That's the only ingredient list I have ever seen on canned beans as well, here in the US. I tend to make beans from dry just because I like to cook them with other vegetables and spices in order for beans to be at all palatable for me.
I wish journalists would run their examples like these by the researchers directly to see if the journalist's conception of the idea matches the researchers.
It is happening already, recently Brazilian woman living in Italy was scammed thinking she was having an online relationship with Brazilian tiktoker, the scammers created a fake profile and were sending her audio messages with the voice of said tiktoker cloned via AI. She sent the scammers a lot of money for the wedding but when she arrived in Brazil discovered the con.
I know that hn is heavily populated by people from the USA, that the author is dutch but a non-english language would be... every other language beside English.
Commenting on the actual text, his solution for the cedilla is awkward and is one of the first things I disable on any computer, because it is a extremely common letter in portuguese.
I agree that typing curly braces and square brackets on non English keyboards suck, but typing non English languages on an English keyboard sucks too. I made the opposite of your decision: I brought a laptop with my national keyboard and I switch to US layout when programming. That has had a curious effect on me: as my editor and my terminals have black backgrounds and everything else has a white one, something in my brain makes my fingers reach for keys according to the color of the background. I make many mistakes when I attempt to program in a white window (eg: type in gedit) or write in my language in a black one (eg: an md file in terminal.)
Hmm. Are you sure that your stack wouldn't accept these discovery packets until after you've successfully authenticated (which is what those chains are for) ?
Take eduroam, which is presumably the world's largest federated WiFi network. A random 20 year old studying Geology at Uni in Sydney, Australia will have eduroam configured on their devices, because duh, that's how WiFi works. But, that also works in Cambridge, England, or Paris, France or New York, USA or basically anywhere their peers would be because common
sense - why not have a single network?
But this means their device actively tries to connect to anything named "eduroam". Yes it is expecting to eventually connect to Sydney to authenticate, but meanwhile how sure are you that it ignores everything it gets from the network even these low-level discovery packets?
I may be missing something, but it is almost a guarantee that you would not receive a RA in this scenario? eduroam is using WPA2/WPA3 enterprise, so my understanding is that until you authenticate to the network you do not have L2 network access.
Additionally, eduroam uses certificate auth baked into the provisioning profile to ensure you are authenticating using your organizations IdP. (There are some interesting caveats to this statement that they discuss in https://datatracker.ietf.org/doc/html/rfc7593#section-7.1.1 and the mitigation is the usage of Private CAs for cert signing).
People keep reinventing LaTeX, but poorly. Most of the issues described have already been solved by it at least 20 years ago, especially the semantics part. The tooling is mature, well understood and supported on all operating systems.
As far as custom shortforms for fully tagged angle-bracket markup is concerned, people are reinventing SGML which can handle markdown and other custom syntaxes since 1986.
Markdown inline syntax is straightforward to capture using SGML SHORTREF. What's more difficult (impossible) are things such as reference links where a markdown processor is supposed to pull text (the title of a link) from wherever it's defined before or after its usage.
Haven't heard about archforms in a while ;) but it's not a technique for custom syntax, and since markdown is specified as a Wiki syntax with canonical mapping to HTML, there's no need for the kind of simplistic element and token renaming possible with archforms.
For example added an <nbsp> attribute to turn all spaces into non-breaking spaces, and used archforms to remove the attribute afterwards.
But yeah, maybe for Makrdown you don't need archforms. On the other hand, perhaps there is some super clever way to use archforms to get your reference links working.
This kind of cynicism is wild to me. Of course most AI products (and products in general) are for end users. Especially for a company like Google--they need to do everything they can to win the AI wars, and that means winning adoption for their AI models.
This is different. AI is an existential threat to Google. I've almost stopped using Google entirely since ChatGPT came out. Why search for a list of webpages which might have the answer to your question and then manually read them one at a time when I can instead just ask an AI to tell me the answer?
If Google doesn't adapt, they could easily be dead in a decade.
That's funny. I stopped using ChatGPT completely and use Gemini to search, because it actually integrates with Google nicely as opposed to ChatGPT which for some reason messes up sometimes (likely due to being blocked by websites while no one dares block Google's crawler lest they be wiped off the face of the internet), and for coding, it's Claude (and maybe now Gemini for that as well). I see no need to use any other LLMs these days. Sometimes I test out the open source ones like DeepSeek or Kimi but those are just as a curiosity.
If web-pages don't contain the answer, the AI likely won't either. But the AI will confidently tell me "the answer" anyway. I've had atrocious issues with wrong or straight up invented information that I must search up every single claim it makes on a website.
My primary workflow is asking AI questions vaguely to see if it successfully explains information I already know or starts to guess. My average context length of a chat is around 3 messages, since I create new chats with a rephrased version of the question to avoid the context poison. Asking three separate instances the same question in slightly different way regularly gives me 2 different answers.
This is still faster than my old approach of finding a dry ground source like a standards document, book, reference, or datasheet, and chewing through it for everything. Now I can sift through 50 secondary sources for the same information much faster because the AI gives me hunches and keywords to google. But I will not take a single claim for an AI seriously without a link to something that says the same thing.
Given how embracing AI is an imperative in tech companies, "a link to something" is likely to be a product of LLM-assisted writing itself. Entire concept of checking through the internet becomes more and more recursive with every passing moment.
reply