Not just that, I've been seeing a huge effort in the Ada community to leverage LLMs to convert a lot of libraries into formally verified SPARK code. One of the biggest issues I see with vibe coded stuff is that it's difficult to review and difficult to prove that it's doing what you think it's doing, but with a strongly type language like Ada and formal verification with SPARK, LLM output is easy to read and easy to prove.
I live out of my RSS reader for everything from YouTube to Reddit, and in this case I leverage this feed for the HN front page:
https://hnrss.org/frontpage
The front page of HN has always been pretty good at filtering out AI generated articles (and if one makes it it's usually flagged by the time I get to it). Then I just glance through the list of articles that made it to the front page and decide which ones I want to read.
I reported a blatantly sexual adult ad on Facebook the other day, they got back to me and said that their moderators reviewed the ad and that it was appropriate.
> A few weeks ago, we conducted hour-long conversations with 10 of our users to dig deep into how you manage your preferences and configurations in Thunderbird desktop
Is only 10 people representative of the population of thunderbird users?
Fwiw, even just going through your software with one user can give quite a few insights about what's not obvious about it. That's not at all to say you never need more, but very few open source projects do user research in the first place, being passion projects that just scratch the developer's/s' itch. More samples is always better, definitely at n=10, but I'd also not dismiss the results and benefits of doing it!
Also, how many people you need depends heavily on whether the thing you're researching affects a lot of people. If settings has a problem that affects 40% of people, then a sample of 10 people would yield results representative of the whole population.
Sample size is a very weird, often kinda counter-intuitive topic.
The video goes into slightly more depth, and at about 1:30 into the video they acknowledge that the participants were not representative and that they would like to conduct further research.
> Is only 10 people representative of the population of thunderbird users?
For very narrow studies it is possible to get representative data with fewer than a dozen interviews, but in this case it is explicitly not representative. In the video they mention that most of the participants have used Thunderbird for over a decade and follow release notes, development, and various forums closely, which to me suggests that they were recruited opportunistically rather than a random statistical sampling.
They do mention that they have plans to engage a larger audience in the future but that can be incredibly expensive. Even large organizations typically have to augment a small number of representative interviews with a large number of surveys and a very large set of user telemetry to properly weight interview feedback.
It's a standard research technique. You can have 2,000 people answer an automated survey but you can't have hour-long conversations with them. Researchers in many fields would like a better solution for in-depth interviews.
I was just watching Tom Scott's latest video, he mentioned firing a trebuchet and the guide pointed this out that you don't "fire" a trebuchet since it doesn't involve gunpowder, you launch it.
Tom's commentary later was that he disagreed with that sentiment. "I disagree with those potential comments. Words change their meaning overtime. In modern English, you can fire an arrow, you can fire a torpedo, we were gonna fire that trebuchet"
In the US Navy at least, you don't "fire" a torpedo, you "shoot" it. The lore is that "fire" has a very specific, very urgent meaning on a warship and you don't use that word unless there actually is a fire.
There are lots of these hyper-specific "reserved keywords" in the military. Another one from the US Army is "repeat", which is the command for an artillery battery to fire again with the same parameters as their previous barrage. Therefore on the radio we only ever used "say again" to ask someone to repeat their last transmission. Even if no one on the radio had ever or would ever be involved in artillery operations, I imagine it's easier to just train the entire force on a single uniform standard.
There's a decent argument for using today's words even for settings that are no the current world. But for fantasy specifically I find it breaks immersion quite a bit. I'm not convinced by the "language changes over time" line of argument. If you take it to a bit more of an extreme it would also be fine to have characters say "cool" or even use some gen-z-ism etc, because hey words change meaning over time after all.
However, I'm also aware that I'm kind of a hypocrite because I'm totally fine with current world grammar and punctuation for example.
I suppose it’s just a matter of the audience and balancing legibility and plausibility depends on whom you’re writing for. For me, if Legolas was aura farming while Gimli rizzed up Galadriel while Saruman wants to hop on a palantir call that wouldn’t work but perhaps a day comes when that’s enjoyable for someone.
The market for these variations is controlled by IP law, not by demand for writing them or reading them. Emily Wilson’s controversial (for silly reasons, in my opinion) translation of Homer demonstrates what’s available in this genre.
Perhaps one day Gimli will rizz up Galadriel and Sam will say “bye-eeee” as Gollum falls into Mt Doom. One might even get to the stage where an LLM sits in place of the text and translates live into your own hyper-personalized idiolect or more usefully into that which will never take you out of the scene.
Years ago I looked into GrapheneOS, and I ultimately didn't go with it because, at least at the time, they only supported pixel phones (specifically pixel phones that Google actively had security updates for). I realized that if I got that OS, I would be at the mercy of Google supporting the device in order to continue using GrapheneOS.
In the end I just opted out of the android ecosystem altogether and went with a flip phone that I used as a hotspot for an iPod touch (we only used over VPN with locked down DNS and nothing google related).
My privacy lasted about two weeks, because unfortunately Spotify was able to fingerprint that device to Facebook.
I would highly recommend LineageOS. Supports non-Pixel phones, some of them many years old (although now that I double check, the Galaxy S3 and OnePlus One aren't on the devices list anymore, I guess they do stop supporting old stuff eventually...). The OnePlus 5 is still supported and was released in 2017. For anything older than that, I guess I'd recommend looking into postmarketOS.
I really appreciate it when services open source their software when they shut down. There's so many amazing sites, software, games, or other neat solitons that are gone forever because there's no way to access it or replicate it.
