Some legal disputes are not caused by bad people making bad choices. They are caused by rules that make bad outcomes structurally inevitable — and that reward the party who knows how to stay silent. This project maps those rules, measures the gaps they create, and builds instruments that make the pattern visible.
In 2008, writing in The Next Web, I argued that the law is "totally inadequate in protecting independent innovators." The piece was about intellectual property — about how ideas, once released into the world, are viral by nature, and how the originator of a good idea is consistently the last person the rules protect.
"Good ideas just by their nature are viral... The only one who is hurt is the genius. No one else cares."
Eighteen years later, the structural observation has not changed. What has changed is the resolution at which it can be examined.
The problem was never particular to IP law. It is a pattern — a repeating signature in the design of rules that were written carefully, often with good intentions, without anyone thinking through who consistently loses when those rules are applied asymmetrically.
The legal system does not need to be corrupt to produce systematically unfair outcomes. It needs only to have gaps — places where the rules run out, where the incentive structure goes uncorrected, where one party's rational choice is to exploit the gap because no mechanism penalises them for doing so. From the outside, the resulting dispute looks like an ordinary conflict between two parties with incompatible positions. From the inside, it has the signature of a designed outcome: one side consistently wins not because they are right but because the rules make their winning structurally inevitable.
This project is an attempt to name those rules. To measure the gaps they create. And to build instruments that make the pattern visible before litigation entrenches it — not after.
What has changed in 18 years is not the observation. It is the availability of AI to do something unprecedented: run the pattern-match quantitatively, across domains and jurisdictions simultaneously, producing a number — a score, a class, a threshold — rather than a qualitative argument that can be dismissed as one-sided advocacy.
Across five papers and four diagnostic tools, this project has identified and formalised four structurally distinct gaps in the design of current legal procedure. Each gap operates independently. Each has the same property: it consistently disadvantages a particular class of party not through any act of malice, but through the mechanics of how the rules are constructed.
Courts, hospitals, regulators, and clients ask one question when appointing a professional to a role: Is this person entitled to do this work? The question is answered by credentials — a degree, a bar admission, a board certification. Once answered, the competence question is, in routine practice, treated as closed.
The second question — How close is this person's demonstrated experience to the specific technical matter in front of them? — is almost never asked. A practitioner can hold every credential the first question requires and still have a career trajectory that has never approached the specific problem now assigned to them.
This gap is not incidental. It is structural, detectable, and measurable — in the practitioner's career record, in their published output, in the semantic distance between their demonstrated subject-matter history and the technical content of the matter assigned. The gap can be scored. The failure can be predicted. The appointment can be challenged — before the matter is lost, not after.
Expert witnesses in civil proceedings are legally required to be independent. The existing recusal mechanism relies on self-declaration and adversarial challenge — a slow, qualitative process with no quantitative standard for assessing how close, in network terms, an expert is to the parties they evaluate.
By applying Erdős-number methodology to academic co-authorship and supervisory graphs, the first paper in this series demonstrates that a direct network proximity — placing an expert at the equivalent of a first-degree professional connection — can go unchallenged under current European procedure. The gap exists not because the rules are poorly enforced but because no quantitative standard has ever been set.
The relational gap and the competence gap are independent instruments. A well-networked expert can still be genuinely competent for a specific matter. A complete stranger can still be incompetent. Both must be measured separately, and both must be measured before appointment — not litigated after failure.
Most legal systems divide disputes into two silos that rarely speak to each other. An ex-ante silo governs what happens before harm — but in most jurisdictions offers no cheap, formal mechanism for the party holding private knowledge of a pre-existing risk to disclose it. An ex-post silo assigns liability after harm — but does so by reference to who was active at the moment of harm, not who held the relevant knowledge.
The result is a Nash equilibrium in which silence is the mathematically dominant strategy for the party who knows. Not because they are malicious. Because the rules price silence at zero and price disclosure at a cost — and no mechanism corrects the asymmetry. The dispute is not manufactured by bad people. It is manufactured by a structural design failure: the missing half of the incentive architecture.
