Hackers' Pub is a place for software engineers to share their knowledge and experience with each other. It's also an ActivityPub-enabled social network, so you can follow your favorite hackers in the fediverse and get their latest posts in your feed.
What an odd (and bad, in my rather uninformed opinion) decision by the Delaware Supreme Court.
Please, people. Stop buying #Tesla. This company is not only broken, it’s literally financing Elon’s evil deeds. The man has starved 600,000 people through DOGE’s destruction of USAID. He’s a white supremacist. He’s a Hitler sympathizer. He’s polluting the sky with junk satellites. He’s poisoning poor people with his data center emissions. What more does he have to do for you to stop sending him your money?
anyway just to be clear I'm not overly negative on rust or anything, that's not why i'm really worried about replies, just that it's a good opportunity for a general "year in review" kind of thing and well, that may get more spicy
I feel like I flubbed an interview because, while I was describing my work with xpaths, the interviewer caught me off guard by asking how I would implement a simple xpath engine. I was caught off guard and answered "I think that's a mistake. I've seen a lot of half-assed xpath implementations. I would rely on the engine optimized by very smart people, as if it were a compiler." Really the interviewer was asking "Do you know about trees and graphs?"
Honestly, if there's anything you want in a senior engineer, it's probably the ability to say "You're about to try something that seems smart but I can tell you from experience that it isn't.'
I feel like I flubbed an interview because, while I was describing my work with xpaths, the interviewer caught me off guard by asking how I would implement a simple xpath engine. I was caught off guard and answered "I think that's a mistake. I've seen a lot of half-assed xpath implementations. I would rely on the engine optimized by very smart people, as if it were a compiler." Really the interviewer was asking "Do you know about trees and graphs?"
My name is Dr. Bernadette “bird” Bowen and I was fired for telling the truth about Charlie Kirk in September. If someone could hire me before I potentially become homeless in a month or two, that would be great LOL
I've tried a few times in the past to build Python bindings for a C library and always gave up pretty quickly. This time I was looking to make Python bindings for https://github.com/fastserial/lite3.
Antigravity + Gemini has gotten everything working over the past hour or so. It's been 20 years since I've done any C, so not sure I'd call any of this production-level code, but it's been fun to play around with it at least.
If you're on a Mastodon server that needs to have the highest level of user safety, you might want to talk to your admin about whether they should activate "limited federation mode".
This mode prioritises safety over connectivity, and means a server manually chooses which servers it allows federation with, while blocking all the rest by default. More info at:
First I did use Opus 4.5 and got nice library and perf results for my CSS parser (compared to ExCSS) but then I run review and optimisations via Codex 5.2 and code is now just so much better perf and quality wise, this is purely vibe coded.
tomorrow is thirteen years since I started with #rustlang
i kinda want to write a blog post but i also kind of don't
mostly because i wouldn't really have time to engage with responses.
we'll see...
The other day I learnt that enabling syncookies in pf on OpenBSD lights up your server like a Christmas tree. Or in other words, it made the server respond to _every_ incoming tcp syn packet with syn/ack, making the bots scanning the internet think the port was open, even though the final ack would be dropped when rules were then evaluated. This increased scanning by a lot! Or so it seemed to me at least.
The reason for this experimentation was I got to experience what I think was a tiny syn-flood attack, and I was trying things to see what I could do to mitigate on my end.
I'll maybe write more about this after the holidays.
Happy #Caturday! Here’s the obligatory “how it started…” post for our new housemate, who is doing quite well as a now-inside cat. But it was a stormy Halloween when he came to the door.
Jessica's playing some Kylie Minogue this morning, and really struck me how much this track feels like a Fantastic Plastic Machine song: https://www.youtube.com/watch?v=FR6C6kNuHKM
Trying to prevent spying is only part of it. We should also try to deal with (potential) manipulation and other negative-sum behavior.
Having open-weight models implemented as open-source software is part of it — but, we also need open-data — the data the models were trained off of needs to be available, too.
I think people who care about PRIVACY should pay a lot of attention to LLMs!
The ability to (both intentionally and unintentionally) spy on, manipulate, and engage in other zero-sum (and even negative-sum) behavior against people will be at a level never seen before — because of how those who want to do those things could (and likely will) use LLMs.
I don't think LLMs are going to go away — not in the way I suspect some of its hater wish it would.
Even after the AI hype-bubble pops — I think LLMs will still be around.
It doesn't matter if you hate them, or hate how they were created, or hate their social impact, or whatever — I think people will continue to use LLMs — both now and in the future.
I think people who care about PRIVACY should pay a lot of attention to LLMs!
The ability to (both intentionally and unintentionally) spy on, manipulate, and engage in other zero-sum (and even negative-sum) behavior against people will be at a level never seen before — because of how those who want to do those things could (and likely will) use LLMs.