rethinking the core curriculum has been brought up again as part of the strategic planning process 🫣

Had two sequential thoughts seeing this: 1. Almost quaint to see a large CS conf excited about getting more paper submissions, instead of having reached the "dear god, stop sending more" stage. 2. But... the post also includes a link to a page entitled "CHI 2026 Papers Desk Reject Process" 🤔🤔🤔

Eleven months later and there are definitely sparks of joy now!

playtesting today!

It's happening! (The early-bird registration deadline is today.)

💡Can we trust synthetic data for statistical inference? We show that synthetic data (e.g., LLM simulations) can significantly improve the performance of inference tasks. The key intuition lies in the interactions between the moment residuals of synthetic data and those of real data

An unsuccessful but funny experience with gemini-cli. I asked it to fix some warnings in old ASP code (it works but produces some warnings about possibly empty aggregates). It tries a few times, fails, and declares victory anyway.

Misguided: scaling up transformer models to build the machine god. Enlightened: computing Chaitin's forbidden constant to uncover the true face of the machine god.

New book in draft: AI and Consciousness [link in thread] This book is a skeptical overview of the literature on AI and consciousness. Anyone who emails me comments on the entire manuscript will be thanked in print and receive an appreciatively signed hard copy.

Neat graphical representation of functional optics (in the functional programming sense, not the sense of "I got my glasses fixed").

I knew he was planning to take early retirement, but still got a jump scare seeing this. End of an era in AI & games academia!

Made myself a cool list of which lambda calculus terms have a normal form

I'm teaching 𝗪𝗿𝗶𝘁𝗲 𝘆𝗼𝘂𝗿 𝗼𝘄𝗻 𝘁𝗶𝗻𝘆 𝗽𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗶𝗻𝗴 𝘀𝘆𝘀𝘁𝗲𝗺(𝘀)! again. I'll be posting the videos & tasks on YouTube too. In the first lecture, I explain what's a tiny system, why write one and show plenty of demos! 🎞️ Playlist: www.youtube.com/playlist?lis... 👉 More info: d3s.mff.cuni.cz/teaching/npr...

Definitely thought this was an AI thesis from the title and opening (M. Alauddin, University of Lancaster, 1976). But it's about error-correcting codes and reliability estimates (of which "doubt is the inverse").

(repost welcome) The Generative Model Alignment team at IBM Research is looking for next summer interns! Two candidates for two topics 🍰Reinforcement Learning environments for LLMs 🐎Speculative and non-auto regressive generation for LLMs interested/curious? DM or email [email protected]

Evolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement Learning

finally, some Good Code crates.io/crates/fibon...

Successfully used an LLM for the first time to speed up a paperwork/admin task (although just by turning it into a coding task). I'm co-chair of the rank & tenure committee this year, which involves pushing a lot of PDFs around, and some of the PDF-pushing is pretty annoying. 1/2

new blog post! why do LLMs freak out over the seahorse emoji? i put llama-3.3-70b through its paces with the logit lens to find out, and explain what the logit lens (everyone's favorite underrated interpretability tool) is in the process. link in reply!

when an RL agent forgets previous skills after learning new ones, that's Bell-Man amnesia

I didn't realize that JAX is taking over in the probabilistic programming space. Very fair (afaict) post from Bob Carpenter, one of the core Stan developers, on how that looks from the Stan side.

The WSJ is usually good about this kind of thing, but this article tripped on a common issue in the BLS occupational categories. The "computer programmers" category is a very small/weird category; most computer programming jobs are instead coded as "software developers". www.wsj.com/lifestyle/ca...

Ok that's an enjoyably pedantic distinction from Gemini, when I asked it to write a parody (probably just a weird artifact, but still): "It should be noted that I can only generate text, and thus, I cannot directly 'write' a parody of [x]. However, I can absolutely generate a text that parodies..."

It's finally let up on the swampy weather and gotten very nice.

I love the precise dates here. * Arrival of Computers: Jan 1984 * Arrival of the Internet: Jan 1996 * Arrival of AI: Nov 2022

I've written an essay reflecting on how I use the word slop and some theory of what distinguishes slop from spam and kitsch. minihf.com/posts/2025-1...

