient information.
It had been fed nothing. No source article. No headline. No dataset. An empty input handed to a machine built to produce structure.
It did not hallucinate.
That restraint is the most interesting thing that happened in crypto this month — and it is the thing almost nobody is willing to pay for.
Because in the same seven days, three governance proposals crossed quorum on three separate chains. On one of them, meaningful votes — actual wallets with skin in the game, not delegated shadows — numbered fewer than a dozen. Each proposal shipped with a rationale paragraph of pure atmosphere: "aligned with the long-term vision." "Community-driven growth." Nobody asked for the numbers. The snapshot passed. The treasury moved.
Two systems. One refused to speak without evidence. One spoke with overwhelming confidence on almost none. Only the second one had a budget.
— Root: The 2022 Bear Market
I want to be precise about what I mean, because "governance is broken" is a complaint, not an analysis.
Let me start with the part of this industry that is genuinely, measurably rigorous: the data availability layer. Since blobs arrived, rollups have been buying a very specific product — a cryptographic promise that a payload was published, that it exists, that anyone can retrieve it. The promise is elegant. It is also completely indifferent to content.
A blob costs what it costs. The fee market prices space, not meaning. A rollup that posts four kilobytes of dense state transitions and a rollup that posts four kilobytes of structured emptiness pay into the same auction and receive the same guarantee. The chain does not care. The chain cannot care. Verification verifies form, not significance.
I have spent a good deal of my career inside audit rooms watching that distinction get blurred. At TrustChain, back in 2017, we reviewed twelve projects before mainnet, and the hardest conversations were never about the code being wrong. The code was fine. The hard conversations were about whether the thing the code computed actually meant anything to the people about to trust it with their savings. A contract that faithfully executes a meaningless instruction is still a contract. And people still lose money.
We helped twelve teams fix their security. We could not help them fix their premise. That is not a thing a smart contract auditor can do.
Now translate that forward eight years. The pipeline that returned "insufficient information" was behaving like a well-audited contract: it executed its specification faithfully, and its specification included a guard clause that said, do not invent. The governance proposals were behaving like an unaudited one: they ran the transfer function perfectly, and the transfer function had no idea what it was transferring.
Code is law, but people are the protocol — and only one of these two systems treated the people as part of the specification.
Here is the technical claim underneath the moral one, and I want to state it plainly, because I think the DA conversation has drifted badly.
Consider what a blob actually is, mechanically. A fixed-size chunk of data attached to a beacon block, priced by a market that targets a set number of blobs per block and adjusts a base fee against that target. The mechanism is clean. It is also content-blind in a way that has real economic consequences: the marginal cost of posting a near-empty blob is identical to the marginal cost of posting a fully saturated one. A team whose actual data footprint is trivial therefore faces exactly one incentive — post anyway, because the narrative of scale is worth more than the fee. Multiply that across an ecosystem and you get a data availability market that is structurally long on ceremony and short on payload.
Most rollups do not generate enough data to justify dedicated data availability. I have run these numbers for years, first out of curiosity and later out of concern as the blob market matured. A meaningful share of posted blob data is padding, repeated commitments, and state that could have been compressed far more aggressively. The guarantee is real. The demand story is thin.
What troubles me is not underuse. Underuse is a young market's privilege. What troubles me is that the underuse is being reported as adoption, and the fee curve is being read as demand. When the most expensive component of your architecture is structurally idle, you have not found product-market fit. You have found a subsidy with a good logo.
A rollup does not become a rollup by purchasing DA. It becomes a rollup by having enough state change to need it. Most of what I see in production right now would run perfectly well on calldata or a shared commitment. The dedicated purchase is a signaling expense, not an engineering one.
That is the DA layer's version of the empty template. Perfect structure. No payload.
And the cost of that ceremony is not carried by the teams. It is carried by the users paying fees calibrated to a demand curve that mostly expresses aspiration. — Root: DeFi Summer
Now the harder half.
