too smart for my own good and i hate it

Truly coding is magic. So glad to have learned basic scripting when I was just 14 years old or so. How do most people even live without utilizing their computer's full potential? This thing can do anything. You just have to phrase it correctly.

People click around, drag-and-drop files out of one program into another; this feels like a remnant of the DOS era, where the system could barely handle one program running at once (and TSR programs seem to be what might have facilitated early multiprocessing, being able to wrest control flow from the running app away for a moment with an interrupt). Applications interacting with each other's data should be seamless, and information flow can and should be automated. Unix pipelines were a step in the right direction; unfortunately, it seems right at the same time or in the immediate future graphical interfaces got in vogue. And users, who were most likely coming from DOS, worked just as they did. Manually.

As far as I understand, even powerful macros or scripting features some pro-level software included were seldom used by anyone but power-users. And I'm willing to bet even some of these power-users might've not utilized the full potential of these features.

And as the software around us gets increasingly dumbed down, useless "AI" features (clearly made by people who fell for the "AI" moniker and do not understand what are language models, and what they can and can't do; and what do users actually need) get shoved into users' faces with no regard for actual utility, I yearn for a time where people actually learn to utilize the full potential of the computer.

One more hurdle in the way is also the restrictions on the devices most people seem to own. The usual thing to have in your pocket is a phone; however, it's not easy to write software for a phone. No scripting environment, arbitrary code execution from outside the application's files is banned by mobile app store gatekeepers across the phone ecosystem. (Yet somehow there's a carve-out for React Native apps, allowing them to self-update bypassing the store review, sometimes changing the entire app completely — talk about unfairness!) No ability to quickly set up shop and code — the device's SDK won't run on the device itself, precluding self-hosting and necessitating a second device for development. Oh, and also the store registration, which I would even compare to the biblical Antichrist (Revelation 13:16 anyone?), if not for a certain billionaire appropriating the Antichrist talk already. Apple charges for it, Google charges too, and recently Google wants to make it really hard to install apps from third parties unless you're registered with them.

I am way above an average user now. My understanding of the tools I use is far above, and that is exactly the problem. It clouds my vision — I am no longer capable of understanding the average consumer and what makes them tick. Therefore I don't even have the ability to answer my own question — how do we put this magic of computing in hands of the people, and teach people to actually use it? How do we build an environment which encourages people to tinker? And the last question, that I might actually have an answer for — how do we ensure this power to tinker is never taken away?

(The answer to the last one is, of course, radical libre software advocacy. The way Stallman did it. He had a point after all. Also printers are evil.)

I love the power of the browser console. Just managed to download a video from a weird web player that yt-dlp wasn't able to handle.

Now I wonder where the subtitles for it are coming from…

ok clearly my reply context parser is horribly broken

Oh so that's what this almost-disappearing picture is in all of your feed articles. It was a tracking pixel! I never even thought this is a thing you could put in RSS…

i feel like i haven't posted to here since forever

but contrary to how it may seem, I still have coherent thoughts that I may want to post later.

unfortunately i don't think my current Micropub client is a good place for anything but a short note; though I must note it was initially designed specifically for shorter-form content, so is it really the fault of my Micropub client?

maybe big posts deserve a bigger window after all.

why would i want to watch black mirror if looking out my window is free?

just found a typo in my old post of mine. gonna go commit seifuku or something

also if i see one more "Co-Authored-by: Claude" trailer i am going to spontaneously combust

i am of the opinion that co-authored-by is strictly for people and LLMs are certainly not people

i of course understand why Anthropic is doing this: this is PR for them (public relations not pull request). this is their way to quantify claude usage in major public projects. and also a way to show off: look how cool we are, our tool is so cool it's making open source contributions!

but for me it's a red flag that you're the kind of user who is so lazy to think you're not overriding that part of the prompt with something more reasonable like the commonly-proposed Assisted-by trailer (see: Fedora policy, post by xeiaso.net and probably more). you're giving up control and letting a random word generator take charge. you cannot accept responsibility for the actions of your model, because you didn't even slap your own name on it, you let fucking claude pretend it did the work. because guess what: that thing doesn't think no matter how much anthropic's papers want to delude you. it cannot think, it cannot feel, it is a machine and a machine should not be allowed to make decisions because a computer can never be held accountable.

just saw a real living person use claude opus 4.6 for a one-line fix. now that's absolute overkill and i think this person is very stupid for many many reasons

first of all who the hell has THAT much money to throw around... i guess anthropic's really subsidizing 'em subscriptions... i wonder what's the catch?

second of all if you're THAT lazy just pass it to haiku, it'll do the job equally well but use way less resources

third of all maybe you should just do it yourself, especially if you're addressing review comments. y'know. out of respect. because i wrote my comment with my hands and not a magic quill that twists words into different words.

lmao apparently i've been using a null pointer for stats and DOS has been just... fine with me writing to there??? what???

this is wild.

shit maybe i should do opengraph metadata for kittybox

opengraph is kinda stupid bc i'm duplicating data but i mean conventional social media IS stupid sooooooo not news

Status: experiencing a mental breakdown over the realization that Windows 95 will only enter Public Domain circa 2090, and unless Microsoft chooses to release the source code earlier like it was with older DOS versions, I won't live long enough to see it happen...

TIL:

GIT_SEQUENCE_EDITOR="sed -i s/pick/reword/" GIT_EDITOR="sed -i '<whatever>'" git rebase -i HEAD~$N

allows to programmatically clean up commit messages using `sed -i`… or anything else that could edit a file.

Very nice for linting commits after-the-fact if somebody (like past you!) did an oopsie that could be easily machine-fixable.

lmao is the LLM age finally making people actually write docs? I'm currently reading an llms-full.txt file of a library I'm playing around with, and ngl it may be better than the official docs with all the code examples and shit

TIL: nix develop $(nix eval --raw nixpkgs#pkgsCross.mingwW64.mkShell --apply 's: (s {}).drvPath')

The DLLs shipped are in $LINK_DLL_FOLDERS, so to test with Wine:

x86_64-w64-mingw-gcc ./hello.cpp
# Set the DLL search path; Wine expects semicolons as a separator
WINEPATH="${LINK_DLL_FOLDERS//:/;}" wine ./a.exe