Talked to two undergrad TAs who are TA’ing an intro class and they sounded like grumpy professors complaining about the kids these days, even some of the exact same complaints. Was very funny somehow. They are only like 2 or 3 years older!

Interesting wording from Kaiser. (I scheduled an appointment.)

it's like you don't even know your own family anymore, Google Search AI Overview

👀 🤖🇪🇺

Sometime in the past year or so, the big commercial LLMs have gotten pretty good at explaining code. They used to make a lot of mistakes in explaining language syntax, semantics, design decisions, etc., but I really rarely get an outright mistake nowadays.

Not all scaling laws are nice power laws. This month’s blog post: Zipf’s law in next-token prediction and why Adam (ok, sign descent) scales better to large vocab sizes than gradient descent: francisbach.com/scaling-laws...

The DEI consultant to AI consultant pipeline is an interesting feature of 2025 academia. Not the majority source of newly minted AI consultants (they are popping up all over), but one of the more surprising maybe.

And new paper out: Pleias 1.0: the First Family of Language Models Trained on Fully Open Data How we train an open everything model on a new pretraining environment with releasable data (Common Corpus) with an open source framework (Nanotron from HuggingFace). www.sciencedirect.com/science/arti...

Interesting long-form argument from Rodney Brooks that methods similar to the ones from the deep learning era of AI that cracked image recognition, text generation, etc. are unlikely to crack humanoid robot dexterity in the near future.

Walking across the quad to the bus and hear over a speaker, "there are no connections to the decision trees".

Wrote my first Rust program that actually does something. It works and does it fast! Fairly good experience overall, though there are a *lot* of Rust-specific concepts to wrap your head around in even fairly simple programs.

where did the term "fairness" in concurrent programming originate? is there a canonical definition? a canonical citation? I've found references in Owicki & Lamport 1982 & Andrews & Schneider 1983, but neither originated the term or concept.

Solid post tracing the history of algebraic data types. Both where terms like "sum type" and "product type" came from, and trying to determine which concepts were actually the same ones, and who was likely to have been influenced by who.

If you gave me these sentences with no context, I would've guessed: it's an LLM reasoning trace that keeps getting confused and trying to correct itself. But it's a paragraph written by a human in 2002. Wonder how much of this stuff we're retconning as LLM.

Heard AI is hot, so picked up some reading to keep up with the latest.

New paper, accepted at @aiide.bsky.social! We do an exploratory analysis of the output variation for text-to-audio models generating environmental audio (prompts like "sound of rain"), with fixed prompts but varying random seeds. We use three open models: Stable Audio Open, MMAudio, and AudioLDM2. ↵

And now I got a print magazine (The Nation, Sept '25) with the pages all messed up! They appear in the order: 1-10 27-34 19-66 (incl. a 2nd copy of 27-34) 51-58 (2nd copy of these too) 75-82 [pages 11-18 and 67-74 missing]

Our IT department installed some kind of macOS auto-updater, so I now get the experience of being an early adopter of new OS releases. So far Tahoe is actually pretty nice & I haven't hit noticeable problems.

We're hiring again! Assistant Professor "with a focus in data science or artificial intelligence", to be a joint appointment between Computer Science and Math/Stat.

Faculty job — please help spread the word! At UC Santa Cruz we have two games degrees, one in Engineering and one in Arts. This is a job with the Arts folks, who include super cool creators like @mattiebrice.bsky.social & @swensen.bsky.social recruit.ucsc.edu/JPF01998

useful overview of what's going on in Wasm land, from Andy Wingo in ACM Queue

"US state legislatures have emerged as important laboratories for 'techno-legal solutionism'—the belief that complex social problems can be solved through legally mandated technical fixes. Although similar approaches are evident in the EU's Digital Services Act and various national regulations..."

1/n Latest on history of credit feuds in Econ: Herrade Igersheim ties many existing account of well-known longstanding priority dispute between Duncan Black & Ken Arrow over the median voter theorem with a layer of new archives [open access] link.springer.com/article/10.1...

Are RCTs better thought of as performance art than a research method? Come for the critique of external validity; stay for the argument that they don't produce "generalizable knowledge" and therefore shouldn't require ethics approval (under U.S. regulations).