Is that something anyone wants? It seems like being able to plug into other editors works well. What are the experiences people are trying to get but currently can't build because ACP sucks or whatever?
LaLiga isn't Cloudflare's customer. They have no relationship. So why would Cloudflare rework their infrastructure just to instrument rapid blocking at their own expense as a favor to LaLiga? And if they don't, ISPs just break the Internet for each soccer match? This is a kind of coercion that makes no sense. Cloudflare has no obligation like this to LaLiga (and neither would a Spanish domestic CDN!).
Cloudflare has not in fact refused to comply with any court orders! The very thing at issue is that LaLiga wants Cloudflare to do censorship on their behalf that Cloudflare, who has no contractual relationship with LaLiga, is not required to do by any legal framework in Spain or the US.
Cloudflare literally wasn't even a party to the ruling by which LaLiga has been compelling Spanish ISPs to do the IP-level blocking. They're just an affected third-party because the blocking scheme the courts have allowed LaLiga to impose on ISPs is on a per-IP basis.
Spain hasn't asked Cloudflare to do anything. Only LaLiga has acted like Cloudflare owes them a huge, expensive rework of their CDN's architecture for the purpose of censoring things for LaLiga purely as a favor to LaLiga. What LaLiga has over Cloudflare isn't a court order. It's a protection racket, or maybe a hostage situation, where court orders involving other parties are the gun held to the hostage's head.
> Cloudflare has not in fact refused to comply with any court orders!
Nor did I say they did.
The question was asked, "why would they [without an explicit order]" The answer is they probably shouldn't, but there's still an obvious incentive here.
I'm not sure why it shouldn't be cloudflare job to make sure they don't host illegal content. If my super market keeps distributing illegal goods, even if they remove it after a court order, they will end up having to close the whole market.
Either they should police the content they serve themselves or they accept the right holders to do it (which sucks for everyone).
Also they certainly willing take all their customers as hostage, as they could certainly split their network into legitimate customers and shaddy ones so the blocking is not so impactful, but I guess they prefer to make it as impactful as possible to be able to complain.
Anyone can report illegal content on Cloudflare and Cloudflare will remove it. The pirate streaming sites pop up only in or just before the first few moments of the game, and LaLiga insists they must be removed instantly in order to prevent their losses. So what they actually want is preemptive removal without meaningful human review or anything else that could take 10 minutes.
That involves more than being responsive when someone reports abusive content or dropping bad customers. That requires becoming a censorship machine that preemptively treats all new customers as criminals, and probably having some unaccountable AI drive the censorship process. (That latter seems to be what LaLiga is pushing Fastly to do.)
That's beyond the legal obligations of infrastructure platforms, bad for the reliability of their service, and just a slice of what they'd have to do to rework their architecture to support this kind of preemptive censorship.
> ” what they actually want is preemptive removal without meaningful human review or anything else that could take 10 minutes.”
Yet this would actually be a better solution for everyone (except the pirates).
10 minutes seems like a reasonable response time that would allow a chance for human review. No football fan wants to have their viewing interrupted because they used a dodgy pirate site to watch it. Currently, pirates can simply use a VPN to get around the IP-level block while the huge collateral damage affects legitimate Cloudflare users.
What he prides himself in (in this context) is craft, which LLM use probably can enable, but definitely isn't commoditized by the kind of vibe coding that Garry Tan is doing.
Pijul seems great and I really want it to flourish. I love reading the blog posts when there are major updates, and its approach seems fundamentally more right than Git's (as it should— while it gets its core theory elsewhere, its UX learns from Git pain).
What I'm waiting for before I switch is mainly a really good bridge to Git. At the first couple tech companies I worked at, I led efforts to switch from Subversion to Git, and I did it by experimenting and I eventually proving a viable path using Git's built-in, two-way SVN bridge, which is excellent.
Whenever I revisit Pijul, I play with importing a giant (pathological, tbh) Git repository. Unfortunately, it runs for days and has never successfully completed for me. But the moment that bridge is good enough, I'm in! (Performance is nice but reliability is obviously the essential thing, especially when it comes to one-time operations like the initial import.)
PS: pmeunier, if you're reading this: I love your work. When Pijul can reliably import Nixpkgs, write a killer blog post about it. I'm sure it'll blow up here on HN. :)
And thanks for your work to advance the state of the art in DVCS! :)
PPS: I wonder if Pijul's fundamentally better merging behavior could become especially salient for teams (or individuals) whose codebases are seeing a lot of churn with heavy LLM use, so that it can help make automatic resolution of merge queues cheaper or easier or more reliable.
There's a difference between a reliable hunch and really knowing something. What is obvious is not always (or even usually) easy to prove. And the process of proving the obvious sometimes turns up useful little surprises.
> Heh, lucky you, at least you get a message. My ISP just drops traffic to the affected IPs. No ping, no traceroute, just a spinner in the browser until it says "page not found".
