What a Law School Evidence Exam Reveals About AI and Legal Writing
The test looked like Frogger. It was really Tetris. AI will play excellent Frogger — and never know a different game is on the table.
By R.R. Hicks · Forlorn Hope Publishing
The Law School Story Behind the Frogger and Tetris Analogy
I was reasonably well known in law school for four things:
- Running Torts Bingo—bingo cards with squares such as “woman in third row, fourth seat from left asks question that was just answered” and “ponytail dude in second row starts with ‘I was gonna ask’”;
- Starting a Contracts fill-in-the-New York Times crossword—$5 in the pot, each person fills in until they are stumped, winner is the last one standing, if more than one hasn't bombed out (hello Tuesdays), the pot rolls over to next class;
- Spotting issues on a dime;
- Being the greeter—first at the bar after an exam, duties included buying the first pitcher, dropping a few bucks in an awesome jukebox, and confirming issues on tests.
For the unfamiliar: law school finals in core courses are the only grade.
One exam, one shot, three or four hours in a room with a fact pattern, maybe an outline, always your own logic.
My section mates and I had a system—finish the exam, get to the bar in Tribeca, debrief.
First question everyone asked two steps into the door: Did I get the issue right?
Which was really asking: am I getting an A/B+ or B-/C?
I was usually one of the first out of the exam room.
Spot the issue, write it up, apply the facts, conclude, get the hell out of there.
Four of us once had a contest to see who could waltz out of Sinclair's Contracts final first—he was funny, pragmatic, and had made it clear that he tolerated neither dicta nor bullshit, and we treated the exam as such.
It was a virtual tie.
It worked like this for a few semesters.
Then came Evidence.
The Exam That Rewrote the Rules
Four-credit course, three-hour exam, one-page exam.
A fact pattern for about two fairly short paragraphs, then a long list—A through X—of short descriptions of evidence and statements.
The next-to-last line was one question: what's admissible, what's not, and why.
The last line was a bonus question good for 10 points (whatever the hell that meant): what work of fiction are the characters from?
The fact pattern was fun, the names were from Gravity's Rainbow, the evidence was downright funny and clever, though I don't remember if bananas were mentioned.
This time, I had the unique experience of writing away as the room emptied.
A couple of my friends slowed down by my desk and gave questioning looks—or maybe they were pissed they'd have to buy that first pitcher.
I was the last one in the room, the last one in the bar.
My friends were worried.
Had I frozen?
Blanked on hearsay exceptions?
Confused myself into paralysis on the difference between a present sense impression and an excited utterance?
I was peppered with questions and couldn't get near a Bass Ale for several concerning minutes.
“Man, it wasn’t that hard—what took so long?” my study partner asked as I finally inhaled an ale.
“Yeah, but it was kinda complicated, right? I mean, I took half the time just to diagram it all out … then it was easy . . . well, easier, anyway.”
She shot me a strange look.
Another friend leaned in: “What are you talking about? It was right there in the question—yes or no, couldn't have been simpler, right?”
Before I could respond, even while a sick feeling was spreading in my stomach that had nothing to do with Bass, someone else added, “I was surprised. I expected a lot more from Jonakait, the guy wrote the book, for Christ's sake.”
The sick feeling intensified, but it wasn't for me.
“Um, guys, yeah, you go through them ‘A’ through ‘X’ and answer yes, no, and why—”
“Exactly!” someone interrupted.
“Then you go back and rearrange some of them, put one in front of another, and suddenly one piece lets another piece in, and then it's kinda an avalanche of evidence …right?”
There were a lot of comments.
Boiled down, they came to:
“Oh . . . shit.”
Tetris vs. Frogger
It must have taken Jonakait hours to craft the piece. The solution was intricate, almost mechanical. The pieces weren't independent. Moving one earlier in the sequence and it changed the shape of what came after. Doors that were closed opened. Evidence that was blocked became admissible. The whole board looked different when you stopped moving through it and started moving it.
The test looked like Frogger. It was really Tetris.
I wish I still had a copy of the test – if I did, I'd feed it into the big three AIs and see how they'd handle it. I have an idea, but I'll let Claude Pro answer for me:
CLAUDE'S ASSESSMENT
"Feed that fact pattern to any major AI system today. It will analyze items A through X, apply the rules of evidence correctly, cite the relevant standards, and produce a thorough, well-organized answer. It would have played excellent Frogger. . . . Tetris requires seeing that the pieces have relationships that the problem didn't announce. That's not retrievable. It has no prior documentation. There's no training signal for it. The model plays the game it's given because it cannot see that a different game is on the table."
Every Law Firm's AI Is at the Bar on Its Third Bud Light
In 2026, every law firm's AI is already sitting in the bar on their third pitcher of Bud Light (yeah, Bud Light), feeling good about their Frogger score. Clean analysis, correct rules, admissible or not, done early. The content looks fine. It looks like everyone else's content, because everyone else's AI played the same Frogger game on the same levels.
The piece that moves the reader is the one that creates, rearranges sequences, and opens doors no one else even knew were there. Someone playing Tetris and, preferably, drinking Smithwick's Ale.
We Play Tetris
Creative Legal Writing finds the angles AI doesn't know to look for. First-person content, unexpected connections, and arguments built from the inside out — for law firm partners who understand the difference between thorough and right.
Looking for Original Legal Content?
AI can help produce words. It can't replace judgment, storytelling, or the unexpected connections that make readers stop and think. At Forlorn Hope Publishing, we create first-person legal content, ghostwriting, and thought leadership designed to reflect the voice of your firm—not the average output of a language model. Contact us today and see how human writing can still stand apart.
Frequently Asked Questions
Modern AI tools can often identify legal issues, summarize rules, and organize information effectively. However, they may not recognize when a problem requires reframing or a different analytical approach.
The comparison illustrates the difference between following an obvious path (Frogger) and recognizing hidden relationships between pieces (Tetris). The article argues that strong legal writing often depends on the latter.
AI can assist with drafting and research, but persuasive legal writing often depends on original insights, judgment, sequencing, and storytelling that go beyond applying known patterns.
According to the article, memorable legal content doesn't simply answer questions correctly. It creates unexpected connections and presents ideas in ways that readers haven't already seen.
