Builds
Each build has a repository or a live tool, a write-up when one exists, and a slot for a short demo video. These are independent projects, not client work, and they show the approach rather than promise results.
A governance layer for AI agents in professional work. Every action gated, receipted, and verifiable. Built on Paperclip, Apache 2.0, first proven on the business side of a legal practice.
Operations tracking that lives where your team already works. A Chrome extension, so there is nothing to install on a server and nothing leaves the browser. Maps the work to the CLOC Core 12 and turns it into a shareable picture of demonstrated skill.
A sparring partner for what I publish. Everything I read, save, and think about goes in. What comes out is a shortlist of what is worth writing, with the sources attached. The writing is still mine.
Three ways to start a build with an AI agent: a one-page CLAUDE.md plus build spec, a five-file harness for a coding agent, and a four-document harness for a chat-based builder. MIT licensed.
A shared synthetic universe of matters, parties, and documents so every exemplar repo and lab runs on realistic data that is safe to publish.
Claude Code features, taught through legal practice. Each feature paired with a task a lawyer or legal ops lead recognizes.
Redline explainer, data room QC, matter intake triage, conflicts screen, pleadings matrix, citation ledger, NDA playbook review, contract tabulator, and more. Small, complete, and archived so the pattern stays readable.
Applied AI learning for professionals who do the work. Build useful things on real work, direct AI tools, and leave with visible evidence of capability. Free tier, paid paths, an AI tutor.
Client transformation work is confidential. Archived repos stay public so the pattern remains readable. Labs and exemplars run on synthetic or explicitly approved material.