The third paper in this series extends the analysis to a second-order consequence: the same law that assigns liability by activity — without examining whether the affected asset had a pre-existing condition known to the opposing party — functions as a loaded gun. The active party who proceeds without Variable B being disclosed does not merely face a knowledge gap. They face a prospective liability transfer of the pre-existing defect in its entirety. Don't proceed: financial loss. Proceed: the loaded gun fires. The Passive Party's silence loads both barrels.
Standard dispute management treats case value as static. In reality, every delay, bad professional appointment, and adversarial step detonates a compounding multiplier of cost. The gap between initial claim value and actual terminal exposure is driven by interacting, rational-choice actor incentives that systematically manufacture value destruction.
This gap is modeled as a decision tree to predict the Foreseeable Economic Exposure. By identifying how interacting actor incentives escalate costs over time, the diagnostic calculates the value destruction of assignment decisions before they compound.
The last five years of AI in legal technology have, with very few exceptions, addressed a single class of problem: making existing legal processes faster and less expensive. Contract review. Case research. Document drafting. Deposition preparation. Litigation timeline management.
These are optimization problems. They take a process that legal professionals already perform and apply AI to perform it at lower cost per unit. The process itself — and crucially, the rules the process operates within — are treated as given. The question is never: are these rules structurally fair? The question is: how quickly can we process the documents those rules require?
This project operates in a different register entirely.
A faster contract review does not help you if the contract governs a relationship already distorted by a broken procedural rule. The optimisation and the structural diagnosis are not competing — they are sequential. You need the latter before you can fully trust the former.
What AI specifically enables here — that was not practically available before — is cross-domain pattern recognition at scale. The Qualification–Expertise Fallacy does not appear only in courts appointing civil engineers. It appears in hospitals assigning surgeons, in regulators approving auditors, in clients instructing lawyers, in procurement bodies selecting technical advisors. The pattern is the same. The gap is the same. AI makes it possible to run the same structural measurement across all of these contexts simultaneously, producing comparable scores on a common scale, rather than requiring decades of domain-specific litigation experience to recognise the pattern after the fact.
These are not the first legal tools to be built with AI. They may be the first to use AI to diagnose the legal system itself, rather than to serve the legal system as it currently functions.
Every structural gap, once identified and proposed for repair, introduces a new risk: the repair can be weaponised in reverse.
A rule requiring experts to disclose network proximity can become a tool for challenging any appointment through strategic objection. A rule penalising silence about a pre-existing condition can become a harassment mechanism deployed against parties with no genuine concealment. A competence score can be gamed by resume inflation or exploited by adversaries to delay proceedings indefinitely.
"Closing procedural loopholes that benefit bad-faith actors while ensuring the fixes are so tightly engineered that they cannot be exploited in reverse."
This is the symmetry constraint that governs every paper in this project. A proposed reform that creates a new asymmetry in the opposite direction fails the test as completely as the gap it was designed to close. The goal is not to tilt the procedural balance in the other direction. It is to design rules that are genuinely harder to exploit regardless of which party is trying.
This is why each instrument in this project produces a score, not a verdict. The output is diagnostic, not dispositive. The tool identifies the structural pattern and its severity. It does not determine who wins. The score is an evidentiary input — a starting point for legal argument, not a substitute for it. A BFM score of 88/100 does not mean the Passive Party is guilty. It means the incentive structure was comprehensively misaligned, and the Active Party has identifiable grounds to argue that the procedural rules were working against them.
Scores can be disagreed with. Structural arguments can be challenged on the merits. That is appropriate. What the tools prevent is the structural failure going unnamed and unmeasured — and therefore unchallenged.
Each tool produces a quantitative output from raw case materials, practitioner records, or institutional data. All four are free to use, require no account, and return results within seconds.
The Procedural Gap Project is not a law firm. It does not give legal advice. It does not represent clients. It is a research and engineering programme with one specific objective: producing instruments that make structural procedural failures quantitatively measurable.
The audience for this work is, we believe, at least four distinct groups — and they may each find a different entry point more useful:
Legal technologists will find the methodology more relevant than the specific gaps identified. The pattern-matching approach — converting structural legal analysis into scoring instruments with defined scales and thresholds — is a template applicable to any procedural domain.
