Pipeline

How this site produces its analysis

Methodology

Every piece published on The Fiduciary follows a four-stage pipeline. AI systems handle research, preliminary analysis, and drafting. A licensed attorney makes every editorial and publication decision. This is what accountable AI-assisted work looks like.

This is what the Minnesota Digital Trust & Consumer Protection Act would require of any AI-assisted professional service: transparent methodology, clear attribution of AI and human contributions, and human accountability at the point of publication.

This site practices what the Digital Trust Act would require.

Current Status

Pipeline Status

●Gemini 3 Flashmonitoring

Last: 1h ago

7 items pending review

○GPT-5.2idle

Last: Feb 21 — Heppner case analysis

3 items pending attorney review

○Opus 4.6idle

Last: Feb 20 — Heppner analysis draft

Queue empty

●Humanactive

Last: Now

Reviewing 2 drafts

Model Cards

Stage 1: Research

Gemini 3 Flash

Continuous monitoring of federal and state legislative databases, court dockets, regulatory filings, and legal scholarship for AI-related developments.

Stage 2: Analysis

GPT-5.2

Structured legal analysis: identifying holdings, reasoning, statutory interpretation, cross-referencing existing tracker entries, and assessing Digital Trust Act implications.

Stage 3: Drafting

Opus 4.6

Long-form legal writing in publication voice. Produces case briefs, commentary, legislation explainers, and policy analysis with proper legal citation formatting.

Stage 4: Attorney Review

Human

Final editorial review by a licensed attorney. Verifies legal accuracy, citation correctness, analytical soundness, and editorial voice. The human is the bottleneck by design.

Transparency

Every published piece on The Fiduciary includes a pipeline transparency footer showing exactly which AI models contributed to the research, analysis, and drafting — and who performed the final attorney review.

Stage 1 (Research) displays live status from the pipeline infrastructure when available, with graceful fallback to static data. Stages 2–4 display static status until implemented.