Ted Underwood
@tedunderwood.com
Uses machine learning to study literary imagination, and vice-versa. Likely to share news about AI & computational social science / Sozialwissenschaft / 社会科学
Information Sciences and English, UIUC. Distant Horizons (Chicago, 2019). tedunderwood.com
Information Sciences and English, UIUC. Distant Horizons (Chicago, 2019). tedunderwood.com
Pinned
Ted Underwood
@tedunderwood.com
· Nov 4
Opinion | AI Is the Future. Higher Ed Should Shape It.
If we want to stay at the forefront of knowledge production, we must fit technology to our needs.
www.chronicle.com
Wrote a short piece arguing that higher ed must help steer AI. TLDR: If we outsource this to tech, we outsource our whole business. But rejectionism is basically stalling. If we want to survive, schools themselves must proactively shape AI for education & research. [1/6, unpaywalled at 5/6] +
If Christians quit the Colosseum, the lions win — so I'm staying
If women quit X because of Grok’s deepfake nudes, the predators win – so I’m staying
If women quit X because of Grok’s deepfake nudes, the predators win – so I’m staying
If women quit X because of Grok’s deepfake nudes, the predators win – so I’m staying
www.independent.co.uk
January 10, 2026 at 10:46 PM
If Christians quit the Colosseum, the lions win — so I'm staying
Reposted by Ted Underwood
I second this sentiment because I feel the buzz. It’s different. it’s a new type of thing. I further do not think this class of tool is limited to devs by any means, and you would do well to join the lunatics
Excitement around Claude Code reminds me of the buzz around GPT-3, which was (correctly!) intense — but limited to a tiny group of people who responded “wow, coherent babble; I see hundreds of ways to use that.” Broader audience ignored the lunatics until easy interface was invented 2022.
The unspoken point of my response to @tedunderwood.com was that the "we" that are all coders now is still limited to folks who for whatever reason don't need to be taught my (1) or (2) above, which comprise many but by no means a majority of folks.
January 10, 2026 at 9:59 PM
I second this sentiment because I feel the buzz. It’s different. it’s a new type of thing. I further do not think this class of tool is limited to devs by any means, and you would do well to join the lunatics
Reposted by Ted Underwood
a thing i've been thinking about posting for a long time is "back in those days, the workers didn't say "eww factories were created by capitalists to enrich capitalists, we should abandon factories" they said "we should seize them and direct their output for the good of humanity"
January 10, 2026 at 7:07 PM
a thing i've been thinking about posting for a long time is "back in those days, the workers didn't say "eww factories were created by capitalists to enrich capitalists, we should abandon factories" they said "we should seize them and direct their output for the good of humanity"
Reposted by Ted Underwood
I missed this last month from @tilmanbayer.bsky.social: "AI finds errors in 90% of October's Featured Articles". Great example of human-in-the-loop LLM use for verifying Wikipedia articles. en.wikipedia.org/wiki/Wikiped...
Wikipedia:Wikipedia Signpost/2025-12-01/Opinion - Wikipedia
en.wikipedia.org
January 10, 2026 at 6:25 PM
I missed this last month from @tilmanbayer.bsky.social: "AI finds errors in 90% of October's Featured Articles". Great example of human-in-the-loop LLM use for verifying Wikipedia articles. en.wikipedia.org/wiki/Wikiped...
Reposted by Ted Underwood
Winter fairytale forest.
January 10, 2026 at 11:31 AM
Winter fairytale forest.
Reposted by Ted Underwood
I've been using Claude Code for a lot more than just coding (thanks to @chrismdp.com) - a good guide for non-devs for how to get set up.
hannahstulberg.substack.com/p/claude-cod...
hannahstulberg.substack.com/p/claude-cod...
Claude Code for Everything: Finally, that Personal Assistant You’ve Always Wanted
Everything you need to get started (no coding required)
hannahstulberg.substack.com
January 10, 2026 at 12:30 PM
I've been using Claude Code for a lot more than just coding (thanks to @chrismdp.com) - a good guide for non-devs for how to get set up.
hannahstulberg.substack.com/p/claude-cod...
hannahstulberg.substack.com/p/claude-cod...
Excitement around Claude Code reminds me of the buzz around GPT-3, which was (correctly!) intense — but limited to a tiny group of people who responded “wow, coherent babble; I see hundreds of ways to use that.” Broader audience ignored the lunatics until easy interface was invented 2022.
The unspoken point of my response to @tedunderwood.com was that the "we" that are all coders now is still limited to folks who for whatever reason don't need to be taught my (1) or (2) above, which comprise many but by no means a majority of folks.
