wallingf

@TheAnnaGat Congratulations to all of my Hungarian friends!

As a Hungarian, I confirm this is true. We should also accept America Ferrera.

@YateghteghSKY I'm glad to see that some researchers are now considering how we might impose structure that improves the performance of these systems, without losing too much of the advantage of finding statistical patterns in data.

@YateghteghSKY This is a big question! My impression is that much of what is hot in AI right now tries to avoid imposing structure on systems, letting statistical techniques find patterns without being biased by what humans think is happening. 1/2

This is the dream: train an LLM on the 20+ year cache of the session notes, assignments, and readings I've created for my classes, then use it to generate lectures, talks, and presentations. I'll replace myself and get out my hammock. gocomics.com/doonesbury/202…

There's a fantastic paper about this called The Discoveries of Continuations by John Reynolds. homepages.inf.ed.ac.uk/wadler/papers/…

Cesar Geronimo

@infinitary “I already have a program, and now I *really* need a programmer.”

@wallingf Are you sure? I could see "can you fix my printer" turn into "can you fix my vibe coded program".😅

Ideas were always cheap. Now execution also became cheap. Marketing is where it's at.

"a computer scientist who doesn't affect the way you think about programming, is not worth knowing" RIP Tony Hoare. blog.computationalcomplexity.org/2026/03/tony-h…

Astronomical spring and meteorological spring are important, of course, but tomorrow is the most important day of March: sartorial spring, when Prof. Wallingford flips the bit and switches into shorts for the year.

i am pretty sure every song i have ever written was inspired by “boys of summer”. a total masterpiece.

Go Tristan and the Panthers! @UNImbb

I have a post-it note on my desktop on which I record 2FA codes that are palindromes, fun patterns, or cool numbers. It's an unusual affectation that gives me a little bit of pleasure.

@JustDeezGuy I loved programming in Eiffel. Would like to have tried CLU.

Scala gets this right, as did CLU, Eiffel, and Sather. That no one has heard of the latter three and the first is in a tiny niche tells you just how pervasive conceptual slop in PROGRAMMING LANGUAGE DESIGN has been, and for how long.

@infinitary I still don't know where it will land. I think there will still be a profession, but for how many of them? The advent of compilers didn't decrease the demand for programmers, quite the opposite, so there is reason to hope, I think.

Had a bit of a think and I'm getting more and more convinced that despite a long tail all of programming is going to go the robots' way. I really don't feel safe for youngsters picking this profession.

@krismicinski @KleeneAlgebra If the admin and current trends have their way, kids won't be able to learn any of that democratic socialist stuff in school.

@wallingf @KleeneAlgebra My parents both have ugrad degrees but I’m banking on my kid aiming higher and being “Zohrans communist enforcer”

To me it was surprising the number of people that have parents with PhDs

@krismicinski @KleeneAlgebra I was a first-gen college graduate, and this gap makes a lot of sense to me. Throughout my career, especially early on, there were so many things I had no clue about. That leads to suboptimal choices sometimes, or simply not pursuing an opportunity that I wasn't aware of.

@KleeneAlgebra yeah academia is weirdly insular wrt parent phds, which is strange because i feel like it doesn't seem that way at a glance

@RobertTalbert Too bad Jones got hurt just as the competition got stronger. I had hoped they might squeeze into the playoffs anyway, but their defense doomed them.

I think I might have been half right on this one. Definitely no more than half.

@krismicinski @tritlo Agreed, though I sometimes wonder... When people use a stock holding as collateral for a loan (used by the ultra-rich in lieu of income), I think that they have realized value. I'm trying to imagine a way to begin taxing capital gains then.

@tritlo Taxing unrealized gains is a clearly-stupid idea, no idea why anyone ever proposes this as a serious idea. On the other hand, I feel the capital gains tax rate should clearly be increased. No idea why the people proposing the first don’t switch to the second.

taxing unrealized gains is insane. They’re not real! They don’t exist! that’s why they’re unrealized!!

@lemire My personal rule for writing a framework is to wait until I have at least three applications. This helps me find the abstractions that generalize over multiple apps and to see the variations of implementation that inform good design choices for speed, space, rtc.

A recurring anti-pattern in software goes as follows. Folks decide to standardize an overly generic programming interface, based on a prototype. They don’t care about performance… why should they? They have no serious users yet. And few people really need the performance ever… because few people have lots of users… You might as well throw the kitchen sink at it; it costs you little because your prototypal implementation doesn’t address performance, and it is all high-level code anyhow. And then we get stuck with a poor design performance-wise and have to do our best with it. It can mean supporting it fully and discouraging its use, or only supporting a subset. But why? Why does performance matter if, as I said, most projects hardly have users? Because the value of software depends on its performance. If you doubt that performance matters, go buy a used iPhone 7. In theory, anything can be made fast… but in practice, the engineering cost to optimize a design that did not take performance into account can be too much for anyone to pay. If you can rewrite your code using an interface that is potentially 10x or 100x faster, while being just as clean and maintainable… you are a much better engineer. Do people notice? You bet they do. So what should you do? Focus on simplicity. Do not make your interface more generic than it needs to. If your first version is missing features, you can easily add them during a revision. And, importantly, benchmark even your prototype: do not propose to replace existing code with some new much slower design.