Lawyers, Humility, and LLMs
Apparently I’ve been raving about The Three and a Half Minute Transaction: Boilerplate and the Limits of Contract Design by Mitu Gulati and Robert E. Scott for 12+ years now without blogging about it.
And now it might be more relevant than ever. So here’s a brief review.
“CONTRACT” by Steve Snodgrass is licensed under CC BY 2.0
#The basic premise
In programmer-speak, the book is about a vulnerability in a contract that was discovered, left unpatched, and exploited again. For programmers, this is a story as old as time: we have bugs, and we know it; so the question is not “why did we have a vulnerability”, the question is “why did we not fix it”.
For lawyers, the story is subtly and importantly different. So let me try to explain it again, in detailed English this time.
Contracts for “sovereign bonds” (bonds sold by countries) are fairly standardized and fairly old. They’re also quite high-stakes: billions of dollars rest on their interpretation. For many reasons, the language in these contracts often stays the same. In particular, the sentence the book is concerned with has been reused, with only slight variations, since the late 1800s.
But: in the late 1990s, “leading practitioners … suggested that no one knew what the clause actually meant”. In 2000, a court said “this sentence means X”, which created an uproar since when pressed most specialists in the area thought it meant not-X. The lawyers who regularly wrote these contracts thought this court’s interpretation was awful, and… continued using exactly the same language. They took no steps to fix (“patch”) the text.
That resulted in a multi-billion dollar litigation—which, in 2016, led law firms to finally start “fixing” the language in new bond contracts.
In other words, the book tells a story of world-class lawyers who were nevertheless resistant to doing the sort of fixes that programmers take to be somewhere between “routine” and “mandatory good practice”.
#Who cares?
I’ve long been interested in this story for many reasons, including what it tells us (or not) about open source licenses, which often copy language and concepts from each other and are notoriously hard to update in response to judicial changes.
I also find the book a good read for a bunch of reasons, including that is a great example of mixed methods: it does deep interviews of the lawyers involved, but also does deep historical research to detail the historical evolution of the clause—and show how the history and the stories told about that history don’t quite match up. The lawyers in question are smart and curious! But also they are humans embedded in a workplace culture, and economically rational actors. So who wants to be the one who spends time and effort saying the emperor has no clothes?
These are admittedly niche interests, so I tend to press this book only on the nerdiest of law nerds.
#Why now?
However, increasingly, the book feels relevant to me because it should be a tale of humility.
If some of the world’s highest-paid lawyers, at the world’s highest-status firms, do deals worth tens of billions of dollars with language they don’t understand, what does that say about the law’s pretensions to high standards?
#In other words, yes, LLMs
Yes, like everything else in 2026 this is actually a post about LLMs.
I do not particularly think that lawyers can be replaced by LLMs, at least not most good lawyers. (See, eg, Gergely.)
But many, many lawyers tend to think that the problem with LLMs in law is that Lawyers Have Incredibly High Standards, and LLMs… do not. Cannot. And that is supposed to settle the discussion.
So I think about this book a lot lately, because this story we tell about our profession’s irreplaceable skill and diligence is, at least in this one case, very demonstrably not true. The story is false for some very good reasons of human nature and economic rationality. But it is nevertheless false.
I don’t know what that means, exactly, for the future of LLMs and law. But I do recommend every lawyer who defends our profession to read this book, and ponder what it might mean to look at new tools with more humility about our own skills—and shortcomings.