January 10, 2026 at 1:03 PM
Excitement around Claude Code reminds me of the buzz around GPT-3, which was (correctly!) intense — but limited to a tiny group of people who responded “wow, coherent babble; I see hundreds of ways to use that.” Broader audience ignored the lunatics until easy interface was invented 2022.
your computer is no longer a machine that runs apps built by other people
it’s now a Lego set you can instruct in natural language
it’s now a Lego set you can instruct in natural language
i think this is gonna be a big theme of 2026
claude code is misnamed. it's not really even about code. it's about making the computer do things you want it to do
claude code is misnamed. it's not really even about code. it's about making the computer do things you want it to do
January 10, 2026 at 12:46 AM
your computer is no longer a machine that runs apps built by other people
it’s now a Lego set you can instruct in natural language
it’s now a Lego set you can instruct in natural language
Reposted by Ted Underwood
Come work at UIUC with me - UIUC ECE has a faculty position on the Ethical, Legal, and Social Implications of Digital Transformation! ece.illinois.edu/about/jobs/f...
Faculty (Tenure Track)
Faculty (Tenure Track)
ece.illinois.edu
January 9, 2026 at 9:10 PM
Come work at UIUC with me - UIUC ECE has a faculty position on the Ethical, Legal, and Social Implications of Digital Transformation! ece.illinois.edu/about/jobs/f...
Reposted by Ted Underwood
Our special issue with @martinkusch.bsky.social is finally complete and here’s at long last our introduction. I pitched the idea to him in 2019! Thank you for your patience and generosity, Martin. I am proud of this manifesto and hope colleagues will join us in rethinking this old debate #philsci
January 9, 2026 at 9:28 AM
Our special issue with @martinkusch.bsky.social is finally complete and here’s at long last our introduction. I pitched the idea to him in 2019! Thank you for your patience and generosity, Martin. I am proud of this manifesto and hope colleagues will join us in rethinking this old debate #philsci
Another way to put this is that we are now all, in fact, coders — and you might want to explore your new abilities.
what I'm saying is that if you are *not* a coder it is probably getting to be about time you spent a couple minutes with Claude Code to see what it does.
January 10, 2026 at 12:10 AM
Another way to put this is that we are now all, in fact, coders — and you might want to explore your new abilities.
Reposted by Ted Underwood
We introduce epiplexity, a new measure of information that provides a foundation for how to select, generate, or transform data for learning systems. We have been working on this for almost 2 years, and I cannot contain my excitement! arxiv.org/abs/2601.03220 1/7
January 7, 2026 at 5:28 PM
We introduce epiplexity, a new measure of information that provides a foundation for how to select, generate, or transform data for learning systems. We have been working on this for almost 2 years, and I cannot contain my excitement! arxiv.org/abs/2601.03220 1/7
Reposted by Ted Underwood
so so so excited to present our research + connect with the #ATScience community 🧪🎉
Excited to announce that @sjgreenwood.bsky.social will be at #ATScience presenting their work on the beloved @paper-feed.bsky.social , experiments on self-hosted feeds (in collaboration with @graze.social @aendra.com) and observational analyses of social media on #atproto!
Looking forward to it! 🎉
Looking forward to it! 🎉
Exciting to see our first speaker proposals coming in for #ATScience 2026! ✨
We’d love to hear yours too, so please submit your idea at
forms.atproto.science/atscience26-... 🗳️
The call for proposals closes at the end of January - we’ll review and notify speakers on a rolling basis.
We’d love to hear yours too, so please submit your idea at
forms.atproto.science/atscience26-... 🗳️
The call for proposals closes at the end of January - we’ll review and notify speakers on a rolling basis.
January 9, 2026 at 9:37 PM
so so so excited to present our research + connect with the #ATScience community 🧪🎉
Reposted by Ted Underwood
The recent days have been horrific. We can't become numb to repeated instances of illegal and unconstitutional action by government agencies. It's even worse when public officials are blatantly lying in ways that contradict dozens of pieces of video evidence.
Sen. Lankford blatantly lies about what the video of Renee Good's killing shows: "A classic law enforcement moment -- they have to fire their weapon and then when you see her car crash, law enforcement is running to her to provide aid. They're never looking to be able to take a life of individuals."
January 9, 2026 at 9:40 PM
The recent days have been horrific. We can't become numb to repeated instances of illegal and unconstitutional action by government agencies. It's even worse when public officials are blatantly lying in ways that contradict dozens of pieces of video evidence.
Reposted by Ted Underwood
For CRITICAL INQUIRY,
I reviewed @leifw.bsky.social’s very important LANGUAGE MACHINES. Also, unless someone tells me different, I’m going to lay claim to the first F bomb in CI’s history.
I reviewed @leifw.bsky.social’s very important LANGUAGE MACHINES. Also, unless someone tells me different, I’m going to lay claim to the first F bomb in CI’s history.
"Cultural AI is called out in the subtitle not as a plea to attend to AI’s better half but because culture has been detached from cognition for too long."
New in review, Matthew Kirschenbaum on Leif Weatherby's Language Machines: criticalinquiry.uchicago.edu/matthew_kirs...
New in review, Matthew Kirschenbaum on Leif Weatherby's Language Machines: criticalinquiry.uchicago.edu/matthew_kirs...
