Start Here
The library holds essays, field guides, papers, thought experiments and case studies written over several years. It rewards reading in order, but not in the order it was written. These three paths are shorter routes in, each built for a different reader and each limited to a handful of pieces.
The path for lawyers starts at the failures that lose motions and ends with tools you can use before your next filing. The path for legal technologists starts with design rules and ends with systems that were built, run and measured in public, failures included. The path for academics starts with the history of computational law and moves to the open empirical questions, where replication and criticism are most useful.
Each piece carries a note on why it is in the path. The paths overlap on purpose: a few ideas, such as proposition usability and the gap between citation checking and filing-grade work, sit underneath everything else. When you finish a path, the Theses and Open Problems show where the arguments currently stand.
Start with the errors that pass every citation check and still lose motions. Finish with protocols you can run, with or without AI, before your next filing.
For: Litigators, supervising partners, law librarians and anyone who signs work that a machine helped produce.
- 01Right Law, Wrong StageThe clearest account of posture mismatch: a real, good-law, on-point case that is unusable because it was decided under a different standard, record, burden or forum.
- 02The Case You Never PulledA pre-filing completeness protocol that turns the four kinds of omission into a matrix in which no cell may be left blank.
- 03Read It Like Opposing CounselWhy adverse authority is a retrieval problem, and the three searches (exceptions, cure paths, distinguishing cases) that find the case against you.
- 04Good Law for What?The proposition-usability model: four verification questions that replace the single question of whether a case is good law.
- 05Case Treatment Classification GuideA field guide to reading treatment by holding rather than by case, with scope, severity and mechanism kept apart.
- 06Twombly Doesn't Live HereFor state-court practitioners: why tools trained on federal practice reach for plausibility pleading where it does not govern.
- 07When the Law Is Not SettledHow to tell a gap in your research from a genuine split in the law, and why the difference changes the advice you give.
- 08The Docket TestA one-afternoon method for evaluating a legal AI tool against a motion your own court has already decided.
Start with the design rules and the argument against treating verification as architecture. Finish with systems that were built, run and measured in the open, failures included.
For: Engineers, product leads, evaluators and procurement teams who build, buy or test legal AI systems.
- 01The Four Laws of System Design for Computational LawThe ordered design rules (confidence, explainability, grounding, testability) against which every other piece in this path can be checked.
- 02Why Retrieval Is Not ReasoningWhy chunk-level retrieval captures topics but not propositions, and what a structured representation of holdings has to record instead.
- 03The Verification Layer Is a Smoke Alarm, Not a Building CodeWhat post-hoc citation checking can and cannot catch, and the architectural principles that follow, starting with the rule that generators never originate citations.
- 04What a Filing-Grade Benchmark Must MeasureHow to design an evaluation that measures dispositive recall under constraint instead of rewarding topical proximity.
- 05Counter-Model BuilderA worked specification for an adversarial system: decomposition, standard first, citation audit, dispositions and gates that warn and log.
- 06isResponsiveAn open methodology for court-defensible document review that ties each design law to an enforceable control, from locked rubrics to validation certificates.
- 07We Ran a Blind eDiscovery Review with No Humans in the LoopA blind run with a hidden answer key, published with its failed gates and the narrower claim the evidence actually supports.
- 08I Built a Tiny Company Out of Robots, Gave Them House Rules, and Nearly Finished a Hard ThingA multi-agent experiment with separated roles and human red lines, and the lesson that stopping an agent needs as much design as starting one.
Start with the history of computational law and the gap between what the field specified and what shipped. Continue to the formal work on openness, inference and interpretation, where replication and criticism are most needed.
For: Scholars in law, jurisprudence, artificial intelligence and law, and measurement who want the theory, the formal claims and the open questions.
- 01The Long Road to Applied Computational LawAn essay on the field's intellectual history, from symbolic statute encoding through case-based reasoning and argumentation to the present.
- 02The Promise FulfilledThe argument that proposition-level representation at corpus scale meets requirements the AI and Law tradition set decades ago, with the distinction between structural and opportunistic reliability.
- 03Detecting Genuine Doctrinal AmbiguityThe shallow and genuine ambiguity distinction and a five-type taxonomy of legal indeterminacy, framed as detection rather than prediction.
- 04Ariadne's ThreadA measurement-theoretic account of legal openness as noise-corrected dispersion across judicial postures, with falsifiable conjectures and a disclosed smoke test.
- 05The Twelve BridgesA catalog of the recognized modes by which legal conclusions may be drawn from precedent, offered as a basis for auditing inference rather than citation.
- 06Canons of Constitutional Interpretation of GibsoniaA constructive counterfactual that enacts interpretive method before institutional power, and a test bed for version control as a model of legal change.
- 07Good Law for What?A verification standard stated independently of any system's architecture, grounded in the empirical literature on citator performance.
- 08What Remains for the Judge When the Machine Has Already Verified the Law?A Russian-language essay that decomposes judicial decision into mechanical verification and discretionary calibration, and asks what remains for the judge.