The LegalQuants Podcast

In this episode

• How Andrew sizes up a new legal AI product: how was it probably built, and does the business case reach beyond a single use case?
• Why retrieval is one of the hardest problems in legal AI, where one changed word can carry outsized weight.
• Why Weil runs closed frontier models alongside open-weight models, and pins some small "tissue paper" apps to avoid churn when a new model drops.
• How a lawyer's prototype becomes a real firm app, and why a working prototype is "the world's best spec document."
• Why Weil decentralizes innovation, and the three hurdles behind "tissue rejection": understanding an idea, believing it can be built, and changing how you work.
• How BenchMark, which analyzes a judge's public record to help litigators prepare, grew out of a client's question.
• Why LASER stays small, what Andrew asks technical hires to explain, and what success would look like in three years.
• Machine-to-machine legal work, the case for an industry standards body, raising kids in the age of AI, and the one thing he won't hand over to it.