Product
Platform Overview Secure Search ZugZwang How It Works Research Pricing
About
Our Story Our Creed Contact Book a Briefing

Litigation Intelligence Engine

Zugzwang

zugzwang — chess  ·  pronounced ZOOGZ-wang
The position where any move worsens your situation.
Forced to act. Every option leads somewhere worse.
Already lost. Does not yet know it.

Three-mode litigation intelligence engine. DEFEND stress-tests your argument under structured adversarial review. EXCAVATE surfaces theories you may not yet have explored. ATTACK ranks weaknesses in the opposing filing before your response. Sessions run on firm-controlled hardware in your office. No external model API calls during session execution.

Request a Confidential Walkthrough Firm-controlled  ·  Air-gapped  ·  Your hardware
ZugZwang — Illustrative Session Running adversarial rounds
DEFEND EXCAVATE ATTACK ✓
Round 01
38
Round 02
57
Round 03
74

Attack — Exploitable Weakness Found

"The motion conflates jurisdictional standing with merits injury — a Rule 12(b)(1) vs 12(b)(6) classification error that gives plaintiff three procedural escape routes before the court reaches substance."

Attack Surface
74/100
Internal session rubric.
Not a case outcome prediction.
Mode: ATTACK
7 weaknesses surfaced.
0 unresolved.
D
Defend

Your argument enters. The engine attacks it, rewrites it under pressure, and converges toward its strongest defensible form. You receive what survived.

E
Excavate

Your seed position enters. The engine follows the implications into territory you have not mapped. You receive the theory that emerged under adversarial exploration.

A
Attack

Their argument enters. The engine inverts. The Adversary attacks their filing from your position. You receive a ranked map of every exploitable weakness before you respond.

The Problem

Too often, the side
with more resources
finds the weakness
first.

That is not a legal problem. It is a logistics problem. The better-funded side has more attorneys, more research hours, more institutional knowledge, and more time to find the weaknesses in your argument before you find them yourself.

ZugZwang closes that gap. Not by adding staff. By running the adversarial pressure the other side would apply -- before they apply it. On your hardware. In your office. Every session logged. During session execution, prompts, outputs, and session records remain on firm-controlled hardware unless your team chooses to export them.

The side that runs ZugZwang walks into court with a clearer map of what the other side is likely to attack. Because they pressured the argument first.

Without ZugZwang

BigLaw files a motion to dismiss. You read it. You respond from memory and experience. The weaknesses you missed become the arguments they reinforce at oral argument.

Attack Mode

Their motion goes into ZugZwang. The engine attacks it from your position. You receive a ranked map of every exploitable weakness before you write a single word of response.

Without ZugZwang

You take a case on contingency. Theory feels solid. Eighteen months in, opposing counsel finds the structural weakness you did not see. The settlement is lower than it should have been.

Defend Mode

Before you commit, the theory goes through ZugZwang. The argument rewrites under pressure. You take the case with a clearer map of likely attack paths -- or you pass.

Without ZugZwang

You have facts and a client but the theory is incomplete. You commit to the obvious angle. Discovery reveals a stronger one. You pivot late and lose momentum.

Excavate Mode

Before you commit, the seed position goes into ZugZwang. The engine follows implications into territory you have not mapped. The stronger angle surfaces before you spend a dollar.

Three Modes

One engine.
Both sides of every
adversarial interaction.

DEFEND and EXCAVATE operate on your arguments. ATTACK operates on theirs. Together they cover the complete litigation lifecycle from intake decision to response brief. Select the mode that matches where you are in the matter.

D
DEFEND

Forge your argument

When you have a theory and need to know if it holds

Your argument enters. The Adversary is configured to generate adversarial attacks based on the matter type, posture, and jurisdiction your team provides. The Author rewrites in response. The Judge scores and surfaces stronger angles. The engine pivots when a better position emerges. The loop runs until the argument reaches its strongest defensible form or stalls under pressure.

Output: The argument at the strongest form it reached under adversarial pressure, plus a full log of attacks identified, weaknesses addressed, and pivots taken. Session scores reflect internal rubric quality -- not predictions of judicial outcome.

E
EXCAVATE

Find your argument

When you have a situation but not yet a theory

Your seed position enters. The engine follows implications from that position into territory you have not mapped. It surfaces angles, downstream consequences, and argument structures you have not considered. EXCAVATE does not harden a fixed position -- it follows the strongest implication wherever it leads. Use it at intake when the theory is incomplete, or when discovery has revealed facts that do not fit your working theory.

Output: An implication map showing where the seed position leads, which angles emerged under adversarial exploration, and what the strongest emergent theory became.

A
ATTACK

Break their argument

When you need to find the holes in the opposing filing

Their argument enters. The engine inverts. The Author defends the opposing filing -- playing their counsel's role. The Adversary attacks it from your position, finding the logical gaps, unsupported claims, jurisdictional errors, evidentiary weaknesses, and structural failures. The scoring rubric inverts: a high score means their argument is weak and your lines of attack are strong.

Output: A ranked attack map identifying the strongest lines of opposition against the filing, with the full adversarial reasoning chain documented. Your attorneys evaluate the findings and decide what to pursue.

The Engine

The methodology that
stress-tests positions
before real stakes are committed.

In quantitative finance, no position goes live without stress-testing it against adversarial scenarios first. You find the failure modes before real money is at risk. The methodology is the same: apply structured adversarial pressure to the position until you know what survives and what does not.

