Show newer

@eniko (is this re: the RMS thing, or did I miss something else?)

@5225225 Wouldn't bother with public-domain-trained LLMs because they consistently perform like crap from what I've seen (the whole 'tech' basically cannot work without mass nonconsensual scraping of data and conversations).

Perhaps something from the ELIZA era? Essentially just pattern matching and perhaps rudimentary knowledge graph processing, which can produce pretty acceptable results AFAIK.

"you cannot continue to victimize someone else just because you yourself were a victim once—there has to be a limit"

— Edward Said, a Palestinian-American academic and postcolonial scholar

🇵🇸

Today 2K Games and 31st Union join the growing list of organized workers in Spain 🔥

At a time when individualization prevails and we only hear talk of profit margins, losses, and layoffs, we believe that the only way to assert our interests as workers is by standing united.

#2K #31stUnion #gamedev #unionize

Show thread

it's kind of messed up to just use an ai transcription service and then leave it at that. you gotta at least go through and edit that thing meticulously. i'm tired of seeing videos put up with lazy captions. the hard of hearing deserve better and it's useful to everyone else too

As an adult I am embarrassed to say that I learned very late how to care for my teeth.

I wrote this short post to share what I have learned.

kokorobot.ca/site/keeping_teet

@jacksonchen666 I feel like this would both create interesting results (something perhaps philosophically closer to the "site of the day" lists that used to exist on the early internet?) and be relatively easy to run a custom variant of on a cheap server, for anyone with a few bucks to spare, moreso with an even smaller index!

Would probably need some unusual scraping logic though, like for example preferring following a random external link on a site instead of an internal link, if there are any, so that most sites would only have one or a few pages actually indexed

Idea: a web search engine that has an index of exactly 256GB. As soon as it's full, it starts throwing away old index data and replaces it with new items. The same query might result in completely different results the next day.

@whreq I imagine their index is a little bigger than 256GB though!

It's okay to be a bit broken sometimes. I believe we all are at various points in our lives. It's not a permanent state. We can work on healing if we want to.

It's not okay to be unaccountable though. When you won't take responsibility for your words and actions, that's when you do irreparable harm to those around you. That's when repair takes a back seat and people have to protect themselves from you.

Good morning. Today I got a very interesting response to a freedom of information request from Transport for Wales.

They pay almost £800,000 per year to an external contractor for "revenue protection" (which means enforcing ticket fines).

They earn ~£80,000 per year from ticket fines and penalty fares.

Which means they're paying ten times what they're making back from this process.

FOI response: whatdotheyknow.com/request/cos

Edit: petition in thread

#trains #transport #politics #wales #cymru

There are times when I can't believe how much effort it takes to argue for things like bike lanes and public transit, in the face of relentless news about climate disasters, record temperatures, and so on. The evidence is all right before us. *Some* of the actions we can take are very cheap. These choices should be completely obvious and undeniable. #Berkeley #ClimateEmergency

Show thread

Idea: a web search engine that has an index of exactly 256GB. As soon as it's full, it starts throwing away old index data and replaces it with new items. The same query might result in completely different results the next day.

Made my first piece of generative art (as in algorithmic art, not the NFT crap), with plain javascript and HTML canvas.

I forgot everything about trig and I really must re-learn it, this one was almost pure luck. But this is exciting stuff!

#drawtober - 14: trails

Heartbreaking: An interesting personal site you've found doesn't have a RSS feed, so you'll forget about it one day.

Show older
Pixietown

Small server part of the pixie.town infrastructure. Registration is closed.