Legal Quants Weekly Digest — Week 15

The LQ Games hackathon dominated the week as mentors completed scoping calls with all participants and published detailed project briefs — contract redlining, regulatory change tracking, invoice review, document assembly, and agentic decision oversight were the most common themes, with mentors steering participants away from oversubscribed playbook-generation concepts. A "LawyerOS" discussion exploded: multiple members independently described building native desktop environments that replace the fragmented stack of Word, Outlook, Lexis/Westlaw, and ChatGPT with a unified, terminal-first platform where Claude generates ephemeral single-use tools on demand. The AI Research channel produced an 80-message thread on applying code-dependency-graph traversal (CC-RLM) to legal agreements, arguing that vector-similarity-based RAG fundamentally cannot follow defined-term chains or cross-reference hierarchies. Legora's $5.55 billion valuation drew pointed skepticism — members noted its Word add-in cannot read footnotes and questioned whether revenue multiples justify the number. Claude Code shipped visual design previews and a 1M-token context window at no extra cost, while members debated whether GPT-5.4 and OpenAI Codex are pulling ahead for certain coding tasks. A court ruling banning AI shopping bots sparked discussion, and a UK judgment caught a witness being coached through smart glasses during cross-examination.