Narrative Systems
I’ve spent the last two years building AI narrative systems that generate briefs, notes, rewrites, continuity, and production docs, with a craft layer that makes frontier models perform like experts. Litigation strategy and screenwriting turn out to run on the same logic — both live or die on the quality of the timeline.
Smithee
A writers' room without the room. Or the writers.
Smithee is built for high-volume content studios with short timelines. Start with anything from a few words to a nearly finished script and it produces the development package in hours: manuscripts, screenplays, and production documents for short-form, features, and minidrama series. A blind 11-axis craft rubric across 40 comparisons puts Smithee’s writing quality at 30% above the consumer AI average. The gap is widest in character DNA, causation, and narrative voice — the elements that separate a story where every choice feels inevitable from one where characters just do things until the plot is done.
Serialized work is where the complexity stacks. Every open hook needs a landing spot. Every cliffhanger has to pay off at the right moment. Every story thread has to hold across sixty episodes without losing what episode three built in. Smithee tracks all of it and keeps the writing moving at the pace the production needs.
Summit6
Litigation Intelligence for Medical Malpractice
Summit6 takes the same habit into medical malpractice litigation. It ingests thousands of pages of medical records, pleadings, and expert reports to rebuild the chronology, and turn it into cited strategy: causation and damages analysis, deposition prep with impeachment threads, settlement intelligence, and draft arguments built from the record. In multi-defendant cases it maps liability allocation, cross-claim opportunities, coalition risks, and insurance tower dynamics.
The citation layer is the spine. Every claim stays chained to its source. In a market where lawyers face sanctions for submitting documents built on LLM confabulations, verified output is the whole product.