Mission accomplished.

Important class updates from @mlittman.bsky.social :

Yikes. Well, at least your house isn't warm, which is an unfortunate consequence of actual fires.

TEAL 2026: Tools for Educational Activities in Logic TEAL 2026: Tools for Educational Activities in Logic @ FLoC2026

Now that people are starting to make travel plans to FLoC: if you're doing education, please consider submitting to Tools for Educational Activities in Logic! Lots of categories: you can even re-send us *existing papers* (please see CFP)! (Please reshare!)
teal.cs.brown.edu/floc2026/

This NYT article about the Venice (Art) Biennale [ www.nytimes.com/2026/04/19/a... ] reminds me that last year's Venice Architecture Biennale's American pavilion, Porch [ www.porchusavenice2025.org ] was simply gorgeous.

Specifically using QR codes? And there are enough safety measures around it (like if you were led to a spoof Pix website and asked to log in, either you would know or the login has other factors that keep you safe)?

Clint Eastwood, Eli Wallach and Lee Van Cleef on location in Spain during the filming of The Good, The Bad and the Ugly.

The Good, The Bad and the Ugly (1966) behind the scenes.

I've been checking on this for a while, and have finally started (as of 2025?) to find panhandlers displaying QR codes. (The security side of my brain sends klaxons blaring every time.)

It’s a cute theory and I like cute theories but I don’t believe this one at all (as a theory).

There is zero evidence for this (yet). It should be provable from git logs?

And the downthread complaints about Claude Code’s own code misses that there can be, and often are, tremendous differences between the code of a compiler and the code it generates.

RIP Luna the rat. After her sis Rizzo passed last month, she totally deteriorated. )-:

La Marche, Flemish, early 18th C. Commemorates the achievements of the Earl of Orkney in the War of Spanish Succession.

Man, old universities are weird places. Went to attend a talk (featuring linguists and computer scientists debating language models) in a room I hadn't been in before, and it turns out they have a full-ass Flemish tapestry just hanging from the wall.

The blossoms are out on campus, and Marcus Aurelius approves.

India’s Computer Science Grads Are Unprepared for the AI Revolution Companies like Infosys are running their new hires through many weeks of training to bring them up to speed on new programming tools.

I've been waiting for articles that point out the existential threat that AI coding technologies pose to the Indian tech sector. Here's one. I fear that the social and cultural—and hence political—consequences in India could be even worse than in the US.
www.bloomberg.com/news/feature...

This is a very good piece, and I'm glad you're updating your priors while maintaining your sense of balance. We need more of this kind of writing.

CC @joshuagrochow.bsky.social the post linked above may or may not tell you new things, but FYI.

Nor does it take into account the loops around the language model. It's not single-shotting a proof of any size. So there is some agent + tool + RL process that seems to be behind the success. I know enough to know I don't know, as opposed to the ones confident that they know why it can't. (-:

I do not! But there are numerous things that are not easily explicable from a "blurry JPEG" account. The blurry JPEG doesn't help me understand how it can single-shot code in a whole new programming language, for instance. Or maintain parallel structures between input and output. ↵

Haha. All I mean is an advanced scientific object that interests us, is weird, but whose underlying technology is too sophisticated for almost anyone to understand, so we use wooly metaphors to muddle our way through, often poorly.

FYI, about your post, bsky said this:

In a way, I feel LLMs° are the new quantum physics.

° Even the terminology is complex because a modern agentic coding system is like 7 layers wrapped around an LLM, so calling it an "LLM" does it a disservice. Indeed, when we had only ~3 layers around the LLM, it didn't do much useful at all.

However, a good deal of the social media slagging comes from people mis-applying or over-applying metaphors. "Blurry JPEG of the Web" doesn't actually help explain how it generates a Lean proof, but it's pretty useful if you want to try to speak knowledgeably about how it can't do anything. ↵

"LLMs are [just] spicy autocomplete" is not what any technically-aware person says. I fully agree that techo folks overindex on the technical aspects ("well akshually the number of attention heads…"), and ignore the social, cultural, political, and other aspects. ↵

The pleasure is mutual! Even though I think you're wrong about your metaphor ascription. (-:

Universities are amazing places. I can walk a few blocks from my office and spend the morning hearing @biella.bsky.social and @steveklabnik.com back-to-back at a humanities workshop about technology at a humanities workshop about technology.
humanities.brown.edu/events/AILab...

who the hell ddoses the millennial posting retirement home, seriously

Major company told our student "not their problem".

It's okay, every only about half-understands iframes.

Hearing today about summer interns struggling to find any kind of housing at all because their work city is hosting World Cup games. )-:

hieroglyphs code #coding #software #programmerhumour #tech #programming

> "Trust me bro, this new tech startup isn't a pyramid scheme"
> Looks at codebase: