Anthropic's David Hershey states it's "deeply unsurprising" that AI is great at software engineering because the labs are filled with software engineers. This suggests AI's capabilities are skewed by its creators' expertise, and achieving similar performance in fields like law requires deeper integration with domain experts.

To ensure accuracy in its legal AI, LexisNexis unexpectedly hired a large number of lawyers, not just data scientists. These legal experts are crucial for reviewing AI output, identifying errors, and training the models, highlighting the essential role of human domain expertise in specialized AI.

A core debate in AI is whether LLMs, which are text prediction engines, can achieve true intelligence. Critics argue they cannot because they lack a model of the real world. This prevents them from making meaningful, context-aware predictions about future events—a limitation that more data alone may not solve.

As AI models are used for critical decisions in finance and law, black-box empirical testing will become insufficient. Mechanistic interpretability, which analyzes model weights to understand reasoning, is a bet that society and regulators will require explainable AI, making it a crucial future technology.

The judicial theory of "originalism" seeks to interpret laws based on their meaning at the time of enactment. This creates demand for AI tools that can perform large-scale historical linguistic analysis ("corpus linguistics"), effectively outsourcing a component of legal reasoning to AI.

Unlike simple "Ctrl+F" searches, modern language models analyze and attribute semantic meaning to legal phrases. This allows platforms to track a single legal concept (like a "J.Crew blocker") even when it's phrased a thousand different ways across complex documents, enabling true market-wide quantification for the first time.

A top-tier lawyer’s value mirrors that of a distinguished engineer: it's not just their network, but their ability to architect complex transactions. They can foresee subtle failure modes and understand the entire system's structure, a skill derived from experience with non-public processes and data—the valuable 'reasoning traces' AI models lack.

A key job for junior lawyers is understanding non-legal context for a case, like a pharmaceutical supply chain. AI excels here by rapidly synthesizing massive amounts of diverse, industry-specific information alongside legal precedent, which is a core part of the value.

To improve LLM reasoning, researchers feed them data that inherently contains structured logic. Training on computer code was an early breakthrough, as it teaches patterns of reasoning far beyond coding itself. Textbooks are another key source for building smaller, effective models.