Fast and furious
A big week for AI, the low-education coalition usually wins, biggest 2024 swings by metro, news influencers, 2026 House ratings are out
No. 348 | February 7th, 2025
🤖 Artificial Intelligence
The announcements from the AI world are now coming fast and furious.
OpenAl’s new Deep Research model can perform complex multistep reasoning tasks with access to the web. Will this replace research assistants in academia?
Tyler Cowen is also emphatic:
I have had it write a number of ten-page papers for me, each of them outstanding. I think of the quality as comparable to having a good PhD-level research assistant, and sending that person away with a task for a week or two, or maybe more.
Except Deep Research does the work in five or six minutes. And it does not seem to make errors, due to the quality of the embedded o3 model.
It seems it can cover just about any topic?
I asked for a ten-page paper explaining Ricardo’s theory of rent, and how it fits into his broader theory of distribution. It is a little long, but that was my fault, here is the result. I compared it to a number of other sources on line, and thought it was better, and so I am using it for my history of economic thought class.
Deep Research also shows a big leap ahead on Humanity’s Last Exam, one of the tests of whether AGI has been achieved. And we are probably looking at ASI — artificial super-intelligence — in the near future.
You can even turn lower-level models into advanced reasoning models by simply adding the word “wait,” preventing it from cutting short its thought process.
PDF parsing is solved at scale.
A fundamental transformation of knowledge work is not a question of if, but when. And like the discussion of Jevon’s Paradox in the context of DeepSeek, what I think this is likely to mean isn’t fewer researchers, but researchers who are able to focus on high-level knowledge synthesis while delegating the gruntwork to AI. Academics are the first in line for disruption, but it’ll be a healthy disruption. The current research and peer review process can take years. When I’m reading a paper on, say, political persuasion experiments, it’s likely the experiments themselves took place in 2016 or 2018 — eons ago. That’s a timeline that doesn’t work at all for practitioners — but that could hopefully change soon as the parts of the research process that can be done well are sped up by AI.
🇺🇲 2024
The chart by Patrick Flynn plotting the party coalitions over the last generation is the essential chart of the 2024 cycle, and should be as influential as Lee Drutman’s scatterplot of the American electorate after 2016.
It’s possible to draw a simple line through it such that you can predict the winning candidate in 7 of the last 8 elections based on the makeup of their coalitions. Generally speaking, the low-education coalition is on the winning side, with the only outlier being 2020.
Trump dominated posts by news influencers across platforms, who were also more critical of Harris. But influencers affiliated with traditional newsrooms were more critical of Trump.
🇺🇲 2026
Cook’s 2026 House ratings are out, suggesting yet another tight battle for control.
📊 Public Opinion
Around 4 in 10 Americans know someone transgender, according to Pew Research, a number that’s higher among adults than teens.
🗺️ Data Visualization
Striking geospatial visualization of income and housing density along the Detroit-Grosse Pointe Park border.
