ZP Rosen
banner
zaqdelinguist.bsky.social
ZP Rosen
@zaqdelinguist.bsky.social
Computational sociolinguist interested in how groups negotiate meaning and social coordination. Because I’ve been working with transformers since 2018, very not into AI hype.

https://zaqari.github.io
That's not what it is and it's never going to be that. It's just AI company marketing copy now, and ground zero for the dead/shambling generative zombie internet. But it could have been a real force for labor equity if things had gone quite differently, and that's an interesting thought.
November 28, 2025 at 10:53 PM
It's funny. LinkedIn though could have been something really cool as far as social media goes. It's a website that centers on employment, and thus to some degree labor. There is a world in which LI could have been a way for social networking as labor praxis.
November 28, 2025 at 10:53 PM
Not a “step too far” but a “hard pass”. I’ve got too much work to do to be distracted by AI outputs. I do not need naviGPT bothering me when I have stuff to do.
a picture of a blue ball with wings and the words `` listen ! ''
ALT: a picture of a blue ball with wings and the words `` listen ! ''
media.tenor.com
November 23, 2025 at 11:51 PM
WOOF. That’s a brutal tagline. Adding this to my queue too.
November 20, 2025 at 3:58 PM
It’s frustrating that the rhetorical move isn’t to just say “well say something worth listening to/platforming.”
November 20, 2025 at 3:17 AM
Self servingly, I would love to read a whole article or commentary about that. Because I struggle with teaching students about voice too, and I really, really want to hear other folks’ stories and takes on it in a post LLM world.
November 19, 2025 at 2:51 PM
Thank you! Dr. Dale and I redid that one with more groups and a whole lot more data and added some commentary too on information and model sycophancy :). we’re working on sending the big updated version out for review.
November 19, 2025 at 2:09 PM
Ooh! Ooh! I don’t know how to treat authorial voice, but I can say that I did a whole lesson on “your voice is better than AI’s and here’s proof” with my students last year and we launched a website + stats visualizations showing that!!

Can share the LP and associated code if you ever want? :)
zaqari.github.io
November 18, 2025 at 10:11 PM
Because less students going tappy tap make number go up for Sam Altman.
November 11, 2025 at 4:01 AM
Even if the paper ends up being rejected, this means we're going to get some feedback from our peers, and every review is an opportunity to make this paper that much more bulletproof. #phdlife #academia
November 6, 2025 at 3:53 PM
No problems! I was in grad school for computational linguistics when the first transformers (BERT and then GPT-1 and 2) came out so this whole era in the field of language modeling is what I came of age in :).
November 5, 2025 at 7:34 PM
I actually compiled a whole list of papers on the history of how we got to LLMs as a model architecture as a prelude to eventually turning it into a class! Here’s my list:

zaqari.github.io/llm-history....
A Seminar/Reading Plan on the History of Language Modeling
zaqari.github.io
November 5, 2025 at 7:27 PM
Prop 50 was made using the Legend of Zelda: BOTW weapon system. It's designed to break after a set period of time so that the user replaces it with another tool that's more relevant to the current context/playthrough.
November 4, 2025 at 5:50 PM