blog
Writing
Research, technical deep-dives, and opinion on AI verification, hallucination detection, and deploying LLMs in regulated industries.
Why LLMs Hallucinate More in Regulated Industries — and What to Do About It
Financial services, healthcare, and legal domains share a common property: the correct answer is often a specific citation, statute, or figure, not a general pattern. That specificity is precisely what makes LLMs unreliable in these settings.
Read more →
Anatomy of a VRL Proof Bundle
A walkthrough of every field in a VRL Proof Bundle — what it contains, why each piece is there, and how to review the evidence record behind a verdict.
Read more →
AI Verification vs. Human Review: When You Need Both
Human review of AI output is not going away — but the question of what humans should review, and when, changes completely once machine-level verification is in place. The right model is a triage layer, not a replacement.
Read more →