Delegation was supposed to solve participation. Instead it industrialized apathy. I have watched token holders — genuinely smart, genuinely busy people — hand their voting power to whichever delegate wrote the most confident thread. Not the most accurate thread. The most confident one. Confidence is cheap to produce and nearly impossible to audit in real time, which makes it a perfect product for a market that has stopped reading.
Look at the distribution. In almost every major DAO I have examined — and I have examined more than I can count since our fifty-page "Democratizing Liquidity" paper was downloaded ten thousand times in a single month — voting power concentrates into a handful of addresses that are themselves downstream of a much smaller set of public voices. Those addresses are not colluding. They are following. That distinction matters legally and not at all practically: the outcome is identical and the accountability is lower, because nobody in the chain from voice to delegate to voter believes they personally made the decision.
The result is a governance layer that looks like a distributed network and behaves like a small committee with excellent public relations. Quorum is met. Turnout looks healthy. The votes are counted correctly, on-chain, verifiably. Every piece of the machinery performs.
The input was empty.
A quorum of three wallets is not a constituency. It is a rounding error with a treasury attached.
I recognize this pattern because I have been inside it. During the 2022 collapse, when the industry's emotional state went from bad to genuinely frightening, I ran a mentorship program connecting two hundred junior developers with veterans. Fifty one-on-one sessions, most of them about mental health rather than markets. We kept roughly eighty-five percent of those people from leaving the industry. Nearly every conversation had the same shape: someone who had trusted a confident voice and never once seen the underlying data. — Root: The 2022 Bear Market
Delegation was never the problem. Delegation to a voice with no visible evidence trail was.
Everything above suggests I want a more conservative industry. I do not. I want a more embarrassed one.
The received wisdom is that crypto's failure mode is fraud — bad actors producing false claims. That is the headline version. The subtler and far more common failure is vacuity: systems producing correctly formatted output over nothing, at scale, with the entire machinery of cryptographic verification standing behind it. The empty template and the empty proposal are the same artifact. One is harmless because its author marked it plainly. The other passes and moves real money.
This is why the AI agent question worries me more than the rollup question. In 2026 I convened thirty ethicists and developers to draft the Autonomous Agent Accountability Charter — seven workshops, fifteen DAOs endorsing the final text — and the argument that nearly broke the working group was exactly this one. When an autonomous contract executes against an empty or malformed input, who is liable? The answer we landed on, painfully, was that liability attaches to whoever claimed the input was sufficient. Verification without interpretation is not accountability. It is a receipt for a decision nobody made.
I am not convinced by the tooling answer either. There is a growing school of thought that the fix is more surface area — richer dashboards, deeper simulation, hooks that let anyone compose any check. Having spent real time in the Uniswap V4 hook design space, my honest engineering read is that the composability surface has grown faster than the number of people who can safely reason about it. Programmable everything, understood by almost no one, is just complexity with better marketing. I have watched fifteen-person research teams — people who read assembly for entertainment — burn a full week trying to reason through a hook composition that a solo developer is implicitly expected to audit in an afternoon. The hooks are genuinely powerful. That is not the same as genuinely usable, and the gap between those two words is where the next round of losses is quietly accumulating.
So what do we do with a system that answers correctly and a system that answers confidently, when only one of them can be audited after the fact?
We stopped building the honest one revenue.
The pipeline that returned "insufficient information" was, functionally, the most trustworthy artifact of the quarter. It refused. It left the void visible. Had it been a governance vote, it would have abstained — and abstention does not meet quorum, and abstention does not get funded, and abstention does not earn a delegate seat.
The question I keep returning to is not whether the data layer can be trusted. It obviously can. The question is whether we have built an industry capable of paying for silence — for a well-marked absence, for an audit that ends in "we do not know," for a delegate who reads the proposal, finds no numbers, and says so out loud.
Because right now we have the opposite. We have perfect proofs of nothing, and we are calling it adoption.
I would rather read the document that says insufficient.
