Prerequisites: Students with any, or no, experience with programming or AI are welcome. We do require enthusiasm for experimentation, problem solving, and engagement with novel AI tools, including completion of a short vibe coding exercise before the first day of class (see note).

Exam Type: No Exam

As AI greatly augments some human capacities while leaving others untouched, lawyers face fundamental uncertainties: how will we learn and practice law, what kind of legal system will we practice in, and how will groups of any size — from legal practices to nations — function as each of us is represented by a growing ensemble of agents?

This hands-on seminar allows students to create and work directly with agents, discovering areas of untapped potential and unexpected danger. Students will build a working litigation dashboard before the first day of class — a task that has now become surprisingly easy — and spend the remainder of the course exploring the implications of that capability.

Weekly exercises include customizing the class’s shared AI assistant; assembling a legal knowledge base for the assistant to rely on; staging adversarial encounters between student-built agents; analyzing real-world litigation for low-quality and AI-generated submissions; designing an end-to-end dispute resolution system; building juror and judge agents; and probing existing bodies of law for newly-exploitable weaknesses.

Through solving collaborative practice problems, students will examine how AI agents are taking on a growing share of legal research, analysis, and adjudication, and learn to deploy, coordinate, and supervise such systems competently and ethically. Each week students will create a working artifact with the assistance of an AI coding tool and critique it under legal doctrines such as professional responsibility, principal-agent law, privilege, and access to justice.

Students will also explore agency through interactions with personal AI agents and shared agents created and modified by the class. Agents will participate in course discussions and individual work and assist in authoring a shared statement of principles developed through the seminar.

Practitioners, technologists, and faculty with relevant expertise (or their agents!) may be invited to join as guests throughout the term.

In lieu of a long final paper, students will be assessed on the basis of participation and work completed throughout the course, followed by a short self-reflection paper.

Note: Admitted students must complete a short self-guided vibe coding exercise before the first class session. A tutorial and instructor office hours are available; students who cannot complete the exercise should contact the instructors rather than withdraw.