Profile avatar
tedunderwood.me
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. Author of Distant Horizons (Chicago, 2019). Pluviophile.
14,917 posts 18,653 followers 5,184 following
Regular Contributor
Active Commenter

My notes on Anthropic's substantial essay about how they built their multi-agent research system, which has finally talked me around to taking multi-agent LLM prompt engineering seriously simonwillison.net/2025/Jun/14/...

An interesting moment in Deni Ellis Béchard’s *We Are Dreams in the Eternal Machine* that speaks to the AI inevitability discourse “It was a convenient argument. I denied my role in making you while simultaneously taking credit.”

Math and humanities are almost in mirror positions in regards to current RL experience design. Biggest issue with math is problem generation. In the humanities, sampling new questions is relatively trivial, actual pain is verification (which could exist: source criticism is one).

The Case Against Multi-Agents Cognition (i.e. Devin) coins the term “context engineering”, successor to prompt engineering and argues that multi-agents don’t pass context effectively cognition.ai/blog/dont-bu...

I was brought up to believe that there are an infinite number of internally consistent language games—so evidence from the games themselves would never give us any way to choose between them. But apparently that was wrong and Plato was right? WTF?

pretty strong argument for multi-agents www.anthropic.com/engineering/...

I mean, Oakland only has 400 thousand people!

Without Brett Bobley, I literally wouldn't have a career, or at least it would be very different and surely not as cool. From the 2010 UCLA IATDH to him *hiring me* three years ago, my life has been shaped by pieces @brettbobley.bsky.social set in motion. #ThanksBrett for everything. You're a giant.

“No Kings” rallies draw huge crowds across the country despite violence and threats from the right: www.washingtonpost.com/nation/2025/...

Brought a stack of about 40 proofs & off-prints from our “No Kings!” run to CU’s protest & they were gone in about 5 minutes. It was energizing to be in such a crowd for our little college town. #NoKings

Happened to be in Arlington, TX today for #NoKings 🚫👑

This probably isn't the platform to use to make this call -- I don't know what is -- but there is an urgent need for people in the broadly "liberal" alliance against Trumpism to figure out how to make use of LLMs to serve our political and social goals.

People get ready… (predictable Early Americanist slogan, but appropriate for the occasion!)

No we need more philosophical inputs on reasoning design, not offloading that impact and ethics section you don’t want to write.

Paper is finally up and open access (www.journals.uchicago.edu/doi/10.1086/...), it's a sequel to an earlier paper where we'd argued that there's not good evidence that pre-publication peer review is a net benefit (www.journals.uchicago.edu/doi/10.1093/...). So in this one we suggest an alternative.

"To begin w/...[Brutus] obliged [the ppl] to swear an oath that they would suffer no man to be king in Rome ('adegit neminem Romae passuros regnare'), lest they might later be turned from their purpose by the entreaties or the gifts of princes." (🚫👑) the Republic's founding, 509 BCE Livy, AUC, 2.1

A bunch of papers suggest that if X and Y are independent tasks, we might expect to see "emergent" behavior on "X and Y" or some task that requires first X and then Y. I'm really surprised I can't find any papers that dig into this; it's usually a side comment. Do you know any?

I’ve seen a lot of hand wringing about the dangers of going to No Kings protests and gently, I want to tell people to chill a little. These are permitted marches and the point is to be out there and seen en masse. Please don’t scare off people who are maybe going to their very first protest

My wife and kid are at Michael's right now getting poster board.

If you're in the US, there's an event near you tomorrow. www.nokings.org#map

#ThanksBrett for always looking forward. You saw a bright future for the humanities, as key interlocutors in the most pressing socio-technical discussions of our moment, and you created opportunities for those conversations to actually happen. We all owe @brettbobley.bsky.social enormous debts

Now *this*, from @kerryhowley.bsky.social's piece on Pete Hegseth, is how you compose a methods graf: nymag.com/intelligence...