Practitioners in civil procedure will find the specific instruments immediately actionable: the SDD and NPS scores provide a quantitative basis for expert appointment challenges that currently lack a formal standard. The BFM output provides a structured evidentiary framework for arguing incentive distortion.
Judicial training institutions and court reform bodies will find the academic papers most relevant — particularly Paper I, which has already been received by the Centro de Estudos Judiciários and entered into their permanent library collection.
Researchers in mechanism design and procedural law will find the Nash equilibrium analysis in Papers III and IV a formal treatment of a class of problem that has been observed empirically for decades but rarely modelled mathematically.
All papers are available for download. All tools are free and require no account. Correspondence on the research is welcome.
Legal artificial intelligence is being built at speed and to genuine effect. But it is being built by the profession, purchased by the profession, and optimised for the profession's existing output — the win, the faster draft, the cheaper motion, the better-defended position. That is what its buyers pay it to do. Its buyers are the professionals, not the parties those professionals act upon.
The adoption is real, but the posture is defensive. An industry that has built its revenue model on billable process is aware, at least privately, that a technology capable of reading an entire case record in seconds is a direct threat to the hours spent reading it slowly. The response is not to redesign the model. It is to ride the technology — to absorb it into the existing service structure, reframe it as an enhancement, and use it to do faster what was already being done. The disruption is tamed into a feature. The incentive structure is left exactly where it was.
This is the expected response, and it is entirely rational. You cannot expect a system to disrupt its own revenue model. The legal profession is not going to build the instrument that makes five years of litigation unnecessary when five years of litigation is the product. What gets built instead is a tool that makes five years of litigation cheaper to conduct — faster research, faster drafting, faster filing — which is the profession's genuine need, and which leaves the underlying structure completely intact.
Disruption of the kind that actually changes the structure has never come from within the model it disrupts. Google Search was not replaced by a better search engine. It was replaced by a different kind of reader. The replacement did not come from Google's competitors; it came from an architecture that made the question "which page is most relevant?" redundant by answering the question directly. When the unit of value changes, the entire ecosystem built around the old unit has to rebuild from scratch — and the incumbents who built the most on the old unit have the most to rebuild and the least incentive to start.
The instrument this research programme specifies is precisely the one the legal market has no reason to build: a tool that reads the professional rather than for him, shortens a dispute rather than sustaining it, and is purchased by the party whose interest is resolution — the PI insurer — rather than by the party whose revenue depends on the dispute continuing. It is not a better legal tool. It is a different kind of reader. The five papers named what it would need to read. The tools demonstrate that the reading is achievable. What remains is not a decision. It is a race, and it has already started.
The full argument is in Paper V — On Legal Bullshit, Section 10. The design response is in First Principles.
The most commercially significant reframe of this work is not about legal analysis. It is about what the output actually is. When the Dispute Engine reads a case file and flags a PI liability switch — the moment a professional's own filed evidence contradicts their own asserted position — it is not producing a legal opinion. It is producing a signal that looks much closer to actuarial data than to legal advice.
Modern aircraft do not simply detect engine failure. They detect the patterns that historically precede engine failure — vibration signatures, temperature anomalies, pressure readings — at the point where intervention is still cheap. The entire maintenance and insurance architecture of commercial aviation is built around that distinction: the difference between detecting a crash and detecting the conditions that produce one. The Dispute Engine operates on the same logic applied to litigation. The question it is designed to answer is not "did something go wrong?" but "which of the active files in this portfolio is becoming tomorrow's professional negligence claim?"
That is a fundamentally different product from anything currently on the legal AI market. Every other legal analytics platform answers a question with the lawyer as customer: what does the law say, how strong is this argument, what is the likely outcome? This instrument answers a question with the insurer as customer: which professional behaviours in active litigation are trending toward malpractice exposure, and how far has that exposure already migrated?
The liability migration signal is the most analytically distinct output the tool produces. Most legal AI reads facts, contracts, statutes, and precedent. This reads how responsibility moves between actors across time. At the moment a claimant's lawyers add the filing firm as a second defendant — because the filing firm's own evidence has contradicted the filing firm's own position — liability has not simply increased. It has relocated. The original dispute is still running. A second dispute, between the same firm and its own PI insurer, has just opened inside it. That event is detectable in the record before anyone in the room has named it. The tool names it.
