Bram Geron
banner
bram.xyz
Bram Geron
@bram.xyz
130 followers 240 following 620 posts
Rust, product. He/they. Nushell&jujutsu fanboi. Live in 🐻 Berlin Used to do research on programming languages. Standing with 🇵🇸 is standing with human rights. @bgeron on other networks.
Posts Media Videos Starter Packs
I think I’ll take it over politics
Reposted by Bram Geron
Proud of the Python Software Foundation for doing the right thing and standing up for their people, even when it’s hard. Especially when it’s hard.
TLDR; The PSF has made the decision to put our community and our shared diversity, equity, and inclusion values ahead of seeking $1.5M in new revenue. Please read and share. pyfound.blogspot.com/2025/10/NSF-...
🧵
The official home of the Python Programming Language
www.python.org
“Are you in favor of the constitution, or not?” Seems very simple the way he puts it.
In Deutschland wird Ordnung großgeschrieben, daher prüfen wir grundsätzlich alles, auch Parteien. Wenn eine Partei als rechtsextremer Verdachtsfall oder gesichert rechtsextrem eingestuft wird, sollte sie geprüft werden. Das ist der nächste logische Schritt.
youtu.be/eSgFEpI9ZDQ?...
Diese Strategie ist der Gamechanger: PRÜF!
YouTube video by Nico Semsrott
youtu.be
Protip: if you ordered by Vorkasse, your payment was only processed once status becomes In Preparation. Otherwise be sure to chase them. (My prepayment went unnoticed, causing a ~3 week delay.)
Congrats! I heard good things about it. Hope you get it quickly.
Very good stuff, at first glance, thanks for sharing!

It reads to me like a non-native speaker (with or without LLM assistance).
You improved the code
The real question is: will any customers leave them for being poorly available 😕🤷🏼
On top of that, the one 3-operand ADD lets you access 32 registers with 6 instruction bytes. With MOV+ADD that would be 4+4 = 8 bytes I think.

The APX announcement mentions more registers can help use less power — 10% fewer loads and >20% fewer stores.
Yes, but those 2 instructions would still be more load on the frontend.

CPUs already regularly run > 2 instructions per cycle, and I imagine decoding doesn’t parallelize very well at all.
That’s a good point, thanks 💚 not my strongest take.
this post brought to you by the YouTube app on Google TV
I wonder if it has something to do with the video game of the same name
In mathematics, yes, the set of functions without arguments into a set is equivalent to that set
(Maybe you meant a constant function which just returns a precomputed value, in which case the last question is moot)
If you call the function twice, do the output values share structure/pointers or not? This can be distinguished through OOM but is it something you want to care about
What’s pure: is malloc observable? Stack allocation? Both can crash a process. Is time observable? Changes in CPU cache or branch predictor? Output binary size / symbols?
As everyone found out: this really depends on the system you measure in and what you count

In Rust, the function could be much smaller or much larger than the value

In Rust it could be a function you can call only once, or many times, thread-safe or not. The output value could be clonable or not
Reposted by Bram Geron
people who think their LLM chats are basically a conversation make me feel like "oh, wow, you're fooled by the puppet bird"

the truth is there is an arm inside that bird and it belongs to a much stranger creature just out of your view
Thanks for the name. So I guess °N is an affine unit, and ° is its torsor.
By which I mean, a Haskell implementation *could* implement IO as such a tree. Seems that for efficiency’s sake, they do it a bit differently
The “nice” way to do this is a Tree enum where every node is either Char -> Tree or IoError.

And essentially, IO () _is_ this tree type. Except it contains more variants of course, and also for some reason they won’t let you inspect on it