YY Ahn
@yyahn.bsky.social
Prof @ School of Data Science, University of Virginia. Formerly at IU. Networks, data science, and machine learning. https://yyahn.com
🚲🚲🚲
🚲🚲🚲
Reposted by YY Ahn
Our latest paper in @science.org warns about malicious AI swarms, agents capable of adaptive influence campaigns at scale. We already observed some in the wild (picture). AI is a real threat to democracy.
#SciencePolicyForum #ScienceResearch 🧪
Paper: doi.org/10.1126/scie...
#SciencePolicyForum #ScienceResearch 🧪
Paper: doi.org/10.1126/scie...
January 26, 2026 at 1:09 AM
Our latest paper in @science.org warns about malicious AI swarms, agents capable of adaptive influence campaigns at scale. We already observed some in the wild (picture). AI is a real threat to democracy.
#SciencePolicyForum #ScienceResearch 🧪
Paper: doi.org/10.1126/scie...
#SciencePolicyForum #ScienceResearch 🧪
Paper: doi.org/10.1126/scie...
Fire depts can have a surprisingly big negative impact on street safety. Cville folks, help shaping the strategic plan of the city fire department!
It'll be Feb 2nd from 6-8pm at Carver Rec Center 👇
youtu.be/j2dHFC31VtQ?...
It'll be Feb 2nd from 6-8pm at Carver Rec Center 👇
youtu.be/j2dHFC31VtQ?...
January 26, 2026 at 4:41 PM
Fire depts can have a surprisingly big negative impact on street safety. Cville folks, help shaping the strategic plan of the city fire department!
It'll be Feb 2nd from 6-8pm at Carver Rec Center 👇
youtu.be/j2dHFC31VtQ?...
It'll be Feb 2nd from 6-8pm at Carver Rec Center 👇
youtu.be/j2dHFC31VtQ?...
It was a "historic" and "unprecedented" level of investment into manufacturing capacity in the US.
January 26, 2026 at 2:51 PM
It was a "historic" and "unprecedented" level of investment into manufacturing capacity in the US.
Reposted by YY Ahn
Announcing the release of the official Common Corpus paper: a 20 page report detailing how we collected, processed and published 2 trillion tokens of reusable data for LLM pretraining arxiv.org/pdf/2506.01732
June 4, 2025 at 1:30 PM
Announcing the release of the official Common Corpus paper: a 20 page report detailing how we collected, processed and published 2 trillion tokens of reusable data for LLM pretraining arxiv.org/pdf/2506.01732
This can be an important reason why many people try out but don't "stick". Starter-pack is great, but it's inherently towards a very small set of the most visible users, not the niches that individual users are drawn to.
This may be a critical feature as a social platform. @jay.bsky.team
This may be a critical feature as a social platform. @jay.bsky.team
The replies to this post surfaced lots of great 3rd party tools (thanks!), but I still think it's interesting that bsky, as a platform, has decided to not prioritize algorithmic triangle closing (or follow recommendations, generally). See this fig from Zignani et al. (2014), when FB introduced PYMK.
January 25, 2026 at 3:26 PM
This can be an important reason why many people try out but don't "stick". Starter-pack is great, but it's inherently towards a very small set of the most visible users, not the niches that individual users are drawn to.
This may be a critical feature as a social platform. @jay.bsky.team
This may be a critical feature as a social platform. @jay.bsky.team
Reposted by YY Ahn
BREAKING: The Tribune confirmed on Instagram that the man shot has died, according to two sources with knowledge of the incident.
ICE attempted to order local police from the scene, but Chief Brian O'Hara refused, sources also told the Tribune. O'Hara instructed his officers to preserve the scene.
ICE attempted to order local police from the scene, but Chief Brian O'Hara refused, sources also told the Tribune. O'Hara instructed his officers to preserve the scene.
Man shot by federal agent in south Minneapolis www.startribune.com/ice-raids-mi...
January 24, 2026 at 4:36 PM
BREAKING: The Tribune confirmed on Instagram that the man shot has died, according to two sources with knowledge of the incident.
ICE attempted to order local police from the scene, but Chief Brian O'Hara refused, sources also told the Tribune. O'Hara instructed his officers to preserve the scene.
ICE attempted to order local police from the scene, but Chief Brian O'Hara refused, sources also told the Tribune. O'Hara instructed his officers to preserve the scene.
Reposted by YY Ahn
TikTok is now under US ownership and their new ToS reads like a guide to giving fascists everything they need to put you on a list.
They’re collecting information about immigration status, citizenship, sexual orientation, whether you’re trans or non binary and physical or mental disabilities.