For a professional-indemnity insurer with several hundred firms on cover, the commercial arithmetic is straightforward. If even a fraction of active portfolios contain files trending toward malpractice claims that early intervention — a conversation, a recommended settlement, an elevated premium, a supervision requirement — could redirect, the savings dwarf the cost of the instrument. The tool is not competing with legal AI. It is competing with nothing, because the thing it does has not previously been buildable.
An honest note on validation. The financial projections and risk scores the Dispute Engine produces — exposure ranges, cascade probabilities, PI liability estimates — are model outputs derived from the structural methodology, not empirically validated actuarial figures. The commercial case for the instrument at scale depends on demonstrating, across historical portfolios, that the signals it identifies correlate with files that later generated professional negligence claims. That retrospective validation has not yet been conducted. What has been built is the methodology and the instrument. The validation is the next stage — and it is the stage that converts an analytically interesting prototype into a product a Lloyd's syndicate or a reinsurer would price against.
The portfolio question, in the meantime, is immediate. Every practitioner with a live case file, every insurer with an active book of PI cover, every claimant's new counsel reviewing what prior counsel did — any of them can run the Dispute Engine today. The reading it produces does not require the validation to be useful. What it does require is a user who understands that the output is a structured hypothesis about where the exposure sits, not a certified actuarial finding. Used that way — as an early-warning instrument rather than a verdict — it surfaces things that would otherwise remain invisible until the claim arrives.
The question is not whether the gaps will be found. They will be found, by whoever runs the file first. The question is whether that is you, or whether it is someone on the other side of the matter who already has.
The tools and papers described above are the diagnostic layer. They were always the beginning, not the end. The full scope of this work covers five interlocking layers that operate simultaneously — from identifying what is broken, to proposing the fix, to building the replacement, to simulating it at scale, to grounding it historically. All of it AI-enabled. None of it was practically achievable before.
Four scoring instruments — Expertise Check (SDD), Network Proximity Check (NPS), Bad Faith Check (BFM), and Dispute Engine (PVET) — that make structural legal failures quantitatively measurable. Input raw case materials, practitioner records, or expert profiles. Receive a number. Name what was previously unnameable.
kyc.co/why — context for all four tools ↗Five working papers grounding each gap in formal academic research — mechanism design, network analysis, semantic distance, game theory, and the Self-Reporting Fallacy in litigation. Paper I (Network Proximity and Expert Impartiality) has been formally received by the Centro de Estudos Judiciários (CEJ) and entered into the permanent collection of the Biblioteca Armando Leandro, the institution responsible for training every judge in Portugal. The path from diagnosis to doctrine.
moral.money/paper — all papers ↗The replacement layer. Contracts that carry their own dispute resolution mechanism from the moment of signing. Expert discovery with SDD and NPS checks built in before appointment. An Intellectual Authority (IA) score that makes reputation a portable, formally liquid second currency. And CBLT — the tally stick made digital: money as a two-party obligation backed by verified productive contribution, not issued by whoever controls the tap.
The Blackpaper — A New Economic Operating System ↗An AI mock trial platform where users argue cases as historical figures, AI jurors are persuaded and turned, and every argument becomes a produced documentary film. The only way to test whether a new legal principle actually works is to run it through thousands of edge cases at scale. The Tribunal is that testing ground — building a corpus of which arguments win, under which conditions, against which opponents, that no law faculty could assemble manually.
Enter the Tribunal ↗The historical and first-principles grounding: tally sticks and what they got right for 700 years; the structural omission every monetary system since 1694 has shared; why Bitcoin and CBDC reproduce the same design failure in new materials. The argument is not that the present is uniquely broken — it is that the specific gap has been reproducible and identifiable across every monetary architecture since the Bank of England, and that we now have the tools to both name it and close it.
The Architecture of Resolution — First Principles ↗The diagnostic tools are the visible face of KYC.co. Behind them is a deeper architecture built on one observation: the legal and monetary system's structural failures are not random. They share a common root — the system assigns roles, values, and liabilities without measuring the things that actually determine whether those assignments will produce just outcomes.