This is generally how the GFW works in China. Instead of an overbearing nanny like a school or corporation's DNS blocker, you're left with a sense that you're on a version of the Internet that is just intermittently and somewhat mysteriously broken.
And indeed, in China, a lot of things that probably aren't fully intended to be blocked are not reliably accessible. Implementation varies, so you get strange routing and peering issues. It feels like an Internet that isn't fully formed, that hasn't finished coming together yet.
Nation states and corporations obviously gain some things sometimes by having Internet censorship/blocking frameworks in place. Maybe, sometimes, ordinary people even benefit, too, if it helps shut down illegal and genuinely harmful businesses.
But it feels like the whole world is gradually trending towards more and more Internet censorship without realizing that we are un-building a miraculous thing that took enormous effort and cleverness and expense to build. I wish we could think about this not only in terms of freedom (and we absolutely should think about it in terms of freedom), but how we are disintegrating the infrastructure of communication and computing.
This is generally how the GFW works in China. Instead of an overbearing nanny like a school or corporation's DNS blocker, you're left with a sense that you're on a version of the Internet that is just intermittently and somewhat mysteriously broken.
Oh boy, an excuse to share my favourite great firewall story on a visit to China. Keep in mind, this is 15 years old, so probably doesn't represent the current state of affairs. At the time, my daily news reading habit had me checking BBC and CBC (Canadian Broadcasting Corporation). The BBC site seemed to be working fine, but whenever I clicked on an article on CBC, it was blocked. A few minutes later, I went to show my wife that CBC articles were blocked, and I clicked on the same one again, and it loaded. I clicked on another: blocked. Tried it again after a few minutes and it loaded. Someone was screening the articles in real time for me. When I was done reading, I clicked on several of the weirdest headlines I could find, and after a few minutes, everything was blocked again including ones that had previously worked.
Your last paragraph: it is sad. But we had successful global networks before the Internet (the PSTN, telegraph) and we'll certainly have global networks after this at some point in human history. Perhaps in the the time between the Internet and what's next, the world will become a bit more mature about a few things.
there's a lot of evidence things do generally get better over time, though. Jon Haidt and his ilk... I forget all involved... have done research into it
obviously it can be bumpy and maybe there's a Great Filter or you happen to live during a bad period but life is certainly much longer and less brutal than it was for 99.9% of human history
> But we had successful global networks before the Internet (the PSTN, telegraph)
These were ripe with espionage, wiretapping and sabotage. Access to it used to be highly restricted as well, up until the 90s for example you were only allowed to connect government-licensed modems to the German PSTN directly.
> These were ripe with espionage, wiretapping and sabotage.
Just like today's Internet. BGP spoofing, CALEA, DDoS.
> Access to it used to be highly restricted as well ...
And this is where the regression or "downfall" is beginning. Access to the Internet (as in ability to send/receive arbitrary data to the wider Internet) is something I bet is going to be increasingly restricted, but most people won't notice because they don't understand the difference between apps and the Internet.
I'd be surprised if direct access to the Internet is possible for consumers in the next 10 years. Everything will have to be through approved apps (age assurance is going to be the catalyst) that work over registered tunnels contracted through ISPs, if there isn't an outright blurring or merger between the concepts of phone/CPE, ISP and CDN. Your non-tech layperson will not know any difference whatsoever if all they use are their phone plan, streaming/banking apps and Facebook.
Surely this was simply the nature of Deutsche Bundespost / Deutsche Telekom? Like, of course you had to use hardware they had approved to connect to their network.
This was the same in many places. The cost of hardware and connection time limited connections, and no one had cryptography except the government and ultra nerds.
We often look back on earlier stages in world history like we're somehow more advanced, or inherently smarter, than past societies. But one of the things made clear by the way this problem lines up perfectly with conflict during the industrial revolution (including the innovators flagrantly violating the law in order to win their advantage) is that for all our technological sophistication, we haven't really gotten better at the hard, human things: social coordination, planning, democracy. (Perhaps that's because we're still living under the same system that the industrial revolution finally birthed.)
For short texts, the translation I usually want the most is fast translation, and local models are actually great for this.
But for high-ish quality translations of substantive texts, you typically want a harness that's pretty different from Claude Code. You want a glossary of technical terms or special names, a structured summary of the wider context, a concise style guide, and you have to chop the text into chunks to ensure nothing is missed. Even with super long context models, if you ask them to translate much at once they just translate an initial portion of it and crap out.
Are you using it for localization or short strings of text in an app? I wonder what you can do to get better results out of smaller models. I'm confident there's a way.
Yea. I agree. In our case we are creating short news articles of max 3 or 4 paragraphs. The texts are translated in multiple passes into various languages.
We use a simple system prompt that instructs the llm to ensure simple authentic language output. With Opus we get seriously good results. The goal is not literal translation, but good translations. I tried hoiku for a while, but its not good in many languages. Sonnet is okaish, but not good enough.
reply