Due Diligence Field Guide

This guide condenses two companion essays — The Quantum Comprehension Premium and The Human-Centered Alibi — into a form you can carry into a meeting or a term-sheet review. It is not a substitute for legal, technical, or financial diligence. It is a first-pass screen for a specific failure mode: the correct sentence used to protect an incorrect programme. Each section pairs a cheap-to-fake answer with an expensive-to-fake answer. The distance between them is the distance between language and discipline.

The one-sentence screen

Ask it before reading the deck

Most diligence documents are written to survive a meeting. The one-sentence screen is designed to survive a cab ride. Before the founder opens the laptop, ask: what is the strongest classical or conventional baseline your result beats, who tuned it, and what would have to be true for the advantage to disappear?

A team that has already done the work answers with numbers and a list of its own failure modes. A team that has not reframes the question as scepticism. The reframing is the signal. It means no instrument is in place that would detect the error before the investor does.

The screen is domain-agnostic. In quantum work it asks for the tuned classical baseline. In biology it asks which conformational ensemble was modelled and what happens when it shifts. In AI products it asks what retention looks like among users who could substitute tomorrow. Same shape, three substrates.

Cheap-to-fake answer	Expensive-to-fake answer
"We are faster / better / more scalable."	Named baseline, named tuner, and the conditions under which the advantage vanishes
"The classical approach doesn't apply here."	A documented attempt to make it apply, and the metric gap it left
"Our experts reviewed this internally."	An outside party with no equity stake who can describe the failure mode

Licensed inference — That the ability to state a falsifier for one's own headline claim correlates with the reliability of later reported results.
Asserted — That this question functions as a practical first-pass diligence screen. Testable against realised outcomes; not tested here.

Three comprehension deficits, three write-down accounts

Where the unpriced risk lives

Every technical claim sits on one or more of three substrates. A deficit on any of them moves cost into an account the narrator does not control.

Physical comprehension is knowing which gains come from the substrate and which from classical pre- and post-processing. Without it, hardware spend is stranded and benchmarks are retracted after the press cycle.

Biological comprehension is knowing that living targets are distributions, not structures. Without it, programmes run against a single conformation and pay in late-stage attrition.

Human comprehension is knowing that a user is a learning system with a plasticity budget. Without it, dependence is read as product-market fit, and the bill arrives as churn, regulatory attention, or brand devaluation.

Cheap-to-fake answer	Expensive-to-fake answer
"We have world-class hardware / biology / UX."	A published map of which gains come from the substrate, the model, and the interface
"The error rate is improving exponentially."	Physical-to-logical overhead, ensemble coverage, or substitutability cohorts plotted against time
"Our users love it."	Retention among users with a real alternative, and a published refusal surface

Established — Late-stage clinical attrition is the dominant cost driver in pharmaceutical development, and target-model error is a recognised contributor.
Established — Proteins occupy conformational ensembles; function depends on dynamics as well as shape.
Licensed inference — That the three shortages produce distinct and separable cost lines rather than one undifferentiated hype cost.

The six questions that cost something to answer

From The Human-Centered Alibi, formatted for use

These are the operational tests of whether a humane AI claim is earned or borrowed. Each has a documentary answer, a date, and a party who can be held to it. A well-resourced narrator can produce the cheap answers instantly. The expensive answers leave a paper trail.

Cheap-to-fake answer	Expensive-to-fake answer
"Humans are always in the loop."	A dated log of the last human override, the named decision-maker, and the reason recorded before the outcome
"We have an escalation process."	A named role, its protected hours for review, and the span it now covers after any flattening
"We invest heavily in upskilling."	Training spend per redesigned role, year over year, beside the headcount change in the same function
"We are aligned on outcomes."	A contract that separates the party defining the outcome from the party measuring it
"The model has guardrails."	A public list of question classes the system declines, with the failure modes that produced it
"Our roadmap is on track."	A changelog of retired claims, each with the evidence that withdrew it

Asserted — That the six-question audit is sharp enough to separate earned from asserted humane strategy in practice. Untested; it needs buyers to run it.

The double bottom line test

Give the second line a unit, a cadence, and an outside measurer

The humane strategy deck usually promises a double bottom line: profit and human flourishing. The test is not whether the promise is sincere. The test is whether the second line has the same reporting discipline as the first.

Profit is audited quarterly, in a standard unit, by a party that can be sued. The social term is usually measured in annual-report narrative, in no fixed unit, by the company itself. When one term is continuously measured and the other is rhetorically measured, the optimiser follows the gradient it can see.

The fix is not more values language. It is to give the second line auditable proxies: training spend per redesigned role, override counts, incidents caught pre-release, and a published refusal surface. Then report them on the same cadence and have them measured by a party that does not also set the target.

Cheap-to-fake answer	Expensive-to-fake answer
"We care about people and planet."	A table with unit, cadence, and outside measurer for each social claim
"We are tracking qualitative impact."	A numeric proxy whose deterioration would trigger a board-level review
"Our culture speaks for itself."	An independent survey with a methodology, a baseline, and a date

Licensed inference — In a two-term objective where one term is externally audited and the other is self-reported, effort concentrates on the audited term.
Asserted — That giving the social term a unit, cadence, and independent measurer is sufficient to make it compete. Untested at firm scale.

