Learning Machines
Lawyers are learning to work with artificial intelligence. Artificial intelligence is learning to work with law. This blog explores how — through pedagogy, practice, policy, and the ethical questions that connect them.
-
From Beginner to Intermediate: Ten Habits for Lawyers Using AI
Most lawyers who use AI regularly have stalled at the beginner stage: they open the same tool, accept its default model, ask for a draft, and judge the result by how well it reads. Ten habits move a user past that point. Among them: know the models well enough to choose one for the task, use the model to attack your work at least as often as you use it to generate ideas, and keep a personal benchmark of your own tasks so that you can test each new release instead of relying on its announcement. The rest concern how full the context window is, the limits of the tool’s sources, when to refine or start over, finding new uses, and the judgment a lawyer has to keep exercising without the model.
Read more -
The Ten-Minute Conversation: What Clinical Supervision of AI Use Teaches Law Firms
A review article in the New England Journal of Medicine asks what an attending physician should do on seeing a resident consult a chatbot between patients, and answers with a structured conversation about how the resident used the tool. The framework translates directly to law firms, where the associate’s interaction with the model does not show in the memo that results. A ten-minute conversation with the associate before reviewing the work product, a habit I have been encouraging at my firm, is that intervention. It reaches the moment a review of the document cannot see, it teaches the associate by making them reason aloud, and it tells the supervisor where to spend the review.
Read more -
The AI Tells Reference Sheet, September 2026
The reference sheet I use to review my own drafts for the words, constructions, and formatting habits that mark prose as machine-written. Almost nothing on it is wrong in isolation, and much of it dates fast: the vocabulary tells of 2023 and 2024 are already weak evidence, the em dash collapsed as an authorship signal within a year, and the paste-through artifacts change whenever a vendor ships a new interface. Current as of September 2026.
Read more -
Durable Skills, Perishable Tools: Who Should Teach the AI-Era Lawyer What
Law firms spent this summer teaching their summer associates how large language models work, while more than half of law schools say they already offer AI classes and a third of students say they are not learning the skills their careers require. The mismatch comes from a failure to divide the curriculum. Law schools should own the durable skills: how models behave and fail, prompt and context fundamentals, verification, and the professional-responsibility overlay. Firms should own the perishable ones: their tool stack and their practice-specific workflows. A graduate who has to be taught the fundamentals at the firm starts every practice skill late, and firms can enforce the division from the hiring side by evaluating AI competency the way they evaluate writing, testing for the understanding that survives tool churn rather than for familiarity with whichever products a school happened to license.
Read more -
From Tokenmaxxing to Token Anxiety
In April, the story was tokenmaxxing: a leaderboard at Meta ranking employees by AI token consumption, as if the meter itself measured productivity. By summer, the story had inverted. Uber capped employee AI spending after exhausting its annual budget in four months, Accenture told staff to stop feeding PDFs into slide generators, and the vendors themselves began publishing rate limits. The emerging workplace condition is token anxiety, and the legal profession has run this experiment before. The research-anxiety literature from the era of per-search Westlaw charges shows exactly how cost-consciousness curdles into worse work, and how to prevent it.
Read more -
Unknown Unknowns: What AI Legal Research Can’t Read
On July 16, 2026, I pulled the robots.txt file for 34 hostnames spanning the legal trade press, the financial press lawyers read alongside it, and the primary sources both report on. The comfortable assumption is that crawler blocking is harmless because the primary materials remain open. The files say otherwise, selectively: enacted law is reachable, but congress.gov blocks roughly 130 AI agents from the entire site, the largest free case-law corpus blocks them from case data, and the financial press is either walled off entirely or open to exactly one AI company’s bots. The gaps are topic-shaped, vendor-shaped, and priced: what AI research tools cannot see maps closely onto what publishers know the profession will pay for, and no tool discloses any of it.
Read more -
Chicago Law's 1L Device Ban and Two Effects It Doesn't Weigh
The University of Chicago Law School will pilot a no-device rule across all first-year core sections this fall, part of a thoughtful AI strategy that gets the sequence right: build judgment without the machine before layering the machine on top. But the classroom ban runs on two assumptions the strategy statement never examines. A disability carve-out does not keep an accommodation confidential; in a room where screens are otherwise forbidden, it exposes the students it exempts. And the case that forcing longhand improves learning rests on a study whose central finding has failed to replicate. If the resulting exams are handwritten, the school risks measuring penmanship alongside the legal reasoning it means to assess.
Read more -
Consumption Pricing and the Return of Cost-Effective Research
Legora moved its most capable product to consumption-based pricing, which means the bill now tracks how the tool is used. Lawyers who remember per-search Westlaw charges already know what that does: it turns efficient use into a billable skill, and it puts the cost of an associate’s fumbling on a dashboard the client can read.
Read more -
The Verification Standard: Training New Lawyers to Use AI
Law firms have rebuilt their workflows around generative AI, and the document review, first drafts, and disclosure-schedule checks that once trained associates are now the tool's work. Prohibiting the tool is unrealistic and counterproductive; handing it over with no change to how juniors are trained lets them skip the work that builds judgment. The standard that fits is verification: let new lawyers use AI from the start, but require corroboration of every substantive output against independent authority, and against more than one source when the training is serious.
Read more -
Three Approaches, One Goal: AI Policy at Berkeley, Texas, and Boston College
Within a few weeks, Berkeley adopted a rule restricting student AI use, the dean of the University of Texas sent his faculty a letter, and Boston College Law named its first director for AI. The three read as a disagreement, but they serve the same goal the ABA already sets for every law school: graduates who can do legal work and exercise judgment, competently and ethically. What divides them is the instrument and the sequence, when and how to build legal judgment and AI fluency into the same lawyer. Berkeley’s ban is the broadest, and because much of it cannot be detected or enforced, it would work better as a statement of values than as a policy.
Read more