Chris Sidebottom
Legal Engineer • J.D., LL.M., CIPP/US • Denver, CO I am very interested in what human reasoning looks like alongside these burgeoning language models. I use both local models on my personal rig and frontier labs to explore this. But rest assured, this is my human reasoning. All em-dashes mine.
Whatever has brought you here, this is my digital home where I put things for my own safekeeping and organizing. But it is also a great place for you to poke around and see what I'm up to.
As the Greater Internet continues to evolve into fractured platforms forcing us to perceive and be perceived through many, many lenses, I've decided to start consolidating all aspects of my conspicuous existence here.
Legal Engineering & Applied Systems
In commercial practice—especially in large organizations with many silos—agreements don't live in isolation. They live in messy, asynchronous stacks: master agreements altered by amendments, conflicting SOW terms, Word redlines, degraded photocopy scans bound by strict NDAs, and every third entry missing the fully executed version the new associate got signed at 4:58pm.
To test and benchmark legal AI without breaching client confidentiality, privilege, or institutional data policies (and to steer clear of third-party frontier data leaks), I built this suite:
-
Fabric8r: Procedurally generates multi-document artificial contract stacks (ReportLab PDFs, Word tracked-changes redlines, degraded scans) with SHA-256 verifiable answer keys (
truth.json), protected by a 6-word shingle similarity filter to prevent boilerplate collapse. The fuzzy PDFs make me feel right back at home in a corporate CLM. - Nova → An interactive property graph workbench that maps complex contract stacks and audits institutional governance rules (like IP assignment gaps and faculty dual-engagements) across entire portfolios.
- Rest8r: An AST-based engine that parses base agreements and cascading amendments to deterministically compile fully synthesized restatements.
What happens when we treat a contract stack not as static text, but as a computable state machine? Currently, I'm exploring comparing extracted knowledge graphs across modalities—using VLMs alongside LLMs to analyze document layout as an artifact of legal reasoning. If we can model contractual operations (assignments, extensions, novations) as mathematical mutations applied to a graph, the drafted contract becomes merely the exported semantic view of an underlying, verifiable system.
On Legal Reasoning & Language Models
In the mid-20th century, as the world settled into a post-war status quo, Western (read: anglophonic) legal philosophy made several attempts at formalizing our human systems of law into logic and computer-legible reasoning. Drawing on Hohfeldian jural primitives, H.L.A. Hart's analytical jurisprudence, and the early rise of cybernetics and symbolic systems, legal scholars sought to model the rules of the land... Most of those efforts stalled or continue to be ground out through painstaking annotation and proof — the slow, harsh work of knowledge creation. Most legal tech for a long time has been largely database management: where are the rules of the land stored, and where are the facts of the case at hand stored?
In recent years, though, the rise of LLMs has also heralded a renewed interest in technological aids to legal reasoning in and of itself. Most legal AI benchmarks evaluate large models on the strength of their semantic encoding. They are powerful legal tools because language is a powerful legal tool. Into our languages, we have encoded our understandings of the world around us, our capacities for poetry and love, and the harshly-wrought rules of logic and reason that we use to govern our legal systems. An LLM can get pretty far with a contract because contracts wouldn't exist without some form of language—some avenue through which minds can meet, so to speak. This work explores the combination of deterministic tools, inferential models, and human reasoning to improve the quality of legal tech and to increase access to justice by lowering barriers to good, correct legal rigor.
Background & Approach
I am a trained attorney by trade, with 12 years of experience in transactional work for research institutions and data privacy. But really, I like to solve interesting problems. Sometimes that's been intellectual property work; at other times, it's been helping streaming companies fix their email marketing stack to maintain legal compliance. Once, I learned 3d resin printing to help a jewelry company make counterfeit-proof designs.
My background in law is more academic and philosophic than most American attorneys; my JD is from LSU and my LLM from Université Jean Moulin Lyon 3. I am fluent in common law regimes, civilian and codal contexts, and in a general comparative legal posture. Anymore, civilian regimes and common law regimes tend to reach similar legal conclusions, but it's a bit like choosing whether to use calculus or finite analysis to arrive at the answer. I'm CIPP/US certified, and while I have no formal software engineering training, I have been breaking and fixing hobby projects since I was 10. The projects above may showcase my approaches in helpful ways. If you're interested in how my particular brand of thinking abreast the box could help you solve interesting problems, don't hesitate to shoot me a message.
Other Explorations & Games
Five Card Charlie → is the beginnings of a Balatro-inspired blackjack trainer. The roguelike playloop is fascinating to me: if Bernard Suits is right that games are real problems turned frivolous, the iterative learning encouraged by a roguelike or roguelite is maybe instructive for solving real life problems. Blackjack as a game has clear levels of skill expression (basic strategy, count maintenance, rule variant tracking, bet management, detection avoidance), so it feels like a natural object of an iterative playloop. I've vibe-coded most of the functionality using local instances of Qwen and Mistral as well as Claude Opus 5, Opus 5.5, Fable 5.1, Sonnet 5 and 5.5, and Gemini 3.8 Flash High for architecture and review.