sigh. followed someone for sensible anti-llm shit. interested to see them call out netbsd taking a sponsor relationship with a fascist org. and like the initial steps of that seemed very productive but then even from my (distant) perspective it looked like they brought forth an internet mob. then they complained about being accused of starting a harassment campaign. at that point i made the mistake of replying to push back on the boost velocity, as gently as i knew how to, and just got a block.

ok i mostly don't regret buying a ticket to Nathan For Theranos. first half's vibe really reminded me of the culty folks i was around in 2014. the "machine that can do anything" fallacy was almost boringly mundane. image came to mind of trying to sell a "table that can hold any book"- you could build a magic trick around that, but ultimately it's not an impressive or true fact.

there is nothing weirder than the reaction when you tell a FOSS nerd that programming harder and under worse circumstances won’t fix a social problem

when people say "ai is just a tool" or whatever i do basically agree with that much. above all, an llm is a tool for obliterating context. it is a new way, as power yearns for, to "know" something without knowing why or from where or how much you actually know it, displacing the thing you wanted in the process of crudely approximating it.

the year is 20XX and all data has been converted to hashed content addressing. every bit is in a random location somewhere in the solar system. storage metadata is 10,000 times larger than the data being stored. caches are obsolete. it's impossible to compute anything in less than a week. we are finally free.

the corporate vision of running a website: lots of blank space, three verbs in a row, watered down apple store aesthetic
the actuality of running a website: you now own 2 square feet of the sidewalk that *everyone* has been pissing on

smells like a git search engine... noticed a bot with a russian ip has been requesting just /HEAD and /info/refs from all my git repos

there's one page that just has hundreds of links to tiny .stl files and the bots are eating those up too

at the least, it's an accidental tarpit

$ find /var/www -name '*.html' | wc -l
130741

who called it a nut check when ballsacknowledgment was right there

ideal browser
- implements web standards
- nothing else
- leaves you alone
- some kind of non-profit and not google or nvidia
- rectangular

nothing good here

There's a dust-up over an AI ad from Bookshop.org. But they didn't use AI; Meta's ad platform automatically created an AI "variation" of the original, non-AI content.

I went to disable this crap in our Shiraki Press ad account, and got served a popup *demanding a reason why*.

Fuck Meta for this. Consent matters. We've just pulled all paid advertising from their platforms.

it looks like there was a database migration that left a whole new column set to nulls instead of calculating the correct values, and then a subsequent object-store-to-local migration fails on any video that has captions

this is only interesting if the bots are scraping text without running a browser but with something that has lenient memory limits on its decompressor. if the bot runs a whole browser there are easy ways to use lots of ram.

could you do a bot tarpit that uses up loads of memory by sending specially crafted compressed content? iirc gzip's worst case memory usage isn't that bad but the bots probably support zstd

circular reference! at the object relational mapper

very wet codebase

peertube seems to be designed for sites where everything is disposable. the opposite of an archival library.

wanted: a programming chat/community that:
- is on an open platform/network (Matrix or IRC preferred)
- is reasonably active, and generally friendly/supportive
- bans AI (both its use/promotion *and* dunking on / discourse about it)
- does not tolerate contempt culture / language bashing (particularly towards JS); this includes not tolerating shitty 'jokes'

boosts appreciated