the next frontier in compilers? ai-powered nasal demons it's not undefined behavior, it's aindefined behavior

if there's one thing I've learned in 25 years of playing and working with computers is that computers sometimes just don't work

"vibe coding is a new level of abstraction" only makes sense given a distinctly impoverished understanding of "abstraction"

I miss living by a real river :( closest thing I have is Strawberry Creek :P

AI aside, the idea that engineers aren't getting appropriately rewarded just because we can't slap a number on their "output" is a complete abrogation of technical leadership ...so totally on-brand for the modern tech industry :/ the solution isn't to a fake a legible metric, it's to not suck

met my parents' new neighbor

This is a very cool paper on zombie ideas about leadership that refuse to die, how to spot them and strategies to defeat them.

Another apposite Jill McDonald Puffin Post illustration

People who attended Monday's Papers in Systems discussion, you might like this, by Kelly Vaughn: www.modernleader.is/p/managing-s...

"What is the ROI of UX?" is shaped like a question, but it is actually two hostile premises: 1) that UX is sufficiently marginal for its value to be openly questioned 2) that the asker has the authority to judge this question, and practitioners are merely consulted on it, as a matter of courtesy

I haven't cooked in ages, then yesterday I made some thoroughly mediocre stir-fry so today I'm disproportionately self-satisfied

SQL is pronounced "squirrel" and WSL is pronounced "weasel"

Here is an example of what I mean in a simple illustration, from our New Developer project. AI Skill Threat is how acutely developers anticipated identity-based and skill-based loss and Contest Culture is how much they rated software as demanding punitive, ruthless competition norms.

I think we have to stop assuming that seeing like a state (synoptic legibility) is the only or even the correct way to see. To drag Adam Phillips into conversation with James C. Scott, we like legibility because it's a promise of omniscience, but (desire for) omniscience is psychic impoverishment.

aduhd which, for historical reasons, tends to be written as audhd

"Correction of Error" just screams "corporate struggle session" so perhaps it's totally on-message for Amazon :P

one thing adhd has given me is a baseline facility with—and exaggerated confidence in—public speaking with insufficient prep :P

It's hard to overstate just how much of a conformist environment corporate America is, particularly at publicly traded companies. Both shareholders, regulators (when they exist), and courts expect you to perform to "industry standards", as understood by e.g. Legal, Audit, Compliance.

Berkeley has some amazing little free libraries

legitimately brilliant use of pumpkins

the ai era has only made me dash ems harder—and I've been an unrepentant dash-addict since high school

hot take: "most technical problems are people problems" is not some fundamental truth, but rather a reflection of our dominant organizations and organizational culture

I will die on the hill that if "the median score of people who use a particular accommodation" is that much higher than "the median score of people who don't use a particular accommodation" it is a sign that the instrument you are using is measuring the wrong thing

(In fact if you see a major bump in score outcomes via accommodations what you are very likely seeing is a gatekeeping event, not a knowledge test.)

We cross our bridges when we come to them and burn them behind us, with nothing to show for our progress except a memory of the smell of smoke, and a presumption that once our eyes watered.

Being good at something and being good at building tools—libraries, DSLs, etc—for something are surprisingly different skills. A lot of teams need the latter but hire and incentivize the former.

Yknow, I never see anyone talk about this side effect of “personalization”—people are weird, and it turns out that it’s hard* to make massive amounts of advertising money catering to specific interests. So platforms have re-created a boring middle, shoved us all into it, and claim it’s “targeted.”

🎉 Unison 1.0 has landed! After years of engineering, design, and community collaboration, we’re excited to announce this milestone! Spread the word!

Assessing the quality of a random number generator is luckily quite simple. Every time it gives you a value, ask yourself: did I expect this to happen?

Gödel's incompleteness theorems prove that everything is made up and the points don't matter.

a joyful discovery today: prefix sum `sum[l,r]=sum[r]-sum[l]` can be generalized to arbitrary group, some of which are good-enough hashes which can be used in subsequence searching. this is exactly the rabin-karp algorithm, but online materials about it seldom mention group theory at all

A realization I've had: programming languages are as much tools managing complexity (ie design + abstraction) as they are for telling computers what to do—and mixing both together is critical for being good at the former. (And PLs are amazing at handling complexity!)

📣THREAD: It’s surprising to me that so many people were surprised to learn that Signal runs partly on AWS (something we can do because we use encryption to make sure no one but you–not AWS, not Signal, not anyone–can access your comms). It’s also concerning. 1/

On certainty of outputs: the Fordist instinct to demand "when will thing be Done" serves no purpose beyond dashboard/report curation, which has replaced actual management work in most orgs. It displaces all other value of the thing except for Done-ness. My Button talk on Thursday is about this.

I designed a cover for Samuel Delany's DHALGREN because the cover of the edition I'm reading is boring.

random bit of nostalgia: libraries here used to have TUIs for searching the catalog, and it definitely gave them some real gravitas it definitely impressed me as a middle schooler :P

New blog post: The illegible nature of software development talent surfingcomplexity.blog/2025/10/08/t...

this is one of those things where I can't tell whether it's so obvious nobody even bothers talking about it, or it's not obvious at all to most people :/

I was doing some software history research and stumbled on this absolutely FASCINATING letter from 1964: dl.acm.org/doi/10.1145/... Some random defense contractor writes in to say "You should deliver a minimal prototype as fast as possible to get feedback and involve users at every stage of labor"

Do I know anyone around Gdansk? I'll be hanging around the area for the next week or two, would be great to meet up for lunch/tea/etc

Empirically studying the effectiveness of static typing is incredibly difficult—and I actually find that reassuring. (The thread before the quoted post is also worth reading!)

#AmeriHac – The hackathon by the #Haskell Foundation in New York City has been announced for the 7th and 8th of February 2026! discourse.haskell.org/t/the-inaugu...

one of the things that finally got me to consider I have ADHD was catching myself painfully procrastinating on... starting a video game I was legitimately excited about >.<

performative hours ≠ excitement if folks were actually excited and motivated, you wouldn't need forced hours, you'd just trust people to work in the best way for them

programming languages aren't just tools for making computers do stuff but also remarkably good as tools for conceptual design fluidly mixing design and execution is part of what makes them so good at managing complexity because that lets us build hierarchies of concepts

my optimistic case for LLMs: the world today is held together with Excel because Excel, for all its faults, lets its user adapt the tool to their needs rather than adapting their needs to their tools LLMs aren't there today, but they might be the foundation for the next Excel

Sol Lewitt, ‘All ifs ands or buts connected by green lines’ (1973)

"Furthermore, although they may exist, we have not been able to find any studies showing that extended overtime (i.e., more than 50 hours of work per week for months on end) yielded higher total output in any field." cs.stanford.edu/people/erobe...

It’s always “AI is great for generating boilerplate code” and never “why do we even need boilerplate code, maybe programming is broken”

conflating "best" with "easiest to manage" is probably one of the worst "honest" management failings ("honest" as in not driven by internal politics or character issues)