Red flags that travel in pairs

Combinations that predict the alibi is functioning

Single red flags can be noise. Pairs are more reliable. When two of the following appear together, the humane claim is probably operating as an alibi rather than as a discipline.

Cheap-to-fake answer	Expensive-to-fake answer
"Human in the loop" + no override log	Named accountability with dated refusals
"Flattened structure" + no rehoused escalation	A named escalation role with protected hours
"Outcome-based contract" + same party defines and measures	Independent measurement clause in the contract
"Reskilling commitment" + no training line item	Spend per role, tracked year over year
"Double bottom line" + no unit for the second line	Auditable social proxy reported on the financial cadence
"Quantum advantage" + no classical baseline	Named baseline, named tuner, falsifier stated

Asserted — That these pairings predict a gap between stated strategy and operational practice. Testable by comparing stated strategy to documentary answers.

Worked example — Topological Golf

Applying the screen to an embodied, non-corporate claim

The framework is not limited to term sheets. A recent side project, Topological Golf (topologicalgolf.com), applies the same structure to a domain most diligence rooms never see: the golf swing. The site claims that a player can operate in a 'Verb Frame' — holding the final state in mind and letting the nervous system collapse the swing into one result — rather than executing a stack of mechanical rules.

The one-sentence screen still works. The classical baseline is conventional golf instruction, motor-learning research, and sports-psychology practice. The expensive answer would name which part of the improvement comes from attentional state, which from reduced anxiety, and which from ordinary practice effects. It would also state what would have to be true for the advantage to vanish — for example, if controlled trials showed no durable handicap improvement from visualization-only protocols.

The three comprehension deficits apply here too. Physical comprehension separates the quantum metaphor (superposition, wave collapse) from the actual substrate (attention, proprioception, flow states). Biological comprehension asks whether the 'club as nervous-system extension' claim maps onto documented sensorimotor coupling or is only a vivid analogy. Human comprehension asks whether the method produces durable skill transfer or just a transient peak experience.

The value of the example is that it shows the screen is shape-agnostic. The same questions that expose an AI alibi can expose a wellness alibi, a sports alibi, or a quantum alibi. The language changes; the structure does not.

Cheap-to-fake answer	Expensive-to-fake answer
"We use quantum superposition."	A named attentional or motor-learning mechanism, and an outcome measure that distinguishes it from placebo and practice effects
"The club becomes part of your nervous system."	A reference to sensorimotor coupling or tool-embodiment literature, with the conditions under which the effect fails
"You enter flow and the shot happens."	Retention data: does the method still work under pressure, fatigue, and after a layoff?

Analogical — The golf example is used to show the shape of the diligence screen, not to validate the claim itself.
Asserted — That the same cheap/expensive-answer structure appears in embodied, non-corporate claims. Untested across a representative sample.

How to use this in a live conversation

Start with the one-sentence screen before the presentation begins. It is disarming because it sounds like ordinary curiosity. A prepared team will recognise it as the question they have already asked themselves. An unprepared team will reach for the language in their deck.

When you hear a cheap answer, do not treat it as a lie. Treat it as a missing instrument. Ask what would have to be in place for the expensive answer to exist six months from now. Sometimes the missing instrument is money, sometimes it is authority, sometimes it is simply that nobody has been asked for it before.

Record the red-flag pairings as you hear them. Single claims are not verdicts. Combinations are. The point is not to catch someone out; it is to notice when the structure of the claim is protecting itself from being tested.

Afterword — The unhireable filter

There is one diligence shortcut that does not require a checklist: make the wrong conversation economically impossible. If the price is a quadrillion per nanosecond, the headhunter who wants a comfortable story will hang up before the deck opens. The best way to solve a problem is not to have one.

This is not a recommendation for everyone. Most buyers, boards, and builders cannot opt out of the market. They need the screen. But for the person who can set the price, the price itself becomes the screen. It saves the time that would otherwise be spent detecting whether the other party wants the truth or wants the truth to sound like what they already believe.

The checklist is for when you must engage. The unhireable filter is for when you can choose not to. Used together, they cover both directions: the conversations you enter are disciplined, and the conversations you refuse never happen.

A worked example of the filter in personal form is available as a Resume as Topological Golf Addendum.

What would show this wrong

If firms that answer the one-sentence screen with numbers show no better claim-durability or milestone attainment than firms that do not, the screen is not a screen and should be dropped.
If firms that publish tuned classical baselines, ensemble models, and retention-by-substitutability cohorts show no differential in cost of capital, attrition, or churn, the comprehension-premium framing is wrong.
If buyers running the six-question audit cannot distinguish earned from borrowed humane strategy within two to three years, the audit is decorative and should be replaced.
If organisations that give their social bottom line a unit, cadence, and outside measurer show no governance advantage over those that do not, the double-bottom-line objection fails.
If the red-flag pairings above turn out to be no more predictive than chance of later remediation or retraction events, this field guide is not a field guide.

Sources

Due diligence — The existing process into which these questions insert themselves at near-zero cost.
Quantum error correction — The overhead that separates physical qubit counts from logical ones.
Conformational ensemble — Why a biological target behaves as a distribution rather than a structure.
Drug development — The attrition economics that make late-stage failure the most expensive error in a portfolio.
Cost of capital — The channel through which credibility discounts become an operating disadvantage.
Externality — The mechanism by which fabulist cost is displaced onto investors, users, and regulators.

Back to all essays