Founder
How and why CaseGenie was built.
What CaseGenie is
CaseGenie is an evidence-intelligence platform for legal AI. It reads a body of documentary evidence, integrates the record as a coherent whole, surfaces chronology, contradictions, entity relationships, and gaps, and produces a first-pass case analysis whose claims are tethered back to the source documents themselves.
The architecture rests on a structural claim about legal AI: that systems organized around chat or fragmented retrieval lose two properties at the moment they begin analysis — the independence of their first reading, and the integration of their reasoning across the record. Once those are gone, no amount of careful prompting recovers them. The architectural response is conjunctive: touchless ingestion paired with whole-record reasoning, supported by a bounded perimeter, source-tethered citation, tenant isolation, and append-only audit logging.
The full argument is developed in two working papers and a five-essay public arc, listed at /research. A U.S. provisional patent application specifying the technical mechanisms was filed on May 1, 2026 and is summarized at /patent. What follows is the story of how this architecture came to be.
The access gap
I did not start using AI because I thought litigation should be easy. I started because the kind of case I believed I had was too expensive to investigate and litigate conventionally.
Lawyers told me complex civil litigation of this kind could cost anywhere from hundreds of thousands of dollars to several million, take years, and require a level of document review I could not fund. The gap was not just between having a grievance and filing a lawsuit. It was between having evidence and being able to pay for the first serious analysis of it.
That gap is where CaseGenie was born.
What attorneys told me
The summaries below are lightly anonymized excerpts from attorney outreach in September 2024. They are included not to single anyone out, but to explain the economic reality that shaped the product.
Plaintiff-side litigator
“Even a relatively contained matter of this kind can come close to $1 million in fees and costs. A larger one could easily exceed $3 million.”
That was the first time the scale of the access problem was stated that bluntly.
Practice-group chair
“This kind of matter is complicated. It would take digging. The other side looks well funded and could bury you in paper. We do not take matters like this on contingency. A retainer would be a substantial undertaking.”
The message was not “you definitely have nothing.” It was: the review itself is expensive.
Litigation partner
“To file and litigate the case you are contemplating would require a great deal of work and could take years. The fee could run as high as $250,000. We will not take it on contingency.”
That confirmed the same reality from a different direction.
The lesson
I used general-purpose AI tools too early in the process — too close to drafting, before a sufficiently grounded analysis of the underlying evidence had been done. The result was the structural failure the working notes now name prompt stain: fluent, confident output whose factual foundation had been shaped by my framing before the system had independently read the record.
That experience did not end my use of AI. It changed it. I stopped asking it to write first and started asking it to read first.
What I built
I spent the next year building the system I wished I had had at the start: one that reads the record, integrates it as a whole, maps the facts, surfaces contradictions and gaps, and builds a trustworthy analytical foundation before a single word of advocacy is written. The contamination problem was named first. The architectural response followed. The architecture was then specified in enough detail to file as a provisional patent.
The system is now in operation at CaseGenie.ai. Four working sample cases — Rivera, Kaplan, Martinez, and Santos — allow the architectural claims to be inspected directly against complete document sets.
Current work
The body of public work continues at beforeadvocacy.substack.com. The two working papers are at /research. The patent summary is at /patent. Press and background materials are at /press.