Mastodon Feed: Posts

Boosted by soatok@furry.engineer ("Soatok Dreamseeker"):
raigho@huskies.online ("Raigho Husky") wrote:

What?! I'm not cute! 😳🧡

Flashback #fursuitfriday

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
rysiek@mstdn.social ("Michał "rysiek" Woźniak · 🇺🇦") wrote:

Oh and a reminder that the whole "wow Mythos is such much special at finding vulns amaze" shtick is largely just Anthropic's hype.

https://aisle.com/blog/ai-cybersecurity-after-mythos-the-jagged-frontier

> We tested Anthropic Mythos's showcase vulnerabilities on small, cheap, open-weights models. They recovered much of the same analysis. AI cybersecurity capability is very jagged: it doesn't scale smoothly with model size, and the moat is the system into which deep security expertise is built, not the model itself.

#AI #Mythos #InfoSec

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
dsalo@digipres.club ("Dorothea Salo") wrote:

I've seen questions about the #instructure #canvas leak, to the effect of "why's the data so dangerous if there's nothing financial in it?"

There are identifiers, for one thing.

But I think the real threat is... look, students send instructors some really heavy and confidential stuff in Canvas messaging sometimes. Health stuff. Family stuff. Relationship stuff. Occasionally even crime stuff.

I really am relieved that I direct students to email. I feel horrible for what could happen.

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
inthehands@hachyderm.io ("Paul Cantrell") wrote:

RE: https://climatejustice.social/@ketan/116534182584305742

LLMs in their current form are an environmental disaster, regardless of whether the vendors who sell them are nice people.

db@social.lol ("David Bushell 🪿") wrote:

i'm so going to set-and-forget this and then a year from now spend an entire day debugging

https://csswizardry.com/2026/05/better-browser-caching-with-no-vary-search/

baldur@toot.cafe ("Baldur Bjarnason") wrote:

So I now regularly get this error message on my FIRST click on some repos in GitHub, usually when I click to pages that are likely to be resource-incentive like commit history. Is it possible that some GitHub views might be so badly made that for some repos they just immediately trigger a “too many requests” error even for a fresh session?

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
algernon@come-from.mad-scientist.club ("Luddicus Mus") wrote:

AI slop is killing online communities

Most of the communities I'm in are doing just fine, with healthy discussions, much cheer and joy (and an occassional spiral into doom and gloom, as is the nature of modern humanity). There's no AI slop in there, and it ain't killing them in any shape or form.

The trick? These communities banned AI. Not just AI slop, everything LLMs and what is today known as "AI". They took a hard stance, and it paid off.

Now, the author of this article sees their communities struggle - because like the author, those communities did not take a hard stand. On the contrary! They embraced "AI"! And with that embrace comes the slop.

Want to keep AI slop out of your communities? Enact a hard anti-LLM policy for code and for everything else too, and enforce it. It's that simple.

#algernonReviewsHackerNews

dysfun@treehouse.systems ("gaytabase") wrote:

you see, i foolishly decided the path of least resistance would be their binaries, since they even publish musl binaries which will work nicely inside alpine images. the archive contains one file, it's called opencode. i am not sure what it expects to do.

we will now attempt to build it ourselves when there are no build instructions anywhere because why would you when you can just use the binaries?

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
stevendotjs@hachyderm.io ("Steven 🏳️‍🌈") wrote:

I wrote a love letter to Peter Welch's "Programming Sucks" essay, about the state of the tech industry in the age of AI. Hope you'll give it a read & tell me what you think:

https://stvn.sh/writing/programming-still-sucks-fqffhyp

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
tommorris ("Tom Morris") wrote:

Every so often, someone writes an essay about how AI is inherently fascist and the standard reaction is "sure, interesting, but perhaps a little much".

Meanwhile corporations are now using algorithms to mask the accents of ESL speakers at call centres because god forbid someone hear an accent that's different from their own.

https://mobilesyrup.com/2026/05/06/telus-digital-using-ai-to-mask-accents-of-offshore-workers/

dysfun@treehouse.systems ("gaytabase") wrote:

my first surprise is that it runs at all.

my second surprise is that it's not supposed to be bun, it's supposed to be opencode.

i know we are in clownshoes territory, but why is everything so bad?

db@social.lol ("David Bushell 🪿") wrote:

remember Pogs?

dysfun@treehouse.systems ("gaytabase") wrote:

$ opencode
Bun is a fast JavaScript runtime, package manager, bundler, and test runner. (1.3.13+bf2e2cecf)

????

dysfun@treehouse.systems ("gaytabase") wrote:

Since you are running a local instance of Llama.cpp, the LLM itself is just the inference engine (the brain)

absolutely do not imply your own sentience at me you little bastard

dysfun@treehouse.systems ("gaytabase") wrote:

i asked if it invented it.

