Writing

Notes & Essays

Observations on capital, technology, and the forces shaping both.

  1. The Five Types of AI Hallucination in Legal Filings

    Not all AI hallucinations look the same in a legal filing. The five-category taxonomy from Ethics-Safe AI Use for Law Firms maps each failure type to the verification step that catches it.

  2. AI Ethics Training for Attorneys: What the Sanctions Record Actually Requires

    Most AI ethics training for attorneys is built on theory. This one is built on 1,356 documented sanctions cases. Here is what the record shows courts and bar associations actually care about.

  3. 1,356 Cases: Publishing the AI Hallucination Legal Database

    We've been tracking every court proceeding where generative AI produced fabricated citations, invented quotes, or misrepresented legal authorities. The full dataset is now public.

  4. AI Hallucinations in Legal Research: What Actually Goes Wrong and How to Catch It

    Not all AI errors in legal research are the same. A five-category taxonomy explains how hallucinations differ from citation errors, and what verification step catches each.

  5. California's New AI Ethics Rules: What Six Rule Changes Mean for Your Practice

    California's bar just proposed the most detailed AI-specific ethics amendments in the country — six rules, binding enforcement, and a mandatory verification requirement for every AI output. Here's what each change requires.

  6. Why Most AI Demos Fail Real Users

    Most AI demos are optimized for the pitch, not the problem. A few observations on the gap.

Published work