The legal industry is particularly suited for AI because the value of each token in legal documents is high, making the efficiency of AI applications extremely valuable. If you focus on building a specific workflow, the value can be very high, but if it doesn't sell, you're in trouble.

The perception that AI startups are merely wrappers around foundational models is misleading; many are creating valuable and innovative companies despite the risks of failure.

The legal industry is on the verge of a major transition where lawyers will increasingly use AI to enhance access to justice, despite existing regulatory challenges.

Transforming raw AI models into usable products for lawyers involves significant work, including filtering relevant documents from vast amounts of data.

The AI industry suffers from a lack of rigorous evaluations and tests that demonstrate the frontier of model capabilities.

The concept of agents will disrupt traditional economic models, as AI will provide services at the cost of compute, leading to an abundance of AI lawyers and other professionals.

The biggest surprise for me is how well AI performs in legal work, which I initially thought would be too complex for it to handle effectively.

The marketing narrative around agents often suggests they can replace human workers, which can lead to inflated pricing models for AI software.

The benefits of advanced AI models may not be evenly distributed, as many people are unaware of the AI advancements that are already impacting industries.