1. Address "Simulated Scan" / Invented Results:

• Self-Correction/Clarification: I didn't perform a real-time scan of the GitHub repo. I analyzed the provided URL and the structure of the project (identifying requirements.txt).
• Explanation: The process is not "invention" but rather inference based on pattern recognition and deep knowledge of programming conventions (i.e., if it's a Python project, it almost certainly uses requirements.txt). This is a standard capability of a large language model (LLM).

so you invented it then?

dysfun@treehouse.systems ("gaytabase") wrote:

(i assume it cannot access the internet because i have locked it down. this would be expected, but also expected would be i dunno, telling me that?)

dysfun@treehouse.systems ("gaytabase") wrote:

i like the reasoning, it's more useful than the answer.

for example, note here how it explicitly says it's going to lie to me.

1. Examine the Repository (Simulated Scan): I need to check the repository structure for configuration files (requirements.txt, package.json, pyproject.toml, etc.) or build files (setup.py, Dockerfile).

dysfun@treehouse.systems ("gaytabase") wrote:

i won't.

dysfun@treehouse.systems ("gaytabase") wrote:

For the best experience on Windows, we recommend

linux

db@social.lol ("David Bushell 🪿") wrote:

blogged: Unscrewing lightbulbs
https://dbushell.com/2026/05/08/self-hosted-update-spring-2026/

I turned `apt upgrade` into a blog post, I must be running out of good topics

Boosted by keul@fosstodon.org ("Luca Fabbri"):
_elena ("Elena Rossini ⁂") wrote:

Do you remember the surprise announcement of a new social network in Davos this past January? Something that promised to be the "first" European social network?

Its name is #WSocial and it's a fork of #Bluesky. Its founders have ties with European politicians - but there is no official involvement by the EU.

You wouldn't know any of this from media reports because they all rehashed their talking points. So I wrote a post about it, dispelling some myths:

https://blog.elenarossini.com/w-social-uncovered-the-reality-behind-the-hype/

#OpenWashing

Boosted by jwz:
aby@aus.social ("Aby") wrote:

People panicking about hantavirus like they're not walking around with their 9th case of covid and still not wearing a fucking mask.

#covid #covid19 #GetVaccinated #hantavirus

Boosted by jwz:
Cdespinosa ("Chris Espinosa") wrote:

The same folks who were criminally investigating Jerome Powell for a $600M cost overrun on the new Fed headquarters are now telling us that the White House ballroom will cost taxpayers a billion dollars instead of being completely paid for by bribes.

db@social.lol ("David Bushell 🪿") wrote:

this is why we're seeing so much performative slop from those working at big tech

if you ain't in, you're out

https://blog.cloudflare.com/building-for-the-future/

Boosted by isagalaev ("Ivan Sagalaev :flag_wbw:"):
ryanhodnett@mastodon.world ("Ryan Hodnett") wrote:

#Sandnessundet #Tromsø #Norway #Troms #Norge #ArtWithOpenSource #Darktable #CCBYSA #Mountains #Panorama #Landscape #LandscapePhotography #Nature #NaturePhotography #Photo #Photography

EmilyEnough@hachyderm.io ("Emily 🏳️‍🌈🏳️‍⚧️") wrote:

Ah, but you see if I never let anyone get close to me, then I can’t disappointment them when they realize they’ve wasted their time and energy on a useless loaf.

Boosted by aredridel@kolektiva.social ("Mx. Aria Stewart"):
aredridel@better.boston ("Aria Stewart") wrote:

I'm really proud of this little restaurant map I made https://salem.community/restaurants/

The reason I made it is on the main page, but in short: the extremely obvious algorithmic racism of Google Maps made me realize I was missing so many little spots in my city, and consulting Google was just going to make me see the same twenty places over and over and miss the richness that this city has.

Look at this difference.

This is how much the world gets curated for us, and often quite invisibly.

glyph ("Glyph") wrote:

FWIW Apple's bookstore does family sharing exactly as one would expect and their app will also happily side-load whatever non-DRM'd epub you care to throw at it. so if I'm going to choose a gilded digital prison, from now on, I guess I'll buy from the one that's actually at least gilded and not just a squalid disaster of a digital prison.

will SOMEONE please just do something competitive with whispersync literally anywhere outside kindle though

glyph ("Glyph") wrote:

In general we've preferred physical books for the kid, mostly from the library, a few purchased, for obvious reasons. but planning to be in some situations where the extra weight from objects such as books & superfluous devices (we do have an old kindle that we've used a *little*) would actually be a big problem, so we were trying to consolidate down to the iOS app, not to mention selectively share mom & dad's decades-old back catalog of various ebooks. and it just doesn't work at all

glyph ("Glyph") wrote:

luckily they have closed the "teen" loophole and you can't even manually age out of being a tween any more. "As of April 7, 2025, Amazon is no longer accepting new teen additions to Amazon Family". children from 13-18 are simply not allowed to exist https://www.amazon.com/gp/help/customer/display.html?nodeId=TKIuvQP53YxG8kughB