nice skies in DC as the thunderstorm line cleared out

An interesting thing about LLMs in Python is that they seem to broadly push code towards some kind of conventional wisdom about best practices, as judged maybe by whoever is setting up the posttraining recipes (I say "in Python" mostly because I notice that more strongly in Python).

lol at this report back by an American visiting Tsukuba Expo '85 onlinelibrary.wiley.com/doi/abs/10.1...

Die Ausstellung “Captured in Code” zeigt die Vielfältigkeit von Fotografie in digitalen Spielwelten. 11. - 21.6.2026 | Depot Dortmund | Eintritt frei Eröffnung: Donnerstag, 11. Juni 2026, 19:00 Uhr Alle Infos, Programm & Anmeldung für die Workshops: www.capturedincode.de

The most productive scientists are driven by resentment, rivalry, and spite. But too many academics work in collegial obscurity. That’s why I’m offering the opportunity of a lifetime to invest in Tinder for Nemeses

Intriguing-looking postdoc on human-centric reinforcement learning, part of "an interdisciplinary project between artificial intelligence, social sciences, and societal partners". At TU Eindhoven with Hendrik Baier, deadline July 9. www.tue.nl/en/working-a...

Now published in open access! Your one-stop shop for the philosophy of language models. It's the spiritual descendant of our two-part preprint from 2024, fully updated. This should be particularly useful for anyone looking for an entry point into this rapidly growing field.

I don't know a lot about world models, but this blog post seems to help disambiguate (at least for me?)

"The Computer History Museum is seeking an organized, creative, and tech-savvy Oral History Program Manager to lead the day-to-day operations of the Museum’s Oral History Program and translate these stories into compelling digital content for broad audiences."

As GenAI systems do more and more, even generate proofs, the focus shifts to getting high-quality specifications. What can a *responsible* programmer do? Any solution should generate work that is *meaningful* and *moderate*. We've been working on that! blog.brownplt.org/2026/06/09/p...

finally some off-the-grid local AI

I've spent months rethinking and rebuilding my programming languages course from scratch for the agentic coding era. I wrote myself a memo explaining what I'm doing. I figure others might be interested in the redesign, so here goes! Feedback welcome of course. docs.google.com/document/d/e...

A thing LLM coding sort of makes more feasible is making *smaller* personalized apps. Like instead of WMATA's big and annoying to use transit app, I'm trying out a custom little thing that just shows me the 2 bus lines I take. Lines hardcoded; stops hardcoded; no configuration; barely any interface.

The Computerized Society (1987)

watched Breaking Bad for the first time. pretty good!

At the Simons Foundation for a book release event for a new book about Lean, "The Proof in the Code: How a Truth Machine Is Transforming Math and AI".

always enjoyable to be on the rails at sunset

In other news, my team is hiring for *11* Deep Learning FTEs on the applied science/research engineer spectrum across the US and Canada (about 1/3 of which must be in East Coast working hours). Also looking for 2 academic year graduate level interns.

Spending a decade fighting a "war on Wikipedia": in retrospect probably not a good use of academia's political & cultural capital.

Assistant/Associate Professor opening at the University of Southern Denmark in games & interactive technologies; focus areas Human-Computer/AI Interaction, data science, and AI. Deadline August 1.

Increasingly, when journalists report “AI is causing X (surge in article submissions, layoff spike &c),” I place zero faith in the story until I see someone’s first-hand analysis of underlying data. Narrative is a hell of a drug.

colleague lent me a book

Small-scale study: 16 law professors were asked to judge short-answer Q&A practice materials written by one of the other 15 professors, or generated by an LLM (Gemini 2.5 Pro). They preferred the LLM materials in 75% of cases, with fairly strong inter-rater agreement. Has some add'l analysis.

2021-era image generation is back!

incredibly clarifying to read that people were telling Alan Turing in 1951 that when he thought the computer did something cool, it was only because he did all the work for it, and that he was stupid for being impressed

I wrote an essay exploring how structuralism and deconstruction have informed recent theory around 'AI' and 'LLMs' including @leifw.bsky.social's Language Machines: olrsupplement.com/2026/06/01/w...

good skies this evening

Truly bizarre stuff happening to the writing styles simultaneously with coding getting better. Tempting to conclude it's the result of a common posttraining mechanism but not sure there's real evidence of that yet.

CAPTCHA asking me to "prove you're not a computer". [*Sweating & trying to remember enough of the Fodor-Pinker debate from grad school to get through.*]

I find this kind of statement interesting. Not necessarily wrong (depending on how you defined "shaped"), but there's plenty of philosophy/history of technology that is less confident in making flat statements like that. www.ft.com/content/acea...

Submission form is now open for Game Poems Magazine Issue 2! Learn more and submit your game at www.gamepoems.com/submit Submission deadline is July 1, but early submissions are welcome. Thinking of submitting something, but haven't started yet? No worries, you've still got a whole month. :-)

A root word collision: the Greek word for "nationalism" is roughly "ethnicism" (εθνικισμός). So if you want to translate the English word "ethnonationalism", it becomes... ethnoethnicism (εθνοεθνικισμός).

Cut though Tenleytown murder alley. Seems pretty peaceful today.

Asked Gemini something about DC apartment logistics, and it recommended I "pop down to the front desk" to verify. Takeover of Google AI by DeepMind Londoners confirmed.

A reason I've pulled back from reviewing for big AI conferences is a feeling that I'm doing unpaid supervision of other people's PhD students. Too many ppl submitting 10+ papers to a single conference where I doubt the prof whose name is on the paper has done a thorough review & revision themselves.

This looks interesting, especially if it lives up to the praise in one of the blurbs that it "deftly avoid[s] both technophobic romanticism and technophilic transhumanism".

if you are wondering about Mistral koenvangilst.nl/lab/mistral-...

Interesting new initiative to increase the proportion of our students who do internships: if they find an internship, but it's unpaid or underpaid, the university will top it up with a stipend.

[David Attenborough voice] Here, on the university campus, the woodlands look peaceful just now, basking in the rays of the fading evening sunlight. But... if we look closer... a lethal predator is resting.

Worth considering.

New-to-me origin story for Nate Silver's interest in politics: he used to make a full-time living beating schmucks in online poker, but some politicians screwed it up for him.

@princeton.edu is seeking to hire AI Postdoctoral Research Fellows affiliated with the AI Lab (@princetonainews.bsky.social) and the Center for Digital Humanities to conduct research focused on societal AI. Applications open through Jul 1! cdh.princeton.edu/news/2026/05...

Odd, our university president resigned only two years in, being replaced on an interim basis by the b-school dean.

I don't really believe in on-paper exams in comp. sci. I get why people are moving to them, but feels very backwards to me, like moving 40+ years into the past. We didn't even do paper exams in my own undergrad two decades ago.

The list of programs SUNY Fredonia is cutting aren't exactly what I would have guessed. I don't know anything about this campus though. www.fredonia.edu/about/office...

Postdoc in "People-Centric AI" at Microsoft Research Cambridge (UK)

looks interesting if you're in London: a four-day festival, 4-7 June, at the Royal Ballet & Opera on music+AI www.rbo.org.uk/tickets-and-...

Isabelle/Isar has a nicely named proof method, blast, that tries to brute force, letting you concisely prove simple but tedious lemmas by just writing 'thus "Q" by blast', which'll go through if in fact blast can prove it. In good cases, LLMs feel kind of like that for me now: pip install by blast.

passing through Philadelphia on the train

A pop AI book where the two people contributing blurbs were Al Gore and Sam Altman. A little microcosm of 2018.