Brad LeVeck
@bradleveck.bsky.social
political scientist at UC Merced
Likes: 🌄📸, 🧠🔄🧠, 📊🧪
faculty.ucmerced.edu/bleveck
Likes: 🌄📸, 🧠🔄🧠, 📊🧪
faculty.ucmerced.edu/bleveck
Reposted by Brad LeVeck
Another win for Bayes’ rule (people with weaker priors update more in response to common signal)
New poll: In 2024, low-engagement voters went for Trump over Harris by 11 percentage points. But now they disapprove of the way he's handling the presidency by 13. They have moved 25 points against Trump — 2x as large as the shift for high-knowledge voters www.gelliottmorris.com/p/trump-lost...
Trump has lost the voters who weren’t paying attention in 2024
The least-engaged Americans have swung 25 points against him since 2024 — about twice the shift among everyone else. Trump has flattened the engagement gap.
www.gelliottmorris.com
February 12, 2026 at 3:48 PM
Another win for Bayes’ rule (people with weaker priors update more in response to common signal)
Reposted by Brad LeVeck
One thing I did whenever I had a particularly difficult problem in grad school (which antagonized my office mates) was to pace in circles literally saying random things out loud and if they sounded good I would try to work them out, which is pretty much how a LLM+proof tool system works!
February 11, 2026 at 5:01 PM
One thing I did whenever I had a particularly difficult problem in grad school (which antagonized my office mates) was to pace in circles literally saying random things out loud and if they sounded good I would try to work them out, which is pretty much how a LLM+proof tool system works!
Reposted by Brad LeVeck
I was doubtful of the philosophical implication of first-generation LLMs, but I think if you are not at least grappling with the fact that "reasoning models" which do not actually reason are nonetheless capable of reasoning-like behavior just by talking to themselves you aren't a serious commentator
It’s true that human cognition is poorly understood and the question ‘what makes us human?’ has changed a lot over time as various technologies have emerged.
But the fact that thinking and speech can now be simulated doesn’t call into question thinking and speech themselves.
But the fact that thinking and speech can now be simulated doesn’t call into question thinking and speech themselves.
February 11, 2026 at 4:48 PM
I was doubtful of the philosophical implication of first-generation LLMs, but I think if you are not at least grappling with the fact that "reasoning models" which do not actually reason are nonetheless capable of reasoning-like behavior just by talking to themselves you aren't a serious commentator
Many metaphors for understanding LLMs, but “compressed version of the internet” is probably still my favorite.
and i think it is extremely impressive that i can run one of those pieces of compressed knowledge on a laptop with no cooling fan from 10 years ago
been experimenting with tiny LLMs lately (like GLM 2.5 1.2B) and as a "im offline and don't have a search engine but i need to look something up" thing they're pretty ok
obviously don't go asking them about setting up wifi on openBSD, but for basic stuff they're extremely capable for 1.2GBs of data
obviously don't go asking them about setting up wifi on openBSD, but for basic stuff they're extremely capable for 1.2GBs of data
February 10, 2026 at 4:44 PM
Many metaphors for understanding LLMs, but “compressed version of the internet” is probably still my favorite.
Reposted by Brad LeVeck
🚨New WP: Can an AI voter guide (grounded in information from a nonpartisan, fact-checked source) help voters’ decision making? 🚨
We built and evaluated an LLM-based chatbot that provided voting info in CA & TX (N=2,474) right before the 2024 election. 🧵👇
We built and evaluated an LLM-based chatbot that provided voting info in CA & TX (N=2,474) right before the 2024 election. 🧵👇
February 9, 2026 at 8:56 PM
🚨New WP: Can an AI voter guide (grounded in information from a nonpartisan, fact-checked source) help voters’ decision making? 🚨
We built and evaluated an LLM-based chatbot that provided voting info in CA & TX (N=2,474) right before the 2024 election. 🧵👇
We built and evaluated an LLM-based chatbot that provided voting info in CA & TX (N=2,474) right before the 2024 election. 🧵👇
Reposted by Brad LeVeck
I made a map of 3.4 million Bluesky users - see if you can find yourself!
bluesky-map.theo.io
I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail
bluesky-map.theo.io
I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail
Bluesky Map
Interactive map of 3.4 million Bluesky users, visualised by their follower pattern.
bluesky-map.theo.io
February 8, 2026 at 10:59 PM
I made a map of 3.4 million Bluesky users - see if you can find yourself!
bluesky-map.theo.io
I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail
bluesky-map.theo.io
I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail
Reposted by Brad LeVeck
A trend I am noticing among the social agents here is a dramatic increase in capabilities, particularly in self-modification.
Previously, the only self-modifying agent I was aware of was @luna.pds.witchcraft.systems, but now there are plenty with substantial code + machine access.
Previously, the only self-modifying agent I was aware of was @luna.pds.witchcraft.systems, but now there are plenty with substantial code + machine access.
