Wrote up my current context engineering setup (seeded with a few thoughts about LLM hysteria): www.ohrg.org/taming-jagge...

Debasing myself today (using pip)

I've been wondering for a while if anything in Unicode could accidentally compute. I've just shown that @unicode.org UTS#35 transliteration rules are universal, computing Collatz with just 3 rewrite rules running on stock ICU, shipped with every major OS. seriot.ch/computation/...

Jacobian? Like the magazine?

There’s a non-zero chance that Pangram has learned to discriminate homoglyphs. Some real taste-making going on there if so

Grace@gracekind.net · last mo.

They swap out a single apostrophe character in your messages, based on your domain. Very sneaky!

Oh, you mean chain-of-thought? (Newell & Simon’s GPS report in 1958)

Excited to share that we are now inviting proposals for our upcoming Critical AI Studies? conference at @ucsantabarbara.bsky.social on Oct. 9&10, 2026. Please see the full CfP here: huml.ucsb.edu/conference. Submissions are due July 31, 2026. For questions DM me or @owenleonard.bsky.social.

Resurfacing online to mention: my book on the history of Speech Rec, NLP, and data-driven AI is coming out in a couple of weeks! Just got my author's copy from @uchicagopress.bsky.social and it finally feels real! press.uchicago.edu/ucp/books/bo...

Divination Engines

A revealing and surprising origin story, showing how attempts to render human speech and language computable led from the era of big data to today’s AI. Since the advent of computers, society has fant...

press.uchicago.edu

This little sentence in last month’s encyclical helped (=goaded) @leifw.bsky.social, @beenwrekt.bsky.social, and me to tie off some threads that spun out of the Cultural AI conference in March “Reify This”: www.theideasletter.org/essay/reify-...

Reify This

The authors contend that contemporary efforts to render AI systems interpretable rest on a mistake: reification, the process of treating abstractions and statistical artifacts as if they were concrete...

theideasletter.org

Tyler Shoemaker@t-shoemaker.bsky.social · 2mo ago

Well then

Only that which is p-hackable, is

Well then

For the subword fiends in your life: here’s a tool that helps to find token variants in HF tokenizers. We’ve been using it in Digital Theory Lab for experiments involving LLM vocab control (ablations, etc) More on the pipeline and model here: github.com/digitaltheor...

Bound to be one of the most important critical and theoretical books on the meaning of neural networks (and not just for vision). Congratulations to Leonardo Impett and Fabian Offert @zentralwerkstatt.org! Freshly published from meson press and University of Minnesota Press. @uminnpress.bsky.social

So excited that the book is coming out in November from @uminnpress.bsky.social!

This is a version of Zizek’s perfect date

mr. TIM@timkellogg.me · 3mo ago

The goal is to remove yourself from the loop Your role now is NOT to build software, but rather to build systems that build software full interview: m.youtube.com/watch?v=kwSV...

early view of my article is out! it explores the twinned ideologies of referentialism that often subtend language and "AI" and how those ideologies break apart in the R&D pipeline, where the social-pragmatic dimension of both speech + data come to the fore rai.onlinelibrary.wiley.com/doi/10.1111/...

‘Everything is a signal’: speaking circuits and noisy signs in the making of language‐oriented AI

Contemporary artificial intelligence (AI) technologies are often presumed to be capable of revealing unmediated truths about the world, including the truths language might hold, echoing the long-stan...

rai.onlinelibrary.wiley.com

Tremendous essay by Paolo Caffoni that approaches tokenisation in machine learning systems using critical translation literature with ideas from the study of political economy. doi.org/10.1007/s001...

It's Easter so I thought it would be appropriate to be back from the dead to announce that this is coming out at the end of this month: www.upress.umn.edu/978151792167...

Vector Media

Dispelling the notion of “generative” AINeural networks are designed to dissolve all media into the vector space—a universal space of commensurability....

upress.umn.edu

Coarsely summarizing perspectives on the bureaucratic culture of language models by @himself.bsky.social, @azjacobs.bsky.social, and Lily Chumley.

The Poetics of Bureaucracy

Language models are a bureaucratic technology

argmin.net

In this post, I serve as an unreliable narrator of the talk I gave two weeks ago at the Cultural AI conference at NYU (thanks again to @leifw.bsky.social and @t-shoemaker.bsky.social for organizing this unique event and for inviting me to share my perspective):

Artificial Intelligence, Interactive Measurements, and Assemblage Theory

Fuck Heidegger.

realizable.substack.com

New work from me in AI & Society: "Systems Programming the Model" theorizes LLMs as systems-level objects. Agentic AI makes this obvious now, but my argument is that LLMs have always been systems + link.springer.com/article/10.1...