Legal Quants Weekly Digest — Week 13

The community announced the LQ Games — a hackathon pairing experienced legal quants as mentors with in-house counsel building their first AI tools. A deep debate on domain expertise unpacked what makes legal AI genuinely hard: tacit judgment, adversarial reasoning, and memory architecture matter more than fine-tuning alone. The "thin wrapper" bubble became a recurring theme as members dissected Harvey, Legora, and the structural tension between flat-rate pricing and model quality. A PDF on AI and professional negligence sparked a novel legal theory: negligent AI architecture as a malpractice cause of action. On the technical side, Claude Code launched mobile support, a study questioned whether CLAUDE.md files actually help, and the Local Models channel debated whether Document-to-LoRA could replace RAG (verdict: not for legal work, provenance matters too much). The community also began its migration to Discord.