Two in the morning in Nairobi. The power flickers, the laptop holds, and I hit run on a scan I've run four hundred times โ nine dimensions, one mid-cap DeFi protocol that has been bleeding deposits since March. Tech. Tokenomics. Market. Ecosystem. Regulatory. Team. Risk. Narrative. Supply-chain transmission.
Nine rows of N/A come back.
Not one dimension resolved. No title. No core thesis. No information points. The pipeline didn't break โ the pipeline was honest. There was nothing to read. No audit since 2023. No governance post in 147 days. No roadmap update since October. No commit that touched a contract.
Over the same seven days, that protocol lost 40% of its liquidity providers.
We don't usually get to see the negative space. We see the silhouette and mistake it for the object.
Nothing about my grid is exotic. I built it the way most analysts built theirs โ out of the abundance of 2021, when the problem was triage rather than scarcity. Every protocol had a litepaper, a Mirror essay, a three-hour Twitter Space, an ecosystem grant, a founder with a podcast habit. Nine dimensions existed because there was too much to read and too little time to read it.
Then abundance inverted. The same grid that sorted surplus now handles scarcity, and it handles it badly, because a framework designed to separate signal from noise quietly assumes there is signal to separate.
I've watched that assumption fail in three rooms. In 2020 I forked a stableswap invariant locally and spent two hundred hours simulating impermanent loss across asset pairs, mostly to convince myself that the math was more honest than the interface built on top of it. In 2022 I ran a proof-generation newsletter and a Nairobi builders' Discord, tracking which teams kept publishing research while their charts went vertical in the wrong direction. In 2024 I sat with fifty executives who asked the same question in different accents: how do we know what's real? The checklist I handed them was this one. Nine rows.
So when the rows come back empty, I don't treat it as a dead end. An empty field is a finding โ it has direction, magnitude, and a cause. The work is figuring out which of the three you're looking at.
Absence has structure. That's the first thing to internalize. In a functioning market, silence is distributed in a pattern: infrastructure teams keep committing code because commits don't care about price, while narrative teams go dark because their entire informational footprint was rented. If you map which fields return null across a portfolio of thirty protocols, you don't get random holes. You get clusters.
The clusters trace a budget line.
This is where the mechanics get uncomfortable. Liquidity mining APY is not revenue. It is rent paid to depositors, and it is typically paid from the same treasury account that funds the content, the grants, the conference sponsorships, and the monthly ecosystem recap. When that account thins, the withdrawal is simultaneous and coordinated: TVL drops in the same week the blog goes quiet. Analysts read those two events as separate stories. They are one story. I spent a full quarter in 2021 modeling where incentives end and organic demand begins, and the honest answer for most farms was that organic demand was the smaller term by an order of magnitude.
A protocol's silence is a function of its revenue model before it is a function of its engineering. The engineering often survives. What dies first is the layer that produced the artifacts my grid was designed to collect.
This is the negative-space problem, and it has a shape. A due-diligence framework asks for narrative residues โ litepapers, audits, team bios, roadmaps, integration announcements. It does not ask for a state root. It does not ask what the LP distribution looked like across the last ninety days, or whether the stablecoin reserves in the treasury contract actually match the number in the slide. So a protocol can be alive and still return nothing. A protocol can be dead and return everything, if someone is still paying a communications contractor.
Follow that logic into the layer wars and it explains something most coverage gets backwards. The visible difference between competing rollup stacks in this market is not the proof system. It's deployment count โ who convinced more teams to launch a chain on their rails. That race produces artifacts by design: foundation grants, launch threads, partnership announcements, a steady drip of names. So those dimensions stay populated even in a downturn, for whoever is spending. A research-heavy team with no business development budget, working on something genuinely difficult, returns N/A across the same grid.
The instrument rewards whoever buys attention. That's a bias, and it's mine, and it's yours too, because the market runs the same distorted scoring function.
Then there's the opposite error, which is worse. Naming without definition. A category can fill every field while being empty of substance โ a familiar label attached to something that shares only a marketing department with its supposed lineage. I've audited enough of these to be tired of the pattern: a project with deep roots in one ecosystem, a fresh domain name, and a rebrand that borrows the cultural capital of another community that never asked for the association. The fields populate. The thesis dissolves on contact with the bytecode. Builders who actually live inside the original community tend not to recognize the newcomers at all.
Which failure should you optimize against? In an expansion, false negatives cost opportunity and nobody notices. In a contraction, false positives cost capital and everybody notices. Bear markets are the wrong time to be generous with the benefit of the doubt. When information is expensive to produce, the burden of proof belongs on the party making the claim, not on the analyst who couldn't find it.
About Me: I write this from the product side, not the podium. My day job is shipping protocol infrastructure; this column is where I test whether the things I build actually survive contact with people who don't care about any of it. That bias is worth naming before you weigh anything above.
So what do you read when the rows come back empty? You read residue. Treasury runway, computed from the actual contract balance rather than the announcement. LP composition over time, watching whether the same three wallets are rotating in and out to farm a points program. Commit recency set against commit content โ a typo fix in a README is not development. Who authored the last governance proposal, and whether they were paid by the foundation. None of this is glamorous, and none of it fits neatly into nine rows, which is precisely why it gets skipped.
I learned the shape of this in 2017, tracing the reentrancy logic behind one of the industry's defining failures across roughly 150 hours of manual review. The exploit itself was almost trivial to spot once you looked. The failure that mattered was upstream of the code: nobody was reading, because everyone was reading the narrative instead. Nine years later, the instrument has changed and the failure mode hasn't.
Here's the contrarian case, and I hold it seriously. The blank report might be the most honest document in the folder. A fully populated deck is a sales artifact โ every field shaped by someone whose compensation depends on the field looking good. Nine N/As are a confession, and confessions are rare enough to be valuable. The analyst's reflex is to re-run the analysis, to fill gaps with inference, to narrate the void into something legible. That reflex โ the compulsion to explain โ is the same engine that inflated the last cycle. The most dangerous person in a research call is the one who cannot tolerate a blank space.
And while I'm dismantling my own tool: nine dimensions is eight more than ever mattered. Two survived contact with reality โ treasury runway, and whether the team has ever shipped anything without a token attached to it. Everything else was rigor as costume.
The bear market didn't delete the code. It deleted the noise around the code, which means the thing we're all staring at now is closer to the real object than anything we saw in 2021. Build a protocol whose record can still be found in a quarter where nobody pays for attention. That's the standard I'd hold my own work to.
When the narrative budget hits zero, what of yours is still standing?