Lady @Lady@cat.family

the way things are going tho we’re more likely to see h·t·m·l adopted as a lingua franca for documents and tools than we are to see an x·m·l parser with wide·spread adoption which is actually good, to say nothing of making reasonable metadata queries of the parsed result

putting licence information in an easily⹀recognizable and ⹀parsable plaintext comment at or near the beginning of a file is sensible and very unixy, but it still irks me a bit when the file in question is x·m·l

that could be structured metadata!

@aschmitz @Satsuma oh to be clear i think that people SHOULDN’T try to get a computer to generate a language, but there are lots of people who want to make real-sounding fake sentences quickly without doing all the work that Building A Proper Language From Scratch might take

and “make real-sounding fake [thing] quickly without work” seems like the sort of problem people love throwing LLMs at, so

large language models aren’t capable of devising and consistently applying a novel grammar OR phonological system but a lot of people probably simply don’t care

there are existing tools out there which, for example, you can put some phonological rules into and have it spit out random words which follow those rules, and i still don’t like them but at least their output will be consistent

are the conlangers using LLMs to generate new words for their conlangs now or have we not quite gotten to that point yet

kind of girl who is so enamoured with her girlfriend that she cosplays as her

it is Very Important that i let my partner sleep

but i want cuddles 🥺

this is true of all queerness, btw

include hot sex tips in all your published code

my point is that i’m not actually that worried about A·I gobbling up my posts because i’m pretty sure anyone who tries to use it will consider my influence on the model to be the noise, not the signal

they need massive amounts of data in order to normalize and sanitize out precisely all the weird shit that gets produced by freaks like me

weird esoteric shit in, not what i was expecting the computer to say out

the thing about having niche interests and a defined but constantly evolving ideolect is there’s a good chance that indexing your content will make indiscriminate large language models worse, not better

this is par for the course for me when it comes to chastity belt

tried listening to the new chastity belt and it just reminded me of a better album by a better band so i went and listened to that instead

@aschmitz oh i mean i would never give access to anyone i didn’t trust OR put anything sensitive on the server ever OR have any kind of uptime or reliability guarantees, if i were to ever do this, to be clear

for my own purposes, i just run make locally and then rsync. but i think that’s a tiny bit too unstructured to be a shared workflow

i think proper CI/CD pipelines with gitea/forgejo/gitlab/github etc are Fine, but wasteful, and also best to do on their infrastructure and not on a friend’s

i have some good makefiles you can use; this gets you a lot

i think what i actually want is :⁠—

• A GitWeb instance

• A number of users which only have git access, and only to repositories they own

• A straightforward means of getting new repositories (and corresponding subdomains) allocated to users [this can probably just be manual, on request]

• A post-receive hook in every repository which, if the repository has a file named Makefile or GNUmakefile, clones the repository and then runs `make install DESTDIR=public`, serving the contents of the resulting `public` directory from a corresponding subdomain

• Maybe a small amount of allowable additional server configurations besides

@aescling you should start a tilde town

i don’t want to operate a tilde town, but i do think all of my friends should have a server they can rsync files to and have them appear on the web