ori

Polished announcements and slick marketing have been anti-signals for about as long as I can remember. CC: @fribbledom@mastodon.social

I reject the term 'core developer'. I have developed zero cores. In fact, I usually throw out the core after I eat the rest of the apple. I have commit access, that's all. CC: @semitones@tiny.tilde.website @nina_kali_nina@tech.lgbt @csepp@merveilles.town

That works so much better than I expected it would. CC: @jonny@neuromatch.social

Those are metric inches. Obviously.

If it's one or two, and the remaining ones haven't been claimed yet, I can grab and ship. I just don't have room for a ton of them in my luggage. Will update tomorrow. CC: @d6@merveilles.town @angelwood@merveilles.town @rek@merveilles.town

Ah, yeah, 'thou', which is short for 'thou lilly-livered excuse for a unit of measurement'. CC: @tezoatlipoca@mas.to @neauoire@merveilles.town

So when are you up streaming this?

Checking Wikipedia https://en.wikipedia.org/wiki/IPv4#Header and https://en.wikipedia.org/wiki/Transmission_Control_Protocol ip[2:2] is the size of the IP packet.((ip[0]&0xf)<<2)) is the size of the IP header((tcp[12]&0xf0)>>2) is the start of data in the tcp packetsSo, this is filtering for tcp packets that have data (is, the TCP data offset is not equal to the end of the IP packet). This is a good example of inspecting packets directly, but definitely needs some additional explanation. Examples should probably not assume a fresh understanding of the protocol header structure. CC: @b0rk@social.jvns.ca

"Forbidden, spammer". I guess Debian doesn't like coffee shops in Asia.

It's public key cryptography, not a simple password. The private data is not sent over the wire, so it can't be stolem. I would like to have public key cryptography in use for authentication, though I have one major and one minor reservation with passkeys. The major reservation is remote attestation: the party that I'm authenticating with should have no idea what device I'm using, as long as it follows protocol. I don't feel comfortable using a protocol with provisions for this; large tech companies have shown that they're not to be trusted, especially where there's an opportunity for lock-in. The minor reservation is the difficulty of backup and recovery. (If the major reservation is resolved, I can run a non-standard passkey agent that backs up the way I want, so this is only an issue if someone has protocol support for constraining the devices I can use) For me, any doors for attestation are a deal breaker. Drop that from the spec, and I would advocate for passkeys without reservations. CC: @filippo@abyssdomain.expert @shaoyu@mastodon.social

At this point, if you're on Linux you're swimming in slop: https://lore.kernel.org/linux-media/CAHk-=wi4zC+Ze8e+p3tMv8TtG_80KzsZ1syL9anBtmEh5Z40vg@mail.gmail.com/

Makes sense -- Facebook must have seen the work he'd done to make Apple's quality match Facebook's standards, and thought "this is the person we want to lead our teams". CC: @khm@hj.9fs.net

But my point is that it's not embedded with our values. You can keep the technology, change the context, and the values "embedded" in it change. And that makes it more dangerous -- if you believe it was embedded with the creator's values, then it would be a great deal safer; it would be harder for someone adversarial to pick it up and run with it, and use it to harm people. Because it's neutral, you can't just consider our values, but the values of every possible person that could pick it up and run with it. CC: @OliviaVespera@spacey.space

The entire goal of the AI project is to replace thought -- it naturally follows that if you've bought in, 5 minutes of thought is too much. CC: @mntmn@mastodon.social @halfbyte@ruby.social

It's worse than that. Technology is neutral, people are adversarial. CC: @mntmn@mastodon.social @OliviaVespera@spacey.space

Made it to Boston for Systems Distributed 2026.

No, because the LLM becomes more efficient, and therefore gets the corp funding, and people will be generating reams of code that is only marginally understood. I don't mean harder in a relative sense, but in an absolute sense. There's a feedback loop here. It's very hard to use LLMs moderately for code generation.

Several years back: https://myrlang.org/summer-of-cod CC: @mjd@mathstodon.xyz

Yes, but the tech companies have the ability to pay the right experts to walk by them all day and provide explanations about how this is ok, really. CC: @pojntfx@mastodon.social @tdelmas@mamot.fr

For me, the essence of the neutrality is the wide range in how it can be used by society. To oversimpify, "if you're not careful, it can be used for anything, including harming the very things you were trying to protect. It doesn't have any care for the things you love built in -- that's the job of the people around it." CC: @OliviaVespera@spacey.space

Right. I saw that email, and I think I just forgot to respond. Anyways, yes -- it's going, but between paying bills, debugging issues that people using it on 9front run into, and just understanding what I'm doing in the VFS, it's slow. Maybe it's time to put up an onlyfans, and see if I can ditch paid work. CC: @coolbean@brain.worm.pink

I disagree with the example -- the atom bomb is just an overly excited source of cheap, clean energy. If you have nuclear energy you have the H-bomb. You take the helpful technology that leads to abundance for humanity, and remove a few safeguards. Now you have a weapon of incredible destructive power. For me, the question is what the social (and, technical) safeguards could look like. It's a genuinely hard question. I don't have a good answer. CC: @OliviaVespera@spacey.space

But what music? CC: @neauoire@merveilles.town

That's three times the management and almost no reasoning about the details. I wouldn't trust the assessments that come out of the process, and I won't use software that was written that way unless someone pays me to. I think I'm probably going to get pushed out of this industry soon.

Made it to EuroBSD.

