Legislative analysis systems engineering
Bills, acts, amendments, ordinances and agency filings enter a fixed pipeline. Nothing reaches publication that a gate has not cleared, and a gate that is registered but never invoked fails the build.
Deterministic engine architecture
Our core engine executes multi-threaded agentic loops within a sandboxed runtime, translating volatile constituency trends into auditable state transitions.
Stochastic projection pipeline
Quantitative work — load growth, rate impact, generation capacity, fiscal exposure to a proposed appropriation — runs parallelized Monte Carlo and Latin Hypercube sampling. Output is published as a distribution with its sample count and seed stated on the artifact, never as a bare point estimate.
Traceable publication path
Every published artifact is read back from the bytes the public actually receives, not from the database record describing them. A publication that cannot be verified against what is served is refused rather than recorded.
Method
Risk sciences, civic subject matter
The methods here come from quantitative risk science: explicit distributions, stated assumptions, uncertainty carried through to the output instead of rounded out of it, and a refusal to publish a central estimate without the spread around it. What is unusual is not the mathematics. It is the subject — legislation, ordinances and agency filings, a field that has mostly been served by assertion.
Legislative analysis fails in a specific way: fluent prose that no one can trace back to a document. Source classification, claim grading and publication are separate stages here, each with its own gate, and none of them takes another's word for it. The output is built for people who will be asked where a number came from — analysts, legislators, county and municipal officials, reporters, and residents reading about a project in their own county.
Review the architecture
Technical specifications, gate inventory and publication-path documentation are available on request.