Writings

Methodology writeups, for the practitioners who actually use this work

Research and methodology pieces on cryptocurrency forensics in contested matters, LLM application security, and adversarial testing — written in language a family-law attorney, a developer, or a security engineer can act on.

Civil Litigation · June 2026

How to Challenge a Cryptocurrency Tracing Report

A blockchain tracing report is expert opinion, not measurement. After the December 2023 amendments to Federal Rule of Evidence 702, the proponent has to show reliable application of the method, not just confidence in it. Six places confident tracing reports tend to break down under Daubert and Rule 702, framed for attorneys.

Family Law · June 2026

Where Hidden Crypto Actually Hides in a Divorce

Standard financial discovery sweeps past self-custodied cryptocurrency because there is no custodian to compel. The strongest first evidence usually is not on the blockchain at all. It is in the fiat trail: bank statements, credit reports, tax filings, and properly directed exchange subpoenas. Two distinctions matter once an asset surfaces, and getting them wrong can send a case sideways.

LLM Security · 576 trials

Does reasoning make LLMs safer against prompt injection? Testing Qwen 3, DeepSeek-R1, and Gemma 4

A 576-trial controlled study across four reasoning conditions (Qwen 3 8B thinking off and on, DeepSeek-R1 8B, and Gemma 4 e4b). Reasoning reduces conventional prompt-injection susceptibility (Qwen 3: 64% to 54% genuine injection), but reasoning-targeted payloads sidestep the defense, and reasoning costs roughly 15 to 20 times more tokens per call. The conclusion: reasoning belongs alongside input and output controls, not instead of them.

LLM Security · 1,280 trials

Temperature is not a defense: indirect prompt injection across four open-weight LLMs

A 1,280-trial study characterizing indirect prompt injection susceptibility across Llama 3.1 8B, Mistral 7B, Qwen 2.5 7B, and Qwen 2.5 Coder 7B at two production-realistic temperatures. Temperature reduction is not a reliable defense. Output-format constraint achieved 0 of 40 injection on the most-susceptible model tested.

Start a Conversation

Has a piece here matched your situation?

If something written here describes a matter you are working on, reach out. A short scoping call will tell us whether it is a fit.