ZugZwang applies that methodology to legal arguments. The position is your case theory or the opposing filing. The adversarial pressure is the engine. The output is not a report about the argument. It is the argument itself, rewritten and hardened through structured opposition -- or the opposing argument, mapped at its most exposed.

"Dean Hoffman built the award-winning Checkmate trading system on one principle: find where the position breaks before real stakes are committed. ZugZwang applies the same principle to legal arguments."

Every session produces a full audit trail: every attack identified, every rewrite, every pivot decision, every score. Hash-based integrity records support documented attorney review. Session scores reflect internal analytical rubrics. They are not predictions of judicial outcome.

01

Select the mode

DEFEND to forge your argument. EXCAVATE to find your theory. ATTACK to map weaknesses in the opposing filing. Each mode configures the engine's role assignment and scoring rubric automatically.

02

Input the argument or filing

Paste your case theory, motion draft, hypothetical, or the opposing filing. Define the thesis, case posture, and jurisdiction. ZugZwang builds the initial position from your material.

03

Adversary applies pressure

In DEFEND and EXCAVATE, the Adversary attacks your argument. In ATTACK, the Adversary attacks the opposing filing from your position. Each round surfaces new weaknesses and attack angles.

04

Author rewrites, pivots, and converges

The Author rewrites in response to each attack. When the Judge surfaces a stronger angle, the engine pivots. The argument evolves through the pressure, not around it.

05

You receive the result

DEFEND delivers the forged argument. EXCAVATE delivers the implication map. ATTACK delivers the ranked weakness map. All three deliver a full audit trail. During session execution, no data is transmitted outside firm-controlled hardware.

This Is Not Legal Research

Legal research tells you
what the law says.
ZugZwang tells you
whether the argument survives.

Other Legal AI Tools

Retrieval and summarization

You ask a legal question. The system searches case law databases, pulls relevant holdings, summarizes applicable statutes, surfaces precedent. It is a faster, more conversational Westlaw. The underlying mechanism is retrieval -- find relevant documents, synthesize into an answer. It tells you what exists. It does not tell you whether your argument survives the attorney who knows the same law and is paid to break it.

ZugZwang — Adversarial Argument Intelligence

Attack, rewrite, and convergence

This is not primarily a search-and-summarize product. Its core mechanism is structured adversarial reasoning configured for legal use. Author constructs or defends. Adversary attacks. Judge scores and surfaces implications. The loop runs until convergence. The output is not a summary of what existing law says. It is the argument forged through structured opposition -- or the opposing argument mapped at its most exposed.

Legal research tells you what authorities exist. ZugZwang shows where an argument bends or breaks under structured opposition. Those are different problems. This is the tool for the second one.

Privacy Architecture

Your litigation intelligence
stays in
your office.

Your DEFEND sessions contain the weaknesses in your argument that you found and addressed. Your ATTACK sessions contain your analysis of every exploitable weakness in the opposing filing. Those are the most sensitive documents you produce in active litigation.

Cloud-hosted workflows can create additional confidentiality, waiver, and discovery questions that vary by jurisdiction and deployment. Prompts, outputs, and logs routed through third-party infrastructure introduce exposure that on-premises architecture avoids by design.

ZugZwang runs on CloseVector hardware in your office. No external model API calls during session execution. Prompts, outputs, and session records remain on firm-controlled hardware unless your team chooses to export them.

A contractual promise that your vendor erases your data is not privacy architecture. Air-gapped infrastructure that never transmitted the data in the first place has nothing to delete and nothing to explain.

01
CloseVector hardware in your office Purpose-built air-gapped legal AI workstation. Configured to your firm's workload. Deployed by CloseVector, owned and controlled by your firm. Specifications discussed under NDA.
02
No cloud dependency during operation Local model inference on-premises. Sessions run entirely on firm-controlled hardware. No external API calls required during a session.
03
Full audit trail on your hardware Every session logged with hash-based integrity records to support documented attorney review. Your firm controls retention, access, and custody settings. Production decisions remain with counsel, subject to applicable law and court order.
04
High-sensitivity confidentiality architecture Local inference on firm-controlled hardware for matters where prompts, outputs, and logs are too sensitive for ordinary cloud workflows. Attorney-client privilege and work product protection remain in the attorney's hands.
05
Patent-pending architecture The CloseVector air-gapped inference pipeline is patent-pending. The architecture that protects your litigation intelligence is proprietary.

The Machine

Purpose-built air-gapped
hardware designed for
litigation intelligence.

ZugZwang runs on CloseVector hardware configured for your firm. The system is deployed by CloseVector engineers and owned by your firm. The machine lives in your office, behind your network boundary, under your control. No sessions route through external infrastructure. No model calls go to third-party APIs.

Deployment On-premises in your office
Model Inference Local, hardware-resident
External Calls During Session None
Data Residency Firm-controlled hardware only
Session Logging Hash-based integrity records on local storage
Specifications Discussed under NDA

Performance tailored to your firm's session volume. Configuration discussed during deployment conversation.

ZUGZWANG

Your argument deserves
pressure from inside.

We run a 45-minute demo in the mode that matches your matter. You enter the material. The engine runs on your facts. No pitch. No generic sales deck. Just the tool running.

Request a Confidential Walkthrough

All sessions run under NDA. Your attorneys control privilege and work product decisions.