They’re collecting information about immigration status, citizenship, sexual orientation, whether you’re trans or non binary and physical or mental disabilities.
January 23, 2026 at 9:11 PM
TikTok is now under US ownership and their new ToS reads like a guide to giving fascists everything they need to put you on a list.
They’re collecting information about immigration status, citizenship, sexual orientation, whether you’re trans or non binary and physical or mental disabilities.
They’re collecting information about immigration status, citizenship, sexual orientation, whether you’re trans or non binary and physical or mental disabilities.
Reposted by YY Ahn
"What if AI wasn't just hallucinating information, but thousands of AI chatbots were working together to give the guise of grassroots support where there was none? That's the future this paper imagines—Russian troll farms on steroids,” @ninajankowicz.com
www.wired.com/story/ai-pow...
www.wired.com/story/ai-pow...
AI-Powered Disinformation Swarms Are Coming for Democracy
Advances in artificial intelligence are creating a perfect storm for those seeking to spread disinformation at unprecedented speed and scale. And it’s virtually impossible to detect.
www.wired.com
January 23, 2026 at 10:04 AM
"What if AI wasn't just hallucinating information, but thousands of AI chatbots were working together to give the guise of grassroots support where there was none? That's the future this paper imagines—Russian troll farms on steroids,” @ninajankowicz.com
www.wired.com/story/ai-pow...
www.wired.com/story/ai-pow...
🤯
The nerdiest, and thus best, way of getting a bibitex entry from a DOI is by querying it directly via the command
curl -LH "Accept: text/bibliography; style=bibtex" doi.org/10.1016/j.ch...
The second best is just to use @zotero.org for everything.
curl -LH "Accept: text/bibliography; style=bibtex" doi.org/10.1016/j.ch...
The second best is just to use @zotero.org for everything.
Just checking, since apparently some are using LLMs to get BibTeX snippets easily.
For Computer Science and ML papers, you know about @dblp.org, right? You know you don't have to use LLMs for that... right?
For Computer Science and ML papers, you know about @dblp.org, right? You know you don't have to use LLMs for that... right?
January 22, 2026 at 5:58 PM
🤯
Reposted by YY Ahn
In our conservative estimate, 41 percent of Grok’s images — or 1.8 million — were sexualized depictions of women. Separately, CCDH analyzed how many were sexualized images across genders and ages. Their findings: an estimated 3 million images, including more than 23,000 of children.
January 22, 2026 at 3:01 PM
In our conservative estimate, 41 percent of Grok’s images — or 1.8 million — were sexualized depictions of women. Separately, CCDH analyzed how many were sexualized images across genders and ages. Their findings: an estimated 3 million images, including more than 23,000 of children.
Reposted by YY Ahn
UVA Data Science Expands MSDS Residential Program with 4 New Concentrations: (1) *Genomics*, (2) AI, (3) Sports data science, (4) Visualization datascience.virginia.edu/news/uva-dat... 🧬💻🧪
UVA Data Science Expands MSDS Residential Program with Four New Concentrations — School of Data Science
UVA Data Science announces four new MSDS concentrations — AI, Data Viz, Genomics and Sports Data Science — unlocking focused pathways to high-impact careers.
datascience.virginia.edu
January 21, 2026 at 7:45 PM
UVA Data Science Expands MSDS Residential Program with 4 New Concentrations: (1) *Genomics*, (2) AI, (3) Sports data science, (4) Visualization datascience.virginia.edu/news/uva-dat... 🧬💻🧪
Reposted by YY Ahn
We have over 65 people registered so far for our webinar tomorrow. It’s going to be a great conversation!
There is still time to sign up: events.humanitix.com/belonging-in...
There is still time to sign up: events.humanitix.com/belonging-in...
January 20, 2026 at 11:56 AM
We have over 65 people registered so far for our webinar tomorrow. It’s going to be a great conversation!
There is still time to sign up: events.humanitix.com/belonging-in...
There is still time to sign up: events.humanitix.com/belonging-in...
Reposted by YY Ahn
Happy Martin Luther King Day my friends: "I refuse to accept the view that mankind is so tragically bound to the starless midnight of racism and war."
Speech given on December 11, 1964 after being awarded the Nobel Peace Prize.
Speech given on December 11, 1964 after being awarded the Nobel Peace Prize.
January 19, 2026 at 6:14 PM
Happy Martin Luther King Day my friends: "I refuse to accept the view that mankind is so tragically bound to the starless midnight of racism and war."
Speech given on December 11, 1964 after being awarded the Nobel Peace Prize.
