Webshit Weekly
September 14, 2025
How Britain built some of the world’s safest roads
2025-09-08 | comments
Our World in Data has decided to rub it in by explaining exactly how Britain built some of the safest roads on the planet. For those of us living in countries where getting a driver’s license is practically a formality that comes with a cereal box, the British approach sounds like a form of psychological warfare. Between the hazard perception tests that require you to click on reflections in shop windows and the general culture of not treating your car like a sentient weapon, the results speak for themselves.
The Hacker News comment section has, as expected, turned into a support group for traumatized expats. American drivers are sharing horror stories about 5-minute road tests that consist of driving slowly in a straight line, while the Europeans are smugly pointing out that their roundabouts are actually a form of advanced geometry that saves lives. The takeaway is clear: safety isn’t an accident; it’s a grind. It requires a society that actually values the lives of pedestrians more than the convenience of “just one more lane” and a licensing system that doesn’t just assume anyone with a pulse is qualified to operate a two-ton death machine. If you want to survive your commute, your best bet might be to move to a country that doesn’t treat “driving while distracted” as a national pastime.
AI might yet follow the path of previous technological revolutions
2025-09-08 | comments
The Economist has released a bold prediction: AI might just be a “normal technology.” This is a direct attack on the venture capitalists currently praying that their $100B bet on chatbots results in more than just a slightly faster way to generate spreadsheet macros. Hacker News is naturally split: the skeptics are reveling in the “I told you so” energy, while the true believers are explaining how the singularity is definitely happening next Tuesday at 4 PM. In the meantime, the rest of us are still just using it to write apology emails we don’t mean.
Anthropic judge rejects $1.5B AI copyright settlement
2025-09-09 | comments
Anthropic’s attempt to pay its way out of a massive copyright headache with a $1.5 billion settlement has hit a major roadblock: a judge who actually read the paperwork. The deal, which was supposed to resolve claims from music publishers and authors, was blasted for being so vague it was “nowhere close to done.” It seems Anthropic thought they could just wave a billion-dollar check and have their past training-data sins forgiven without actually defining what “fair compensation” looks like in the age of generative scrapers.
The Hacker News crowd is predictably cynical about the whole thing. While some commenters are calculating their potential $1,000 windfalls for books that never earned out their advances, most are recognizing that this settlement was less about justice for creators and more about clearing the legal runway for Anthropic’s next funding round. The judge’s rejection is a rare moment of sanity in a process that usually favors companies with enough capital to simply “settle” their way out of legal questions about ownership and fair use. Anthropic claims to be the “safe” and “ethical” AI company, but it turns out that their idea of ethics is just another line item in a venture-backed budget that hasn’t quite balanced yet. If the judge has their way, the lawyers are going to be billed for a lot more hours before this “settlement” is anything more than a PR stunt.
Pontevedra, Spain declares its entire urban area a “reduced traffic zone”
2025-09-10 | comments
Pontevedra, Spain has declared its urban center a “reduced traffic zone,” which is European for “get your noisy metal cage away from my tapas.” While the residents are reportedly thrilled with the silence and the lack of diesel fumes, Hacker News is busy explaining why this would never work in a suburb of Phoenix where the nearest loaf of bread is a forty-minute drive away. The debate has devolved into the usual “just one more lane” vs “trains are the only way” theological war, while the people of Pontevedra presumably just enjoy being able to walk to a bar without getting killed by an SUV.
We can’t circumvent the work needed to train our minds
2025-09-10 | comments
There’s a new blog post going around warning us that using AI to remember things is making our brains turn into oatmeal. Hacker News, being a sanctuary for people who build the exact technology being criticized, mostly agrees — while simultaneously trying to sell you a coaching course on how to better organize your digital attic. The irony of using a “Second Brain” app to avoid brain-rot is lost on approximately zero of the commenters, yet the allure of a tool-based shortcut for basic cognition remains undefeated.