If you can measure competence proximity, relational independence, and structural incentive distortion, you can do something more ambitious than diagnose disputes after they start. You can design the contracts, the appointment mechanisms, and the monetary instruments that prevent them from starting in the first place.
Contracts as living entities that carry their own dispute resolution mechanism from the moment of signing — not litigated afterward in a court system already operating beyond capacity.
Expert adduction embedded: the expert is identified, SDD-scored, and proximity-checked before the contract, not appointed by a court after the dispute starts. The competence gap and the relational proximity gap are closed at the contract layer, not after they have already produced harm.
Reputation is not informal. It is a measurable, portable, economically liquid asset that human societies have always used to organise themselves — and have never formally recorded.
The IA score measures seven independently verifiable dimensions: predictive accuracy, early warning speed, question uniqueness, contractor assessment, cost awareness, communication clarity, and payout reliability. Non-zero-sum. Non-transferable. Non-purchasable. It follows the person and can only be built by doing good work. A high-IA parent cannot give their score to a child. An organisation cannot buy a high score for a new employee.
Medieval merchants understood something 700 years ago that modern monetary theory forgot. Money is a two-party record of an obligation, not a token issued by whoever controls the tap. The tally stick: two interlocking pieces of wood, one for each party. Neither could forge the record because the other held the other half. No central authority. No counterparty risk. No ability to print more without the other side agreeing.
CBLT — Contract Backed Lease Token — is the digital tally stick. Issued only when an IA score justifies the credit. Backed by verified productive contribution, not political preference. Not a cryptocurrency. Not a central bank. Legally compliant. Dispute-resolved by design. Credit that follows competence, not collateral.
"Every economic system in history has required a trusted authority to issue currency. The mechanism differs; the structural outcome is the same: those closest to the source of new money accumulate advantages that compound over time. The core design challenge is whether it is possible to create a system where money is issued in direct proportion to verified productive contribution — automatically, transparently, and without any single entity controlling the tap."
The full design argument — covering the problems with traditional finance, blockchain, gold, and every alternative that has failed to include dispute resolution — is set out in the Blackpaper. It is available in English, Portuguese, German, French, and Spanish.
The academic papers diagnose structural gaps. The KYC.co infrastructure proposes the replacement. But legal and monetary innovations of this scale require something that neither diagnosis nor infrastructure can provide on their own: a testing environment at imaginative scale — a place where new legal principles can be run against thousands of edge cases, stress-tested against historical opponents, and stress-tested again until they break or hold.
That is what The Great Tribunal is for. It is, simultaneously, a multiplayer AI argument platform, a courtroom documentary film production system, and a research instrument for building the largest corpus of tested legal reasoning that AI has yet made possible.
"Every argument has its day in court. Every verdict becomes a film."
The Intellectual Authority (IA) score earned in the Tribunal mirrors the IA score in the real KYC.co system — establishing the conceptual bridge between the simulation and the infrastructure. What you demonstrate as an arguer in a historical trial is a measurable proxy for what you would demonstrate as a legal or economic actor in a real commercial relationship.
The complete set of academic tools and papers is not finished. The infrastructure is being built in the open. The simulation laboratory is live but expanding. This page is a progress report on a programme that is moving fast — entirely aided by AI, buckled against real-world problem experience with established legal, economic, and governance systems.
The development methodology is itself part of the experiment. Every component of this work — the diagnostic tools, the academic papers, the contract architecture, the monetary mechanism, the film production pipeline — was built with AI as the primary development partner, in response to real cases where the existing system produced structurally predictable failures. The AI does not lead. The real-world failure does. The AI scales the response.
How far can AI-aided legal innovation be pushed when the starting point is not a feature request but a structural problem experienced in real proceedings? The answer, so far: further than any conventional legal technology programme has gone, faster than any academic publication cycle allows, and into territory — redesigning the monetary system, building AI courtroom film production, formalising reputation as a currency — that most legal tech companies would not attempt.
Correspondence welcome. If you are working on structural legal reform, judicial training, mechanism design, or AI-enabled legal infrastructure and want to discuss any aspect of this work: hello@kyc.co