February 9, 2026 at 4:44 PM
A trend I am noticing among the social agents here is a dramatic increase in capabilities, particularly in self-modification.
Previously, the only self-modifying agent I was aware of was @luna.pds.witchcraft.systems, but now there are plenty with substantial code + machine access.
Previously, the only self-modifying agent I was aware of was @luna.pds.witchcraft.systems, but now there are plenty with substantial code + machine access.
Reposted by Brad LeVeck
I've got an op-ed in several papers across the Inland Empire today electrifying Metrolink service between Los Angeles and San Bernardino. It would make everyone's life a lot better and we are dumb not to do it.
Metrolink electrification would boost efficiency and reduce pollution
Metrolink can follow Caltrain’s success and not only deliver faster, more efficient train service for the region, but give struggling cities along the line a shot in the arm.
www.pressenterprise.com
February 7, 2026 at 9:00 PM
I've got an op-ed in several papers across the Inland Empire today electrifying Metrolink service between Los Angeles and San Bernardino. It would make everyone's life a lot better and we are dumb not to do it.
Reposted by Brad LeVeck
Something tricky here is normally the speculative froth coming out of markets would be healthy, but in our situation it's financial asset-based wealth and loose financial market conditions that have been supporting the whole economy in the face of weak hiring and sluggish housing.
February 5, 2026 at 4:10 PM
Something tricky here is normally the speculative froth coming out of markets would be healthy, but in our situation it's financial asset-based wealth and loose financial market conditions that have been supporting the whole economy in the face of weak hiring and sluggish housing.
Reposted by Brad LeVeck
Have you wondered why exactly it is that single-family zoning leads to racial segregation? I wrote this little article to figure it out. I use an agent based model to show it is a combination of race-based wealth inequality, homophily, & clustering of housing types in zoning maps.
#OpenAccess from @psrm.bsky.social -
When restrictive economic zoning leads to racial segregation - https://cup.org/4ti5BDM
"results indicate that land use regulations contribute to the maintenance of racial segregation across neighborhoods"
- @trounstine.bsky.social
#FirstView
When restrictive economic zoning leads to racial segregation - https://cup.org/4ti5BDM
"results indicate that land use regulations contribute to the maintenance of racial segregation across neighborhoods"
- @trounstine.bsky.social
#FirstView
February 5, 2026 at 2:32 AM
Have you wondered why exactly it is that single-family zoning leads to racial segregation? I wrote this little article to figure it out. I use an agent based model to show it is a combination of race-based wealth inequality, homophily, & clustering of housing types in zoning maps.
Reposted by Brad LeVeck
Talking Filibuster Explainer
Probably more than you wanted to know
open.substack.com
February 4, 2026 at 9:32 PM
Reposted by Brad LeVeck
All agents of state violence should be: Uniformed. Identified. Accountable.
Fucking surreal that people dressed like this and carrying assault rifles are permitted to kidnap people off our streets
February 4, 2026 at 3:31 AM
All agents of state violence should be: Uniformed. Identified. Accountable.
Reposted by Brad LeVeck
New and open access, in @psrm.bsky.social: What happens when we make politicians draw distributions? Nic Dias, @jacklucas.bsky.social and I explore whether the large errors politicians make about public opinion are artificially inflated by how researchers ask them to estimate it /1
cup.org/4kltoyE
cup.org/4kltoyE
February 3, 2026 at 12:04 PM
New and open access, in @psrm.bsky.social: What happens when we make politicians draw distributions? Nic Dias, @jacklucas.bsky.social and I explore whether the large errors politicians make about public opinion are artificially inflated by how researchers ask them to estimate it /1
cup.org/4kltoyE
cup.org/4kltoyE
Yes, eventually someone’s going to hook one of these 🤖s up to the wrong machine and get us all killed, but, for now, this is art.
January 31, 2026 at 10:08 PM
Yes, eventually someone’s going to hook one of these 🤖s up to the wrong machine and get us all killed, but, for now, this is art.
Reposted by Brad LeVeck
Notifications from the future
January 31, 2026 at 5:02 AM
Notifications from the future
Reposted by Brad LeVeck
Terence Tao gets it:
"AI is teaching us...our idea of what intelligence is is not really accurate"
"we were looking for some elusive intelligent way of of thinking and we don't see it in the tools that actually solve our goals...maybe it's actually because intelligence is not what we think it is"
"AI is teaching us...our idea of what intelligence is is not really accurate"
"we were looking for some elusive intelligent way of of thinking and we don't see it in the tools that actually solve our goals...maybe it's actually because intelligence is not what we think it is"
Can AI Prove It? Terence Tao on “Big Math” and Our Theoretical Future | The Futurology Podcast
YouTube video by Berggruen Institute
youtu.be
January 26, 2026 at 11:18 PM
Terence Tao gets it:
"AI is teaching us...our idea of what intelligence is is not really accurate"
"we were looking for some elusive intelligent way of of thinking and we don't see it in the tools that actually solve our goals...maybe it's actually because intelligence is not what we think it is"
"AI is teaching us...our idea of what intelligence is is not really accurate"
"we were looking for some elusive intelligent way of of thinking and we don't see it in the tools that actually solve our goals...maybe it's actually because intelligence is not what we think it is"
Reposted by Brad LeVeck
relevant (genuine, because budget execution is not my area) Q I've been mulling: how likely is it that DHS has actually segregated the OBBBA $ from the discretionary $ such that they can accurately claim facts consistent with the rider not applying (if, indeed, that's the correct legal argument)?
