Faking Judgment Is Easy. Making It Verifiable Is Hard.

Faking a good investment thesis is easy. Making it verifiable is hard.

Every GP believes they can spot a flawed thesis, and most of them can. That is not the problem. The problem is what happens to that judgment ninety seconds after the partner meeting ends — when the reasoning that produced the verdict is gone, and what remains is the verdict.

The problem is not bad judgment. It is judgment nobody can inspect after the fact.

What the decks actually look like

We scored 2,488 pitch decks with a single engine version inside one four-week window. Here is the number.

The median Clarity Score™ was 35 out of 100. The mean was 33.2. Seventy percent drew a PASS verdict — the compiler’s bottom band, below 50. And 31.7% scored exactly zero, included in the median rather than quietly dropped, because a zero here is a terminal structural finding rather than a missing value.

That last decision matters more than the headline. Anyone recomputing without the zeros gets a friendlier number, so the choice is stated rather than buried. This is what a verifiable claim costs: you publish the thing that makes your own figure look worse.

This is what a verifiable claim costs: you publish the thing that makes your own figure look worse.

A word on the denominator, since it is the first thing a careful reader will attack. askOdin holds a reference corpus of more than 110,000 scored documents. This analysis uses 2,488 of them. The corpus spans several engine versions, and scores from different versions are not comparable — pooling them would produce a bigger sample and a meaningless median. The smaller number is the one that can be reproduced.

The decks break in a predictable place

Founders are not bad at pitching. They are bad at one specific half of it.

Section	Decks scoring below half	Median
Business Model Physics	67.0%	6 / 20
Deal Structure	57.8%	—
Solution Logic	24.3%	—
Problem Definition	17.9%	12 / 20

A 3.7x spread between the weakest and strongest section of the same document. The half that describes what is broken in the world is coached, rehearsed, and workshopped. The half that describes how this company makes money is not.

Here is why that asymmetry is dangerous rather than merely interesting: the coached half is the half that persuades. A deck can clear every narrative test a human reader applies and still carry a business model that does not close at any plausible scale. The reader is not being careless. They are reading the part that was written to be read.

Invisible judgment

Consider what survives a diligence process.

A conclusion survives. A memo survives. A number in a model survives. What does not survive is the chain of reasoning that connected the documents to the verdict — which assumption was load-bearing, which contradiction was noticed and set aside, which claim was checked against the model and which was accepted because the founder sounded certain.

That reasoning was real. It was probably good. But it lived in one person’s head for the forty minutes it took to form, and then it was gone. Two years later, when the position is under review and someone asks how the committee got comfortable with the revenue assumption, there is no answer that can be produced. There is only a memo asserting that the committee did.

This is not a rigor problem. Every firm in this market is rigorous. Rigor that leaves no inspectable trace is indistinguishable, after the fact, from rigor that never happened.

Rigor that leaves no inspectable trace is indistinguishable, after the fact, from rigor that never happened.

Seven shapes, and the honest limits of the map

Deals do not fail in infinite ways. The Taxonomy of Venture Conviction names seven structural archetypes across three vectors: Fundamental Solvency, where narrative cannot negotiate with physics; Structural Coherence, where the economics may close but the company is assembled in a way that makes execution unlikely; and Cognitive Alpha, where physics and structure have cleared and the remaining question is whether you see something the market does not.

Five of the seven produce a stop verdict. Two — the False Negative and the Deep Tech Winner — are conviction archetypes, which is the part most summaries of the framework get wrong. The taxonomy is not a catalogue of ways to lose. It is a grammar for deciding what a deck actually is.

It is also a framework for human review, and worth being precise about that. It is how an experienced reader should structure their attention. It is not a claim that a machine enumerates these seven shapes and assigns one — that is a different assertion, and it is not the one being made here.

What verification actually requires

To make judgment inspectable you have to move from a conclusion someone asserts to a record someone else can walk.

RUNE Protocol™ is the compiler. It strips narrative polish out of unstructured materials and renders what is left as business logic. It does not summarise and it does not guess. It compiles.

U.S. PATENT PENDING 63/948,559

RAVEN Protocol™ is the triangulator, cross-referencing claims across documents so that a number appearing three different ways in three different files becomes a finding rather than an inconsistency nobody indexed. The architectural mechanics of RAVEN’s triangulation engine are protected under U.S. Provisional Patent No. 63/994,876 and are not publicly disclosed.

JUDGE Protocol™ is the halting condition. When compiled logic violates a structural constraint, JUDGE stops and flags the violation rather than producing a confident number anyway. It does not return a probability. It returns a verdict. U.S. Prov. Patent No. 64/017,488 | IPOS §34 National Security Clearance (Issued 2026-03-26)

The output is the Defensible Audit Log™ — every claim tied to the page and line it came from, every contradiction preserved rather than silently reconciled. A record that can be handed to an investment committee, to an LP, or to a regulator, and read by someone who was not in the room.

A Dialogue on Institutional Judgment

The Judgment Gap is an existential threat to funds facing the mathematical crisis of scaling capital and deal flow. In the AI era, running on artisanal, unscalable judgment processes is no longer a viable strategy. We are building the infrastructure to solve this.

If you are a partner or principal at a growing venture capital fund and are committed to building a more scalable, defensible, and rigorous investment process, we invite you to a confidential discussion.

Schedule a Confidential Briefing

What good looks like

The firms that compound through the next decade will not be the ones holding the most data. Data is abundant and getting cheaper. They will be the ones whose diligence produces a record that survives the analyst who wrote it.

That means a process that preserves contradictions instead of resolving them into a cleaner story. It means a runtime that halts on a violated constraint rather than optimising around it. And it means accepting that a defensible “we do not know” is worth more at an investment committee than a confident number nobody can trace.

Capital scales smoothly. Human verification does not. When volume rises, every fund improvises its own internal standard for what counts as sufficient reasoning — and none of those standards are auditable across firms, across LPs, or across years. Financial reporting has GAAP. Audit has PCAOB. Cybersecurity has SOC 2. Investment judgment has each firm’s private convention and a memo.

The tools of persuasion have been fully democratised. Anyone can generate a fluent, confident, structurally hollow narrative in an afternoon. The premium has moved to the infrastructure that can tell the difference — and prove it did.

Venture capital is the last unaudited asset class. askOdin provides the infrastructure to close the gap.