There you have it! AI just beat law professors at their own job. Not by a little. By a landslide.
June 3, 2026 · 0 likes · 0 comments
Workforce AI
There you have it! AI just beat law professors at their own job. Not by a little. By a landslide.
Stanford law professor Julian Nyarko ran the experiment the right way. 16 contracts professors from 14 U.S. law schools. They wrote 40 exam and office-hours questions. They wrote their own answers. Then they blind-judged 2,918 head-to-head matchups between human answers and AI answers.
The AI won 75.33% of the time.
And here is the part that should stop you cold.
The professors flagged AI answers as potentially harmful only 3.53% of the time. They flagged their own colleagues' answers as harmful 12.06% of the time.
When they didn't know who wrote it, they trusted the machine more than they trusted each other.
This wasn't a chatbot hallucinating fake cases. It was Gemini 2.5 Pro and a NotebookLM trained on the actual casebook these professors teach from. Purpose-built. In the exact domain being tested.
Now do the math on the credential. Top law school tuition is $70,000 a year. $200,000 over three years. A huge chunk of that pays for access to faculty expertise — in exactly the instructional context where AI already wins three out of four times. And that wasn't even a great LLM choice!
If you have a kid in law school right now, you should be asking what, precisely, you are paying for.
Don't take my word for it. Take theirs.
On June 2, the world's mathematicians issued a formal warning — the Leiden Declaration, endorsed by the International Mathematical Union, the body that hands out the Fields Medal. These are not people afraid of calculators. They build the verification systems that would make AI math trustworthy in the first place.
Their warning: AI produces "plausible but unreliable" proofs that are cheap to generate and hard to tell apart from the real thing. Bury one error in a widely-cited result and it doesn't stay buried — it metastasizes through the literature for decades. Cryptography, finance, engineering, medicine all sit on top of that foundation.
So read both signals together.
One field is being replaced from the bottom up. The other is being polluted from the top down. Same technology. Same year.
I have been saying this everywhere I go: the displacement doesn't announce itself with robots marching through the door. It shows up as a 75% win rate in a blind study, and a tuition bill that hasn't moved yet.
The steam drill always wins eventually. The only question left is what we choose to build after it does.
My book, REPLACEMENT, coming July 28, will help you!
Full breakdown on UnbiasedHeadlines.com:
https://lnkd.in/eE4ZqfED
What are your thoughts?
Stanford law professor Julian Nyarko ran the experiment the right way. 16 contracts professors from 14 U.S. law schools. They wrote 40 exam and office-hours questions. They wrote their own answers. Then they blind-judged 2,918 head-to-head matchups between human answers and AI answers.
The AI won 75.33% of the time.
And here is the part that should stop you cold.
The professors flagged AI answers as potentially harmful only 3.53% of the time. They flagged their own colleagues' answers as harmful 12.06% of the time.
When they didn't know who wrote it, they trusted the machine more than they trusted each other.
This wasn't a chatbot hallucinating fake cases. It was Gemini 2.5 Pro and a NotebookLM trained on the actual casebook these professors teach from. Purpose-built. In the exact domain being tested.
Now do the math on the credential. Top law school tuition is $70,000 a year. $200,000 over three years. A huge chunk of that pays for access to faculty expertise — in exactly the instructional context where AI already wins three out of four times. And that wasn't even a great LLM choice!
If you have a kid in law school right now, you should be asking what, precisely, you are paying for.
Don't take my word for it. Take theirs.
On June 2, the world's mathematicians issued a formal warning — the Leiden Declaration, endorsed by the International Mathematical Union, the body that hands out the Fields Medal. These are not people afraid of calculators. They build the verification systems that would make AI math trustworthy in the first place.
Their warning: AI produces "plausible but unreliable" proofs that are cheap to generate and hard to tell apart from the real thing. Bury one error in a widely-cited result and it doesn't stay buried — it metastasizes through the literature for decades. Cryptography, finance, engineering, medicine all sit on top of that foundation.
So read both signals together.
One field is being replaced from the bottom up. The other is being polluted from the top down. Same technology. Same year.
I have been saying this everywhere I go: the displacement doesn't announce itself with robots marching through the door. It shows up as a 75% win rate in a blind study, and a tuition bill that hasn't moved yet.
The steam drill always wins eventually. The only question left is what we choose to build after it does.
My book, REPLACEMENT, coming July 28, will help you!
Full breakdown on UnbiasedHeadlines.com:
https://lnkd.in/eE4ZqfED
What are your thoughts?