Show more

'It was embarrassing for Americans to welcome a notorious wanted war criminal on their territory. It was humiliating to watch an American president act like a happy puppy upon encountering the dictator of a much poorer, much less important state, treating him as a superior.'

14-day gift read

theatlantic.com/ideas/archive/

#politics #uspol

as a software entwicklerin i have assigned some dumb names to things, either in lack of better names or as a means to inject silliness into things that are too serious and unfun by default

but at least i never named a fish species "stripey"

If journalists were doing their jobs, they would call the deployment of National Guard troops to Washington, DC, from three deep-red states what it is: an invasion and occupation force.

wsvn.com/news/us-world/3-state

While I'm generally very pro-EU, this #ChatControl initiative they keep pushing - despite experts having persistently debunking the rationale behind it - is indefensible.

Instead, this initiative will create an extremely invasive surveillance state where everything you type, view and receive is automatically scanned & potentially (falsely) flagged for human review.

This would mean the end of privacy.

If you are an EU-citizen, please consider writing to your MEPs.

fightchatcontrol.eu/

I'd like to start manufacturing more LLM poison, but I'm a little bit sniped on where to get good source data.

Basically, I want to run Dissociated Press (markov-chain text generation) over some input sources, but I want to use sources that I can either a) find as public domain, or b) use under a combined license. For example, a bunch of code under the MIT license (the *same* MIT license) could all be churned up together.

I also want to ensure that the text's words and topics cover the space that LLMs will be called on to generate within. One approach would be to use English Wikipedia, for instance. Another would be to grab a bunch of Rust or Python source code that's all under the same license.

Those both involve more curation than I was hoping to do. Are there other sources I should be thinking of?

#LLM

If you want help collecting some Gaps in the Linux Phone App Ecosystem, please contribute to this pad:
https://pad.hacc.space/XP5opplBQ3mWEpAcv47oSg#

(It's part of this #FrOSCon #FOSSonMobile Talk: https://programm.froscon.org/froscon2025/talk/0a60d401-814a-473d-af0c-5d6ac5913065/ (which won't be recorded, sadly).

#LinuxMobile

Support our comrades in publishing. Get 20% off all Distro titles thru 9/1/25 with code DISTRO at checkout.
Browse the entire catalog here: pmpress.org/index.php?l=produc

While the new ChatGPT might not be 20X better than the old versions, it just might take 20X the power to use.

nakedcapitalism.com/2025/08/ho

Firefox, AI, PSA 

Wishing I was in DC so I could buy a drink for the Subway Slinger. He’s going to drink free for the rest of his life. The highlight of this week’s news.

“We’ve already seen AI systems effectively collude to raise rent prices and point the finger at innocent black men based on faulty facial recognition tech,” EFF’s Kit Walsh told @SFPublicPress. AI recreates patterns in its training data, including biases. sfpublicpress.org/california-p

This is great. It reminds me of some of the complex hinged 3D-printed objects that come out as a single print.

hackaday.com/2025/08/15/libera

Long thread/13 

Show thread

Newly released documents shed more light on previous efforts by Trump’s DOJ to undermine a key law protecting internet users’ speech. eff.org/deeplinks/2025/08/new-

Show more
Librem Social

Librem Social is an opt-in public network. Messages are shared under Creative Commons BY-SA 4.0 license terms. Policy.

Stay safe. Please abide by our code of conduct.

(Source code)

image/svg+xml Librem Chat image/svg+xml