Jill Smith
@jaslawlib.bsky.social
Head of Reference, Georgetown University Law Center Library. Opinions my own.
Welp, you know, my Georgetown colleagues all know about my sideline, so it's not like I'm being coy.
It's just more that I spend time on the author/publisher side of social media.
It's just more that I spend time on the author/publisher side of social media.
November 18, 2025 at 11:05 PM
Welp, you know, my Georgetown colleagues all know about my sideline, so it's not like I'm being coy.
It's just more that I spend time on the author/publisher side of social media.
It's just more that I spend time on the author/publisher side of social media.
And I'll re-post from this account which is lower profile but very much more DC-centric!!
November 18, 2025 at 10:58 PM
And I'll re-post from this account which is lower profile but very much more DC-centric!!
And there are tons of great RAs! But we're often the ones training them, sometimes helping to supervise them, and (yes) sometimes bailing them out when they don't get it.
"Vocational awe" is not something I love, but a sincere thank-you is not the same thing and is much appreciated.
"Vocational awe" is not something I love, but a sincere thank-you is not the same thing and is much appreciated.
November 18, 2025 at 9:39 PM
And there are tons of great RAs! But we're often the ones training them, sometimes helping to supervise them, and (yes) sometimes bailing them out when they don't get it.
"Vocational awe" is not something I love, but a sincere thank-you is not the same thing and is much appreciated.
"Vocational awe" is not something I love, but a sincere thank-you is not the same thing and is much appreciated.
They weren't so used to the whole "for exposure" con, and were shocked.
November 13, 2025 at 2:35 PM
They weren't so used to the whole "for exposure" con, and were shocked.
I had no interest in working with such a transparently unethical outfit. I also let my team know what had happened and said, "Obviously, you can do as you like, but here's the rate I was paid >10 years ago to create and present one of these and it wasn't recorded."
November 13, 2025 at 2:35 PM
I had no interest in working with such a transparently unethical outfit. I also let my team know what had happened and said, "Obviously, you can do as you like, but here's the rate I was paid >10 years ago to create and present one of these and it wasn't recorded."
So, Westlaw can now do ad hoc 50-state surveys using their AI. I used the weirdest thing I learned when studying for the NH bar 30 years ago to test it.
tl;dr I wouldn't trust it for a final answer (not that I would with any AI, but...)
tl;dr I wouldn't trust it for a final answer (not that I would with any AI, but...)
October 6, 2025 at 6:32 PM
I've started keeping a file on CoCounsel's errors and hallucinations. We don't use it a lot (because we already have the skills to do more efficiently in most cases what it's supposed to do for you), but it's still worrying how much I've snagged in a short time.
October 6, 2025 at 6:26 PM
I've started keeping a file on CoCounsel's errors and hallucinations. We don't use it a lot (because we already have the skills to do more efficiently in most cases what it's supposed to do for you), but it's still worrying how much I've snagged in a short time.
I also did this: bsky.app/profile/jasl...
So, Westlaw can now do ad hoc 50-state surveys using their AI. I used the weirdest thing I learned when studying for the NH bar 30 years ago to test it.
tl;dr I wouldn't trust it for a final answer (not that I would with any AI, but...)
tl;dr I wouldn't trust it for a final answer (not that I would with any AI, but...)
July 25, 2025 at 1:09 PM
I also did this: bsky.app/profile/jasl...
And in the "transcript" of the interaction, the AI reworded the original prompt!
July 25, 2025 at 1:08 PM
And in the "transcript" of the interaction, the AI reworded the original prompt!
One of my colleagues tested it by asking CoCounsel what a law review article had to say about jury instructions. It produced a lengthy paragraph and helpfully highlighted the original.
Which had nothing at all to say about jury instructions. The entire paper was about victim impact evidence.
Which had nothing at all to say about jury instructions. The entire paper was about victim impact evidence.
July 25, 2025 at 12:55 PM
One of my colleagues tested it by asking CoCounsel what a law review article had to say about jury instructions. It produced a lengthy paragraph and helpfully highlighted the original.
Which had nothing at all to say about jury instructions. The entire paper was about victim impact evidence.
Which had nothing at all to say about jury instructions. The entire paper was about victim impact evidence.
And that quote that doesn't exist in the original is exactly the kind of thing we've seen Westlaw's CoCounsel do.
July 25, 2025 at 12:50 PM
And that quote that doesn't exist in the original is exactly the kind of thing we've seen Westlaw's CoCounsel do.
Please point to a legal citation that says you can't "mention the color of their skin or their faith or their gender or anything else when you run against them"
May 23, 2025 at 8:19 PM
Please point to a legal citation that says you can't "mention the color of their skin or their faith or their gender or anything else when you run against them"
This part of the inquiry especially made me want to walk into the sea:
"including using GenAI to teach, grade, interact with students, draft correspondence"
What.
WHAT.
No, I don't use a Furby to teach, grade, interact with students, draft correspondence... Next question???
"including using GenAI to teach, grade, interact with students, draft correspondence"
What.
WHAT.
No, I don't use a Furby to teach, grade, interact with students, draft correspondence... Next question???
April 29, 2025 at 5:43 PM
And even the stuff that is specifically created for that purpose, with a limited dataset gets stuff wrong. So you have to check it, which means doing the work yourself. How exactly does this save time??
bsky.app/profile/jasl...
bsky.app/profile/jasl...
So, Westlaw can now do ad hoc 50-state surveys using their AI. I used the weirdest thing I learned when studying for the NH bar 30 years ago to test it.
tl;dr I wouldn't trust it for a final answer (not that I would with any AI, but...)
tl;dr I wouldn't trust it for a final answer (not that I would with any AI, but...)
April 29, 2025 at 5:42 PM
And even the stuff that is specifically created for that purpose, with a limited dataset gets stuff wrong. So you have to check it, which means doing the work yourself. How exactly does this save time??
bsky.app/profile/jasl...
bsky.app/profile/jasl...