Mastodawn

Today, thanks to a former PhD student, I learned that LLMs are terrible at generating Whitespace code: https://arxiv.org/pdf/2603.09678

Also today, I was invited (and I am accepting) to join a group on redesigning our undergraduate curriculum, which includes discussions on our choice of first programming language.

I am not saying that these things are related. I just happen to be mentioning them both at once because I saw them both today. You get a lot of characters in one toot after all.

What?

Max Rees Mar 13

Danielle Foré Mar 13

LLM code is the lead pipe of our generation of software engineers. They’re going to be replacing it for decades and wondering how we were so foolish

Paco Hope Sep 5, 2025

One of the ways that LLM-authored code improves productivity is by merely SAYING it does things. It's way faster than the whole time-consuming process of actually doing things. This is real code someone sent to me for review.

Max Rees Mar 7

Cassandra is only carbon now Mar 7

"It writes boilerplate faster."

Is that all we aspire to now? Not abstraction or templating or better language design, just... output boilerplate faster.

It's depressing.

Max Rees Mar 6

e. hashman 🇵🇸Mar 6

Someone said something about ingesting libc and this is what popped into my head

Max Rees Mar 5

LLMs are expensive and impossible-to-secure malware vectors, part 4000: https://lwn.net/Articles/1061548/

A GitHub Issue Title Compromised 4,000 Developer Machines (grith.ai)

The grith.ai blog reports on an LLM prompt-injection vulnerability that led to 4,000 installati [...]

LWN.net

Max Rees Mar 5

Gabriele Svelto Mar 4

A few years ago I designed a way to detect bit-flips in Firefox crash reports and last year we deployed an actual memory tester that runs on user machines after the browser crashes. Today I was looking at the data that comes out of these tests and now I'm 100% positive that the heuristic is sound and a lot of the crashes we see are from users with bad memory or similarly flaky hardware. Here's a few numbers to give you an idea of how large the problem is. 🧵 1/5

Max Rees Feb 22

mhoye Feb 22

I’ve been thinking about this for days. Incredible stochastic algorithm, gets more reliable the larger your input, incredibly fast, trivial to implement and deterministic on its inputs. It really has so much going for it.

(Via @jonathankoren )

Max Rees Feb 20

James Thomson Feb 16

Writers: Generative AI models were built on our stolen works, are deeply unethical, and risk devaluing our entire profession.

Artists: Generative AI models were built on our stolen works, are deeply unethical, and risk devaluing our entire profession.

Developers: Wheeeeeeeeee!

Homepage	https://www.cs.purdue.edu/homes/mcrees
Pronouns	they/them
Interests	C programming, systems integration, build automation