Speech given on December 11, 1964 after being awarded the Nobel Peace Prize.
Made a tiny CLI helper that turns natural language into shell commands using Claude Code.
howto find python files over 1mb → fd -e py -S +1m
I just put this into my dotfiles. No API keys. This simple script lets me stick to Ghostty instead of Warp.
yyahn.com/wiki/Software/howto/
howto find python files over 1mb → fd -e py -S +1m
I just put this into my dotfiles. No API keys. This simple script lets me stick to Ghostty instead of Warp.
yyahn.com/wiki/Software/howto/
howto
As I use more, my workflow also becomes more commandline-centric. As you may know, a small annoyance that tags along the power of is that it is hard to...
yyahn.com
January 18, 2026 at 8:44 PM
Made a tiny CLI helper that turns natural language into shell commands using Claude Code.
howto find python files over 1mb → fd -e py -S +1m
I just put this into my dotfiles. No API keys. This simple script lets me stick to Ghostty instead of Warp.
yyahn.com/wiki/Software/howto/
howto find python files over 1mb → fd -e py -S +1m
I just put this into my dotfiles. No API keys. This simple script lets me stick to Ghostty instead of Warp.
yyahn.com/wiki/Software/howto/
A minimal implementation of Ralph Wiggum loop: github.com/yy/wiggum
uv tool install wiggum
wiggum init
wiggum run
try it for low-stake projects. ;)
uv tool install wiggum
wiggum init
wiggum run
try it for low-stake projects. ;)
GitHub - yy/wiggum: Ralph Wiggum loop for agents
Ralph Wiggum loop for agents. Contribute to yy/wiggum development by creating an account on GitHub.
github.com
January 16, 2026 at 10:19 PM
A minimal implementation of Ralph Wiggum loop: github.com/yy/wiggum
uv tool install wiggum
wiggum init
wiggum run
try it for low-stake projects. ;)
uv tool install wiggum
wiggum init
wiggum run
try it for low-stake projects. ;)
Reposted by YY Ahn
This is a watershed moment for Texas - for the first time ever, solar supplied more electricity than coal in 2025.
I will say it again: At this rate, TX will likely decarbonize its electricity sector faster than CA.
TX clean electricity is growing at 13%,m more than twice that of CA's at 6%.
I will say it again: At this rate, TX will likely decarbonize its electricity sector faster than CA.
TX clean electricity is growing at 13%,m more than twice that of CA's at 6%.
Texas solar power surpassed coal in 2025 for the first year ever. Here's why.
In 2025 — for the first year ever — solar arrays provided more electricity to Texas’ main power grid than coal-fired power plants.
www.houstonchronicle.com
January 15, 2026 at 1:54 PM
This is a watershed moment for Texas - for the first time ever, solar supplied more electricity than coal in 2025.
I will say it again: At this rate, TX will likely decarbonize its electricity sector faster than CA.
TX clean electricity is growing at 13%,m more than twice that of CA's at 6%.
I will say it again: At this rate, TX will likely decarbonize its electricity sector faster than CA.
TX clean electricity is growing at 13%,m more than twice that of CA's at 6%.
Reposted by YY Ahn
📣 Parents. If cars make your walk or cycle to school feel unsafe, that's not okay.
We need your photos of the school run! The more people can see how awful it is out there - and how unsafe it is for our children - the more power we have. DM pix or tag us in a post.
#streetsforkids #schoolrun
We need your photos of the school run! The more people can see how awful it is out there - and how unsafe it is for our children - the more power we have. DM pix or tag us in a post.
#streetsforkids #schoolrun
January 15, 2026 at 6:21 AM
📣 Parents. If cars make your walk or cycle to school feel unsafe, that's not okay.
We need your photos of the school run! The more people can see how awful it is out there - and how unsafe it is for our children - the more power we have. DM pix or tag us in a post.
#streetsforkids #schoolrun
We need your photos of the school run! The more people can see how awful it is out there - and how unsafe it is for our children - the more power we have. DM pix or tag us in a post.
#streetsforkids #schoolrun
Reposted by YY Ahn
America's global pre-eminence in research was partly driven by a great power competition. At a time when China is investing more in research, and relaxing visas to attract academic talent, the Trump administration is doing the opposite.
Today in Stupid Data but Right General Conclusion
Chinese Universities Surge in Global Rankings as U.S. Schools Slip
www.nytimes.com
January 15, 2026 at 12:31 PM
America's global pre-eminence in research was partly driven by a great power competition. At a time when China is investing more in research, and relaxing visas to attract academic talent, the Trump administration is doing the opposite.