AirPods live translation blocked for EU users with EU Apple accounts
2025-09-11 | comments
Apple’s latest “Think Different” move is blocking live translation for anyone holding an EU passport. Depending on who you ask on Hacker News, this is either a cowardly surrender to the DMA or a brilliant piece of malicious compliance. Either way, the ‘universal language’ promise of AI has been successfully defeated by a combination of regional settings and lawyers. If you want to understand your Spanish waiter, you’ll just have to move to a country that doesn’t care about the privacy of its own citizens.
Ships are sailing with fake insurance from the Norwegian Ro Marine
2025-09-12 | comments
The Norwegian maritime industry has discovered a exciting new ‘innovation’: fake insurance. Ro Marine has managed to get over 100 ships sailing with policies that are about as legitimate as a three-dollar bill. Hacker News is marveling at the bureaucratic balls required to run a shadow insurance racket for oil tankers. It turns out that ‘trustless’ systems are a lot easier to build when you just lie to everyone about being insured. If a tanker spills in the North Sea, the “recovery plan” is presumably just a very loud shrug.
Proton Mail suspended journalist accounts at request of cybersecurity agency
2025-09-12 | comments
Proton Mail, the company that has built its entire brand on being a “Swiss fortress” for privacy, has once again demonstrated that even a fortress has a side door for government agencies. The Intercept reports that Proton suspended the accounts of journalists from the Russian outlet iStories simply because a cybersecurity agency asked them to. It wasn’t until a social media storm broke out on X that Proton suddenly found the “technical error” or “mistaken flag” needed to restore them.
The discussion on Hacker News is a predictable mix of cynical pragmatism and betrayal. Many users are pointing out the glaring gap between Proton’s marketing — which implies you’re safe from the global surveillance state — and its legal reality, which is that they’ll fold the moment a CERT sends an email. It’s a classic case of security theater: you get the end-to-end encryption for the content, but your account exists at the whim of whatever bureaucrat decides your reporting is “malicious activity.” If you’re using Proton for high-stakes journalism, you aren’t just trusting Swiss law; you’re trusting that Proton’s PR department stays more scared of Twitter than they are of government requests. Good luck with that.
‘Overworked, underpaid’ humans train Google’s AI
2025-09-13 | comments
Google, a company with enough cash to buy a small country and still have change for a fleet of private jets, is apparently struggling to pay the people actually building its AI. The Guardian has pulled back the curtain on the “ghost work” economy, revealing a miserable landscape of contractors who are expected to train the models that will eventually replace them — all for wages that would make a fast-food worker look like a high-net-worth individual.
Hacker News is, as usual, split between genuine horror and the kind of detached silicon-intellectualism that only develops when you’ve spent too long in a Herman Miller chair. Some are pointing out the obvious: why on earth would a subject matter expert train an LLM to do their own job for twenty bucks an hour? Others are busy arguing about whether this even qualifies as “work” or if it’s just a sophisticated form of Mechanical Turk. The reality is that the “magic” of Gemini is actually just tens of thousands of desperate people in high-cost-of-living areas being ground into dust by a bureaucracy that treats human feedback as a telemetry metric rather than a livelihood. It’s the digital equivalent of a sweatshop, but with more “synergy” and a much better cafeteria for the people who actually have badges.
Safe C++ proposal is not being continued
2025-09-13 | comments
The final nail in the coffin for “Safe C++” has been driven home, and to the surprise of exactly no one, it was the C++ committee itself holding the hammer. Sean Baxter’s ambitious proposal to actually fix the language’s fundamental memory safety issues has been unceremoniously dropped because the committee would rather spend the next decade bike-shedding over “profiles” that nobody will ever actually turn on.
The Hacker News consensus is a mixture of “I told you so” and a weary recognition that C++ is basically the software equivalent of a fifteen-minute bagpipe dirge—it doesn’t matter how much modern vinyl you press it on if the music itself is a disaster. The core issue isn’t technology; it’s culture. The committee views memory safety as a nice-to-have feature, like a slightly better way to format strings, while the rest of the world has realized it’s an existential requirement. As one commenter noted, anyone who actually cared about safety left the committee years ago for Rust. What’s left is a collection of legacy curators maintaining a museum of undefined behavior and pretending that a voluntary “best effort” at safety is the same thing as a guarantee. If you’re still waiting for C++ to become safe, you might as well wait for a shark to become a vegan.