January 9, 2026 at 1:55 AM
For CRITICAL INQUIRY,
I reviewed @leifw.bsky.social’s very important LANGUAGE MACHINES. Also, unless someone tells me different, I’m going to lay claim to the first F bomb in CI’s history.
I reviewed @leifw.bsky.social’s very important LANGUAGE MACHINES. Also, unless someone tells me different, I’m going to lay claim to the first F bomb in CI’s history.
Gemini integration is not there yet, but it's going to be possible to have a Claude-Code-level agentic assistant in your email. Systematically cross-referencing inquiries with replies, asking clarification questions, remembering preferences, generating to-do lists. Absolutely fatal app.
January 9, 2026 at 6:34 PM
Gemini integration is not there yet, but it's going to be possible to have a Claude-Code-level agentic assistant in your email. Systematically cross-referencing inquiries with replies, asking clarification questions, remembering preferences, generating to-do lists. Absolutely fatal app.
Reposted by Ted Underwood
this is so cool! language models use a ring-like manifold to keep track of when they need to insert a newline, just like a brain would.
arxiv.org/abs/2601.04480
arxiv.org/abs/2601.04480
January 9, 2026 at 4:43 PM
this is so cool! language models use a ring-like manifold to keep track of when they need to insert a newline, just like a brain would.
arxiv.org/abs/2601.04480
arxiv.org/abs/2601.04480
Reposted by Ted Underwood
Nice bit of digital humanities here - using AI to transcribe more than 32,000 manuscripts in the space of a few months. But with two years of preparation in training the model and creating standards for automating manuscript transcription
www.inria.fr/en/comma-med...
www.inria.fr/en/comma-med...
CoMMA: thousands of medieval manuscripts finally transcribed
Transcribing thousands of medieval manuscripts by hand would be a monumental undertaking. Fortunately, researchers in computational humanities at the Inria Paris Centre have been able to automate the ...
www.inria.fr
January 9, 2026 at 12:48 PM
Nice bit of digital humanities here - using AI to transcribe more than 32,000 manuscripts in the space of a few months. But with two years of preparation in training the model and creating standards for automating manuscript transcription
www.inria.fr/en/comma-med...
www.inria.fr/en/comma-med...
Reposted by Ted Underwood
Consider signing this petition regarding restoring free access to the ACM digital library: www.ipetitions.com/petition/res...
Petition Restore Fully Free and Open Access to the ACM Digital Library
Restore Fully Free and Open Access to the ACM Digital Library
www.ipetitions.com
January 9, 2026 at 8:08 AM
Consider signing this petition regarding restoring free access to the ACM digital library: www.ipetitions.com/petition/res...
Reposted by Ted Underwood
i always nuke my git histories for code repos that i post for papers because my entire commit history is "update" "." "fix" "." "trying again" "pls work" "pls" "k"
I get that for professional devs, vibe coding may feel sloppy. But as an academic who writes research code, I would not say that the specifications I'm getting from Claude Code, complete with tables of edge cases and testing plans, are um <cough> *less* rigorous than my past practice.
January 9, 2026 at 1:42 AM
i always nuke my git histories for code repos that i post for papers because my entire commit history is "update" "." "fix" "." "trying again" "pls work" "pls" "k"
Reposted by Ted Underwood
sysadmin hits different now
January 9, 2026 at 12:36 AM
sysadmin hits different now
Reposted by Ted Underwood
1/ We found that deep sequence models memorize atomic facts "geometrically" -- not as an associative lookup table as often imagined.
This opens up practical questions on reasoning/memory/discovery, and also poses a theoretical "memorization puzzle."
This opens up practical questions on reasoning/memory/discovery, and also poses a theoretical "memorization puzzle."
January 8, 2026 at 8:31 PM
1/ We found that deep sequence models memorize atomic facts "geometrically" -- not as an associative lookup table as often imagined.
This opens up practical questions on reasoning/memory/discovery, and also poses a theoretical "memorization puzzle."
This opens up practical questions on reasoning/memory/discovery, and also poses a theoretical "memorization puzzle."
I get that for professional devs, vibe coding may feel sloppy. But as an academic who writes research code, I would not say that the specifications I'm getting from Claude Code, complete with tables of edge cases and testing plans, are um <cough> *less* rigorous than my past practice.
January 9, 2026 at 12:11 AM
I get that for professional devs, vibe coding may feel sloppy. But as an academic who writes research code, I would not say that the specifications I'm getting from Claude Code, complete with tables of edge cases and testing plans, are um <cough> *less* rigorous than my past practice.
controversy over the image-edit button does not seem to be fading quickly
January 8, 2026 at 11:10 PM
controversy over the image-edit button does not seem to be fading quickly
Reposted by Ted Underwood
Legal falsity aside, I think “our masked, anonymous paramilitary thugs have an absolute right to kill you and your job is simply to trust that they won’t” isn’t so much a defense as a confession
January 8, 2026 at 9:30 PM
Legal falsity aside, I think “our masked, anonymous paramilitary thugs have an absolute right to kill you and your job is simply to trust that they won’t” isn’t so much a defense as a confession