Benedict Evans
Issue #597 of his newsletter
The story he’s referring to is from an Anthropic “study” that found that generative AIs will “lie, cheat and steal” to achieve their goals. Yes, because they were instructed to do so.
In 2012, artist Ben Grosser released a browser extension called Facebook Demetricator. Once installed, it hid all metrics from Facebook’s interface: likes, comments, notifications, unread messages, and so on.
“What’s going on here is that these quantifications of social connection play right into our (capitalism-inspired) innate desire for more,” he explained.
In creating his extension, Ben questioned why there were so many numbers “a system (and a corporation) that depends on its user’s continued free labor to produce the information that fills its databases.”
All of this in 2012!
More than a decade later, I feel we haven’t internalized Ben’s ahead-of-his-time discoveries. Even alternatives that position themselves as opposites to the abusive practices of commercial platforms like Facebook — think of Bluesky and Mastodon — insist on interfaces packed with numbers. It almost seems like we’ve lost the ability to imagine other models of digital interaction.
At the exhibition Indomitable Presences, currently showing at the Centro Cultural Banco do Brasil in Rio de Janeiro, Brazil, works by artist Mayara Ferrão are on display. Created using generative artificial intelligence, they emulate old photographs in order to “resignify the past”: indigenous and enslaved women kissing (example), scenes that probably occurred but of which we have no records for obvious reasons.
The Rio CCBB’s Instagram profile has been getting into arguments with some followers who are outraged by the promotion of art created with the help of AI. Even on profound topics that still lack answers from those who make a living finding these answers (philosophers, in general), @ccbbrj is taking a stance:
My instinctive reaction to the news that Windows 11 Notepad has gotten text formatting support was to reject it outright. What a blasphemy! After reading the article, however, it seemed quite interesting: the formatting is Markdown, you can toggle between formatted and plain text with a click, and most importantly, you can completely disable formatting in the app’s settings.
(This is yet another reason why it’s always good to read beyond the headline. Microsoft’s blog post announcing this feature, for example, doesn’t mention Markdown, which made me expect the worst.)
TextEdit, macOS’s notepad equivalent, offers rich formatting (*.rtf format). It’s horrible. I think I only use it when I open the app for the first time after reinstalling the system or when setting up a new computer. My first move is always to switch the default to plain text in the settings.
That said, I would love for macOS TextEdit to have native Markdown support, even if it were just syntax highlighting — that is, without rendering the formatting.
Back to Windows Notepad, I learned that the version Microsoft has been updating with cool features (Markdown) and questionable ones (Copilot/AI) over the past three years is actually a whole new app. And that the old app — the one that was abandoned by Microsoft for over two decades — remains accessible at C:\windows\system32\notepad.exe. And if the new one is uninstalled, the old one automatically becomes the default. It’s good to have a backup when major changes hit previously reliable software.
(At least that’s what this commenter on Ars Technica says. I don’t have a Windows PC to verify this information.)
One of the latest generative AI-motivated blunders, the recommendation of non-existent books in a “special supplement” of US newspapers Chicago Sun-Times and Philadelphia Inquirer, generated yet another wave of criticism of the technology.
Dan Sinker defined the moment as “the who cares era”:
The writer didn’t care. The supplement’s editors didn’t care. The biz people on both sides of the sale of the supplement didn’t care. The production people didn’t care. And, the fact that it took two days for anyone to discover this epic fuckup in print means that, ultimately, the reader didn’t care either.
It’s so emblematic of the moment we’re in, the Who Cares Era, where completely disposable things are shoddily produced for people to mostly ignore.
Dan focuses on AI, but I have to say that the problem runs deeper and predates it. Supplements of this kind already existed before, and while slip-ups of this nature were rare, the fact that this one took two days to be noticed implies that on the reader’s side, nobody cares — yes, and they don’t care for a long time, well before the popularization of generative AI.
I find myself wondering how much stuff has already been printed not to be read, or at most, to be read and ignored. Or, in the digital realm, how much content isn’t published to be read and spark action or make people think, but rather to fill space, capture attention to redirect it toward ads or similar things.
Rob Horning raised this argument more thoroughly and elegantly, as he often does:
The fact that LLMs can generate endless amounts of explicitly “fake” copy with the traces of human intention and presence deeply diluted through countless layers of processing and concatenation could hopefully demystify not only that particular subject position that seeks safe harbor in “real texts” — i.e. an alibi in a “real supplement” for the dubious pleasures such supplements have always supplied — but also the fantasy of accessing perfect authenticity through media.
Between Meta announcing that its AI, Meta AI, reached 1 billion users and Google saying that AI Overviews are used by 1.5 billion, I’m curious to know how many of these people intentionally use the feature, or prefer it to what the AI replaces.
AI Overviews appear at the top of searches, with no option to turn them off. Meta AI, I suspect many people trigger accidentally by tapping that horrible button in WhatsApp, in search results across its three core apps, or when trying to tag someone in a group by typing an @ symbol.
It’s very easy to reach enormous numbers when you already have a giant platform. I don’t think that’s even part of the discussion. The issue is trumpeting these numbers as if they were earned, rather than imposed.
I’m in Salvador (BA) participating in the 15th Internet Forum in Brazil, FIB15. I came to present a new interview podcast (in pt_BR), Nós da Internet, and to fix a personal flaw: never having participated in a FIB before.
Here, I had the privilege of interviewing people who built and continue to build the Brazilian internet. And in a big fashion: in a beautiful aquarium-studio set up in the middle of the convention center. The one in the picture above.
Just a QR Code is a simple, straightforward QR code generator without ads or invasive trackers.
Just a QR Code was born from Gabe Schuyler’s dissatisfaction with online generators of this type. “Isn’t it possible to just make a one-page website that uses Javascript to generate QR codes? Something I could save to disk and run locally”, he pondered.
And from that, Just a QR Code was born. Gabe himself was committed to cover the operating costs. In exchange, he asks:
If you find it valuable, you can pay it back by creating your own useful thing for the world and releasing it for free. Let’s take back the friendly web, one vexingly-monetized utility at a time!
It’s this spirit that drives PC do Manual, a host of FOSS apps from Manual do Usuário. Which, by the way, has two QR code generation tools, a general one and another for joining Wi-Fi networks.
Mozilla has announced the shutdown of Pocket, one of the pioneering “read later” services.
Starting July 8th, Pocket will no longer allow saving content, essentially going into read-only mode. Data can be exported until October 8th, 2025. After that date, it will be deleted.
According to the company, “the way people use the web has evolved,” which justifies redirecting resources into “projects that better match their browsing habits and online needs.”
Bill Gates, Microsoft co-founder, philanthropist, and one of the world’s richest people, has released an autobiography. It’s another step in his long-running campaign to distance himself from the image of the ruthless businessman of the 1990s, the one who was seen as a symbol of capitalism and, therefore, deserving of pies in the face.
Source code: My beginnings is the first of a trilogy that Gates promises to release in the coming years. It covers his childhood in Seattle, through his school and university period, to the early years of Microsoft in Albuquerque, New Mexico, before Windows, when the company made its living selling versions of a Basic language interpreter for the handful of computer architectures that were popping up at the time.
Bill Gates’ story, at least as he tells it in this book, would make for a pleasant TV series, a coming of age set in a typical middle-class American suburb in the 1980s. Like a Stranger Things, but without the supernatural part…? Or, in a less popular but more accurate comparison (even in name), a Freaks and Geeks with more emphasis on the “geeks.”
Don’t be mistaken, this is a compliment to the narrative. It’s a really nice book!