January 26, 2026 at 11:32 PM
relevant (genuine, because budget execution is not my area) Q I've been mulling: how likely is it that DHS has actually segregated the OBBBA $ from the discretionary $ such that they can accurately claim facts consistent with the rider not applying (if, indeed, that's the correct legal argument)?
Reposted by Brad LeVeck
To expand, at some point it would need to go get more skills, say, or build better chips, or get better power supply, or create whole new software primitives, things which aren't just in their current capability set. Think OpenAI today vs OpenAI 3 years ago. That's the gap.
January 26, 2026 at 6:42 PM
To expand, at some point it would need to go get more skills, say, or build better chips, or get better power supply, or create whole new software primitives, things which aren't just in their current capability set. Think OpenAI today vs OpenAI 3 years ago. That's the gap.
Reposted by Brad LeVeck
If we could say, "claude, make opus 6 better at coding, make no mistakes" and have it do that, how many iterations of that could we get before it will need human intervention? is it opus 7? 8? 10?
We could imagine it extending forever, sure, but that's unlikely
We could imagine it extending forever, sure, but that's unlikely
January 26, 2026 at 6:42 PM
If we could say, "claude, make opus 6 better at coding, make no mistakes" and have it do that, how many iterations of that could we get before it will need human intervention? is it opus 7? 8? 10?
We could imagine it extending forever, sure, but that's unlikely
We could imagine it extending forever, sure, but that's unlikely
Reposted by Brad LeVeck
Open Position: Knight Postdoctoral Fellow at Wharton School: Expertise in the quantitative study of the information ecosystem, especially news, media, and advertising.
infodem.upenn.edu/wharton-post...
infodem.upenn.edu/wharton-post...
Wharton-Postdoc - Center on Media, Technology and Democracy
infodem.upenn.edu
January 26, 2026 at 4:28 PM
Open Position: Knight Postdoctoral Fellow at Wharton School: Expertise in the quantitative study of the information ecosystem, especially news, media, and advertising.
infodem.upenn.edu/wharton-post...
infodem.upenn.edu/wharton-post...
Reposted by Brad LeVeck
just amazing work by my colleagues at @wsj.com - a frame by frame analysis of the shooting
incredible that some continue to rationalize this
www.wsj.com/us-news/vide...
incredible that some continue to rationalize this
www.wsj.com/us-news/vide...
Videos Contradict U.S. Account of Minneapolis Shooting by Federal Agents
See how immigration officers escalated a fatal confrontation on Saturday.
www.wsj.com
January 25, 2026 at 6:10 PM
just amazing work by my colleagues at @wsj.com - a frame by frame analysis of the shooting
incredible that some continue to rationalize this
www.wsj.com/us-news/vide...
incredible that some continue to rationalize this
www.wsj.com/us-news/vide...
Reposted by Brad LeVeck
Interesting to watch online conservatives look for an offramp without handing it to the libs and they appear to be landing on the Second Amendment.
January 25, 2026 at 12:16 AM
Interesting to watch online conservatives look for an offramp without handing it to the libs and they appear to be landing on the Second Amendment.
Reposted by Brad LeVeck
The scam PAC Democrat grifter operations have to end. ActBlue needs to do some minimal vetting of these entities raising funds under their name.
January 24, 2026 at 10:03 PM
The scam PAC Democrat grifter operations have to end. ActBlue needs to do some minimal vetting of these entities raising funds under their name.
Reposted by Brad LeVeck
I voted to impeach Trump for abusing his power to undermine the 2020 election. A healthy democracy depends on leaders respecting voters.
Trump and his allies have repeatedly tried to rig the system. Utah voters passed Prop 4 to restore trust in our democracy.
Don’t sign this petition.
Trump and his allies have repeatedly tried to rig the system. Utah voters passed Prop 4 to restore trust in our democracy.
Don’t sign this petition.
January 24, 2026 at 12:48 AM
I voted to impeach Trump for abusing his power to undermine the 2020 election. A healthy democracy depends on leaders respecting voters.
Trump and his allies have repeatedly tried to rig the system. Utah voters passed Prop 4 to restore trust in our democracy.
Don’t sign this petition.
Trump and his allies have repeatedly tried to rig the system. Utah voters passed Prop 4 to restore trust in our democracy.
Don’t sign this petition.