I'm playing with 9front's locks. Our spinlock, as far as I can tell, should be laughably bad, but rough microbenchmarks are actually giving me reasonable results, partly because the locks are hilariously unfair. Does anyone know of any good, representative benchmark suites I should be stealing ideas from, or even better, porting?

Yes, it would be nice if we stopped building hell so people can roast a few marshmallows. Marshmallows are nice, but not that nice. CC: @pozorvlak@mathstodon.xyz

There's two XLs remaining. CC: @neauoire@merveilles.town @d6@merveilles.town @angelwood@merveilles.town @rek@merveilles.town

bourne> "How many people have actually written shell scripts here?"
entire room's hands go up
bourne> "I'm so sorry"

(NYCBUG, May 13 2026, https://youtu.be/Anr2UAs7ifY?t=3523)

I just started to re-read Brave New World. I,atypically, read the foreword too, and this sentence caught my attention: If I were to rewrite the book, I would offer the savage a third alternative. Between the utopian and primitive horns of his dillema would lie the possibility of sanity. There are many other sentences that caught my attention. I've linked a copy of a PDF with it: https://dn790000.ca.archive.org/0/items/BraveNewWorld-and-BraveNewWorldRevisited/Brave-New-World_-_Aldous-Huxley.pdf

Can you tell when I enabled the bot scraping protection on https://shithub.us ? I'm a bit proud of the approach. It's JS-free, stateless, and at least moderately annoying to handle on the scraper side.

Looks like my talk was accepted. https://events.eurobsdcon.org/2026/talk/NVMSCJ/

Ah, well, shit, I owe someone else an apology then. CC: @coolbean@brain.worm.pink

I pay for, and regularly test, the most recent Claude code. It can be coaxed to produce working code, but the process remains about as fun as sifting through turds for nuggets of corn. And that's the goal; the entire point of Claude is to make the process of writing code feel like micromanaging an idiot savant, but cheaper. As Anthropic releases new versions, the further you get from puzzle solving and the closer you get to management. The reason I write code is to solve puzzles, and the details are important to coming up with a good solution.

I voted 'first' because it's close enough; it's usually second after '-?' (-? is not be a valid argument in many programs, but most dump their usage on an invalid argument, and it's easier than typing --help)

If it's "yes AI" people will bail. The split is inevitable, the only question you get to answer is who you keep. CC: @deobald@fantastic.earth @pojntfx@mastodon.social @jorge@hachyderm.io

I've (in the distant past) taken multiple courses on the ethics of technology, so I've definitely been exposed to this viewpoint. I disagree with it. I admit, I haven't read the specific book you pointed at, but I would imagine that it mirrors Ursula Franklin, and to a lesser extent, Neil Postman. Rejecting the idea that technology has inherent ethical value isn't the same as absolving it of harm. Remove the people, and technology is just a very intricate pile of rocks and metal. Deciding that technology has inherent moral value outside of how people chose to use it makes it too easy to avoid reckoning with the social context of technology. Rejecting that forces you to reckon with the full interaction of technology and humanity. The harms and benefits technology cause are intricately intertwined with social structure, ideology, politics and human behavior. Accepting technology's neutrality, though, means that you need to reckon with the complexity of social interaction. You need to defend not just against the things you label as "bad", but also the things you label as "good". Moral neutrality is not the same as having no social impacts. And while some, er, ballistic impact, translates pretty directly to social impact, a huge amount of indirect impact is far more ambiguous. Even if we can use it for huge good, we need to be careful to defend against downsides. Trying to label technology outside of social structures is too dangerous for me to accept. CC: @mntmn@mastodon.social @OliviaVespera@spacey.space

Every time the LLMs get more effective, they make it harder for humans to meaningfully contribute.

I'm comfortable with management, but telling people what to do is the part of it that sucks. Being in a position to nurture the growth of their abilities is what's fun. Claude doesn't have that. It's a seeded deterministic function. How do you feel about trying to nurture the personal growth of an overly talkative calculator?

The problem is that people forget to ask "work well for who, exactly?" CC: @mwl@io.mwl.io

15.9 TB Huh -- if that's all of github, it's surprising to me that it fits on one large-ish consumer hard drive. CC: @concretedog@mastodon.social

'misprompting'? CC: @david_chisnall@infosec.exchange

A French person once told me: "You shouldn't found a bakery, it's just going to be a source of pain".

Yes. However, You didn't ask what they should do when they're clicked. CC: @neauoire@merveilles.town

Can confirm. CC: @mra@mathstodon.xyz @rek@merveilles.town

Why would it do that? Claude can already do a reasonable job of reasoning about code that I don't understand, and the Mythos vulnerabilities show it's only likely to get better.

There are far too many people who consider it a personal success. And, if you're relying on LLMs, you're stuck in that ramping up stage because it's someone else's code; your ability to understand the code is curtailed, but it's also very easy to fool yourself into thinking you understand the code. How long do you think it would take an engineer to ramp up on a codebase if they were mostly reading it, and relying on colleagues to edit? Would you trust their grasp on the architecture?

Yes. My point is that you can't assign a moral value to the technology in a vaccuum; the technology and the society are co-designed for the outcomes that you want. It all comes down the to people. The people in charge are one of the most important components defining the moral value of a technology. The technology itself can't be evaluated morally without the context it exists in. CC: @OliviaVespera@spacey.space

Again, every improvement makes the agents work better, and the situation for people who actually like working on the puzzles worse. I also haven't talked about the way that using agents heavily leaves you perpetually ramping up on "your" code, so that you can't really contribute meaningfully by hand without a huge deal of effort.