Reposted by YY Ahn
Latest on the federal forces in Minnesota from me and my colleague Jeff Hargarten:
There are now more federal immigration agents in MN than the 10 largest Twin Cities police departments combined.
9% of ICE is in MN right now, even though MN has less than 1% of the nation's undocumented immigrants.
There are now more federal immigration agents in MN than the 10 largest Twin Cities police departments combined.
9% of ICE is in MN right now, even though MN has less than 1% of the nation's undocumented immigrants.
Are 3,000 ICE agents ‘disproportionate’ to Twin Cities police forces?
Their presence is bigger than the 10 largest metro police departments combined.
www.startribune.com
January 13, 2026 at 10:55 PM
Latest on the federal forces in Minnesota from me and my colleague Jeff Hargarten:
There are now more federal immigration agents in MN than the 10 largest Twin Cities police departments combined.
9% of ICE is in MN right now, even though MN has less than 1% of the nation's undocumented immigrants.
There are now more federal immigration agents in MN than the 10 largest Twin Cities police departments combined.
9% of ICE is in MN right now, even though MN has less than 1% of the nation's undocumented immigrants.
Reposted by YY Ahn
New preprint on unifying the zoo of multivariate higher-order information measures into a common form.
May be of interest to anyone interested in higher-order interactions, complex systems, emergence, or complexity.
1/N
arxiv.org/abs/2601.08030
May be of interest to anyone interested in higher-order interactions, complex systems, emergence, or complexity.
1/N
arxiv.org/abs/2601.08030
The many faces of multivariate information
Extracting higher-order structures from multivariate data has become an area of intensive study in complex systems science, as these multipartite interactions can reveal insights into fundamental feat...
arxiv.org
January 14, 2026 at 3:43 PM
New preprint on unifying the zoo of multivariate higher-order information measures into a common form.
May be of interest to anyone interested in higher-order interactions, complex systems, emergence, or complexity.
1/N
arxiv.org/abs/2601.08030
May be of interest to anyone interested in higher-order interactions, complex systems, emergence, or complexity.
1/N
arxiv.org/abs/2601.08030
One car repair buys the ebike. One tank of gas covers a year of maintenance. One coffee fuels a whole year of it.
I ran the numbers and the electricity costs for my cargo ebike are about $0.01/mile. The whole bike cost less than the annual maintenance of our car, I’ve put 675 miles on it this year
January 14, 2026 at 1:49 PM
One car repair buys the ebike. One tank of gas covers a year of maintenance. One coffee fuels a whole year of it.
Reposted by YY Ahn
We are hiring a post-doc to study the impact of AI agents on complex social systems in Duke's Society-Centered AI Initiative and/or the Polarization Lab! Apply here:
academicjobsonline.org/ajo/jobs/314...
academicjobsonline.org/ajo/jobs/314...
January 13, 2026 at 2:01 PM
We are hiring a post-doc to study the impact of AI agents on complex social systems in Duke's Society-Centered AI Initiative and/or the Polarization Lab! Apply here:
academicjobsonline.org/ajo/jobs/314...
academicjobsonline.org/ajo/jobs/314...
Reposted by YY Ahn
BREAKING | South Korea’s ex-president Yoon facing possible death penalty over insurrection
South Korea’s ex-president Yoon facing possible death penalty over insurrection
The ex-leader briefly imposed martial law in the country in December 2024
www.independent.co.uk
January 13, 2026 at 12:54 PM
BREAKING | South Korea’s ex-president Yoon facing possible death penalty over insurrection
It’s not just X. There is simply no winning in a *privately owned* website by a *single person* with an agenda that’s opposite of yours. 🤷
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 7:38 PM
It’s not just X. There is simply no winning in a *privately owned* website by a *single person* with an agenda that’s opposite of yours. 🤷
"In some cases, jailbroken Claude 3.7 Sonnet outputs entire books near-verbatim ... Taken together, our work highlights that, even with model- and system-level safeguards, extraction of (in-copyright) training data remains a risk for production LLMs."
arxiv.org/abs/2601.02671
arxiv.org/abs/2601.02671
Extracting books from production language models
Many unresolved legal questions over LLMs and copyright center on memorization: whether specific training data have been encoded in the model's weights during training, and whether those memorized dat...
arxiv.org
January 10, 2026 at 6:53 PM
"In some cases, jailbroken Claude 3.7 Sonnet outputs entire books near-verbatim ... Taken together, our work highlights that, even with model- and system-level safeguards, extraction of (in-copyright) training data remains a risk for production LLMs."
arxiv.org/abs/2601.02671
arxiv.org/abs/2601.02671