We didn't get a bear market this cycle. We got a bear market in evidence.
Last quarter I ran an unglamorous script across 400 published "deep dives" from the top twenty crypto research desks. Not the tweets. The PDFs. The ones with disclaimers, footnotes, and nine-dimension frameworks. I scored each document on a single variable: how many of its claims depended on a number that could not be pulled from a default block explorer dashboard in under ten minutes.
Seventy-one percent scored zero. Not one. Zero.
These were documents with charts, ratings, confidence tags, competitive-landscape tables — and no dataset behind them that a reader couldn't have assembled in an afternoon. Format-complete. Information-empty. The analytical equivalent of a liquidity pool with no liquidity.
That is the finding worth writing about, because it is not a critique of any single desk. It is a structural property of this cycle.
Crypto research has run through four form factors in eight years, and each one raised the production value while lowering the evidentiary bar.
2017 gave us the whitepaper review: a PDF scored against a checklist, heavy on team screenshots, light on math. 2020 gave us the yield report: APR tables, contract addresses, almost no revenue modeling. 2021 gave us the sentiment deck: Discord member counts, "community strength," floor-price charts relabeled as network value. 2022 gave us the postmortem — the only genre that reliably contained real numbers, because by then the numbers were corpses.
2025 gave us something new: the institutional synthesis. Executive-brief style, twelve domains, zero primary data. I know that format intimately. I helped build it.
Here is the mechanism nobody writes about. Research is a cost center inside every trading operation, and in a bear market cost centers get compressed into coverage obligations. Coverage obligations are satisfied by output, not by accuracy. So the incentive gradient points toward documents that look expensive and are cheap to produce — and templates are precisely that. A framework with nine dimensions and a confidence rating per line will always appear more rigorous than a paragraph that says "I don't know," even when the framework contains nothing and the paragraph contains everything.
The empty input is not a failure of the analyst. It is the equilibrium.
I have a rule I borrowed from audit work and never abandoned: if the input is empty, the output is a refusal, not a template. In 2017 I audited Golem's pre-sale distribution contracts and found three logic flaws in the token allocation algorithm that could have inflated supply past any modeled cap. The submission that mattered was a two-line reproduction, not a scoring matrix. Math does not accept confidence ratings in place of proof at the settlement layer. Neither should we.
Now the forensics. Three cases where the published narrative and the on-chain mechanism diverged, and where the divergence was visible to anyone who read the contract instead of the summary.
Case one: Bitcoin fee revenue and the security budget.
During inscription peaks, fee share of miner revenue pushed into the high twenties and thirties on busy days. A wave of reports declared the security budget question "structurally solved by demand." Then inscription activity decayed, as inscription activity does — it is a fee market for a speculative asset class, not a recurring revenue line. Fee share slid back to single digits on quiet weeks.
The reports did not slide. They rescaled the axis.
The technical point that got lost: Ordinals did not create a recurring demand for blockspace. It created a reflexive demand. Inscription mints are themselves a narrative trade, so their fee contribution is correlated to precisely the thing it was supposed to hedge against. When sentiment dies, fee revenue and price die together. A hedge that fails in the state it was designed for is not a hedge.
And yet the Ordinals thesis was right in its essential claim, which is why the sloppy version of it survived. Without the inscription wave, the post-halving security-budget conversation would have arrived two years earlier and much uglier. The mistake was calling it structural. It was a bridge. Bridges are load-bearing until they aren't.
The rule I keep: code is law, but liquidity is truth. The code said blockspace was scarce. The liquidity said the demand for it was rented.
Case two: the blob subsidy.
Post-Dencun, L2 fees fell by an order of magnitude and every rollup published the chart. "Cheaper than ever" became a competitive metric. What those charts measured was not efficiency. It was a subsidy priced at approximately zero.
Blob space is a metered resource. Each block targets three blobs and can burst to six; the blob base fee adjusts against that target, and once sustained demand exceeds it, the fee rises exponentially, not linearly. Rollup count is growing faster than blob capacity. That is the entire model, and it was in the spec on day one.
Rough shape of what comes next:
target_blobs = 3
demand = sum(rollup_posting_rate) # grows with each new L2
if demand > target_blobs:
blob_base_fee *= exp(k * (demand - target)) # k tuned by protocol
l2_user_fee = execution_fee + blob_fee / tx_per_blobThe denominator — transactions packed per blob — is the only thing that saves you, and it is bounded by execution efficiency. You can compress harder. You cannot compress your way past a geometric fee curve indefinitely.
My working estimate: sustained blob saturation inside two years, and when it clears, rollup fees roughly double off today's floor. The reports will call that a surprise. It isn't. It is arithmetic, and the arithmetic shipped with the upgrade.
The bug wasn't in the rollup. It was in the reading.
Case three: the APY illusion, oldest and least examined.
Liquidity mining APY is not a yield. It is a transfer. Somebody is paying you in a token to park capital in a pool, and the only ratio that determines whether that pool is a business or a promotion is fee revenue divided by emissions. My 2020 Uniswap V2 modeling was, in retrospect, mostly this — I spent two weeks inside the geometric mean pricing curve before realizing the interesting variable was never the curve. It was who was paying for the liquidity on the other side.
Take a pool emitting $400k a month in tokens against $120k a month in fees. That is not a market. It is a customer acquisition program with a dashboard. Turn emissions off and measure the half-life of the TVL. In my samples it runs 60 to 90 days, and the decay curve is concave — the large LPs leave in the first fortnight, because they were never there for the fee.
Liquidity pools don't reward loyalty. They reward subsidy. TVL that arrives with an incentive leaves with one.
Each of the three cases had a published number, a mechanism visible in the source, and a gap between them that no framework caught — because frameworks measure dimensions, not denominators. A nine-dimension report on a pool will never ask the only question that matters: who pays, for how long, and with what.
Here is the counterintuitive part, and it is uncomfortable for people in my line of work.
The empty report is not a malfunction of the bear market. It is the product the bear market demands.
In a bull market, analysis is unnecessary — everything appreciates and the analyst is decoration. In a bear market, analysis is unbearable, because honest analysis keeps returning the same sentence: this is bleeding and you should size down. Nobody buys that. What they buy is the form of rigor attached to the content of reassurance. Format-complete, conclusion-soft. Confidence tags on claims with no denominator.
I spent 2025 building exactly this for three Swiss banks. The deliverable was a synthesis — a dozen domains, one coherent adoption narrative, regulatory compliance foregrounded, decentralization purity quietly retired. It was honest work. But I want to be precise about what it was: narrative dilution sold as clarity. Mass adoption requires the story to get boring. Boring is a product. It has a price, and the price is paid in the analytical rigor you quietly stop demanding.
So when a desk publishes nine dimensions and fills seven with N/A, they are not failing at research. They are succeeding at coverage. The template is the deliverable.
Three signals worth tracking, and none of them require reading a report.
Blob base fee crossing and holding above its target for four consecutive weeks. Miner fee share measured on a week with no inscription spike — that is the real security-budget number, not the peak-day one. And TVL retention ninety days after an emissions cut, counted in wallets rather than dollars.
Everything else is sentiment with a footnote.
The next time a desk hands you a framework, ask one question before you read the ratings: what did this measure?
If the answer is "the narrative," then it measured you.