Posts

Every major AI lab prices inference below cost. When the venture capital subsidizing your five-cent contract review runs out, your AI economics change whether you're ready or not.

Krafton's CEO bypassed his lawyers and asked ChatGPT how to avoid a $250 million payout. A Delaware court used those chat logs to rule against him — and the case is a warning to every executive treating a chatbot as a confidential advisor.

Federal courts are working out how attorney-client privilege and work product apply to AI prompts. The doctrine hasn't changed — *Hickman*, *Upjohn*, and *Kovel* still control. Here's how each case applies the existing elements.

A practical walkthrough of Claude Cowork across the litigation lifecycle — organized around Projects for matters and Skills for recurring tasks — plus the privilege question every firm needs to answer first.

Every major legal AI vendor shipped autonomous agents in Q1 2026. Here's what they actually do, what can go wrong, and why your ethical walls weren't built for this.

Ten research-grounded predictions for legal AI through the end of 2026 — from the first disbarment for hallucinated citations to the collapse of point-solution vendors to the pricing collision between AI-enabled firms and their clients.

The PPP fraud pipeline worked because the SBA released everything. Medicare's public data is fragmented, de-identified, and missing the features detection needs. Here's what exists on GitHub, where it falls short, and what CMS would need to release to let outside analysts do for healthcare fraud what one Python repo did for PPP.

How Quinn Emanuel, Mayer Brown, Outten & Golden, and others deployed AI on real matters — with real numbers

Public data can source prosecution leads. An open-source fraud-scoring system, run against the full SBA PPP dataset, identified the same lenders, geographies, and loan populations that DOJ prosecuted — using nothing but a downloadable CSV and a standard laptop.

Chinese labs aren't just catching up — they're pioneering the techniques Western models adopt, sharing them under open licenses, and training them on chips that weren't supposed to exist.