I like that there's a list of primary sources at the bottom.
Kagi's AI assistant has been satisfying compared to Claude and ChatGPT, both of which insisted on having a personality no matter what my instructions said. Trying to do well-sourced research always pissed me off. With Kagi it gives me a summary of sources it's found and that's it!
Great little to-the-point writeup. Thanks! This is helpful for me choosing my own small-scale stack, which I've been having trouble with lately.
One major difference is that I'm currently hosting a bunch of very small apps off the same hetzner box just by myself, and I'm using Docker Compose instead of something like Hatchbox (or another bring-your-own-PaaS like Coolify or Dokploy, which are both less specific than Hatchbox). I'm not running any production-level apps, though; maybe Hatchbox would provide better stability/teamwork?
I've switched to a flatpak-based immutable distro lately. It's great when it works. But so many niceties don't work, and then it feels like my computer is not really the fantastic tool it should be. For example:
- I had to run around with a distrobox running WINE and a bunch of permissions and kludges to run an external tool for Godot
- I gave up on the flatpak for Firefox because it can't talk to my KeepassDX flatpak
- The Godot and Krita flatpaks are oddly unstable and crash more than they did on Windows (may just be Gnome or something)
- non-flatpak tools like AppImages and .rpms feel pretty dang grungy
I want to see more cool stuff with Flatpak so seeing the state it's in is kind of a bummer.
I don't know anything about compliers, so this note about reasons and intentions from the presentation link was helpful to me:
> YJIT can make Ruby code run faster, but this is a balancing act, because the JIT compiler itself must consume both memory and CPU cycles to compile and optimize your code while it is running. Furthermore, in large-scale production environments such as those of GitHub, Shopify and Stripe, we end up in a situation where YJIT is compiling the same code over and over again on a very large number of servers, which seems very inefficient.
> In this presentation, we will go over the design of ZJIT, a next generation Ruby JIT which aims to save and reuse compiled code between executions. We hope that this will help us eliminate duplicated work while also allowing the compiler to spend more time optimizing code so that we can get better performance.
Seems pretty cool. I haven't run into any these limitations in my own usage really, but I'm working at very small scales.
This is really surprising to me. Wouldn't this make it a good deal more expensive to produce most kinds of apparel and some merchandise? Maybe I'm out of date on modern printing and embroidery tech.
Oh wow! It took me a bit to realize this was DHH of Ruby on Rails fame (among other things). I like this message. After a career in Java, bash, Python, PHP, and a half-dozen config-definition languages, I've never found anything that evokes the joy-of-actually-writing-code like Ruby does.
Well, other than few tiny niggles like my code editor not being sure what kind of variable something is. But usually that doesn't matter.
I read manga extensively on my Kobo Forma with koreader. I wrote a script with imagemagick to scale, trim, adjust contrast, map to 16 colors, dither, and repack, all without me having to interact with it... something I'm hoping to open-source sometime, although it's very specific to my use case.
Very cool! You've quite nicely rendered some of Aseprite's attractive window chrome.
I draw "pixel art" in Krita with my "Flipnot Brushkit" brushes [1], which I prefer for Homestuck-style low-res painting where traditional pixel art programs kind of fail. Krita has a lot of nice power features for transforms and layering, like the batch_export plugin for exporting tons of layers at once.
I observe that the letter of the law intends to severely restrict pornographic materials from being made available to minors: see "harmful content" [1] and obscenity [2] definitions for the state of Texas.
Realistically however, fringe content and the risk of misinterpretations of the law place a huge burden on the distributors of any kind of media, where they could be at risk for a suit at any time.
This seems to point towards the need for age-checks at the book store, just to quash all possiblities---which is absurd since community-sourced content like tiktok is already so much worse for them, and that's not half as restricted!
Rather few of the stereotypical hacker crowd may be ultimately affected by this---unless Mr. Friede injects himself with some trojans and worms as well! (Forgive me)
Very cool. I'm one of the few (it seems) who likes these styles of chart, so that was a nice surprise.
I love Ruby, and one of my few qualms about using it is that it doesn't really have any cross-platform GUI libraries. Someday I'll try building one in JRuby...
"Crown of Ash and Honor," the sample book linked on the "Start Writing" page [1], seems like a counterproductive example. With "novel-length drafts" being the main selling point for the app, this draft seems acceptable for the prose itself (it's on par with most AI writing) but doesn't really demonstrate novel-length strength.
The story begins with a boy yearning to be a knight in a war-torn fantasy world apparently without magic, then there are extra-terrestrial invaders hunting a technological artifact and he suddenly knows how to fix a shield generator by crossing wires, then it suddenly turns again into a dark fantasy story with corrupted zombie-ish warriors.
On the technical front, clicking on a link in Section 2 of the story's ToC takes me to that number in Section 2.
Great, original article. I didn't notice at first that this blogger is the very same author behind Blue95: https://github.com/winblues/blue95
I used to love theming my desktop environment, but the joy faded when I realized the UI felt much more magical than anything I was using it for. Wonderful application of the tech, though.
The app looks and feels great. Very smooth, I can see you were considerate of the zen experience. It's very easy to misread the name as -isv.
Small ux bug: when starting a new chat, the title field gets focus, but if I start typing before the UUID slug in the URL changes, it'll cut out a chunk of my input.
Also very similar to a lot of Obsidian plugins I've seen going around recently---that will be one of your competitors.
"The Laundry Files" is one of Charles Stross's (the author of this article) fun series, although it's more a kind of urban (bureaucratic) fantasy than sci-fi.
I'm surprised he doesn't touch on AI at all in this article. To me, AI companies heavily depend on the implications that the "AI" moniker (as well as "learning", "neural network," etc.) carries from many years of human-like reasoning machines appearing in science fiction, and I think there would be much less momentum if they were called something else.
BR confuses me because when I look at blocks of BR-enabled text, I feel less mental strain. Yet the studies and counter-studies (like this article) I've seen only report on comprehension scores. It definitely feels like BR is in scam territory, but I'm really curious about what it is about it that catches me.
Maybe it's just because bold text would have the same effect; or because I usually skim while reading anyways.
Kagi's AI assistant has been satisfying compared to Claude and ChatGPT, both of which insisted on having a personality no matter what my instructions said. Trying to do well-sourced research always pissed me off. With Kagi it gives me a summary of sources it's found and that's it!