Show newer

Easy to fix though, and would only be exploitable by a malicious Matrix homeserver affecting their own media, so no point. I was keying thumbnailed cache entries as $mxc-$crop-$widthx$height, but those are all characters that could be added to a (malicious) media id on the same homeserver

Show thread

lol lmao just realized there's a glaring cache poisoning vulnerability in my refactored code

very sudden urge to work on my fediverse client again even though I have so many more pressing things to work on

@cblgh that's a good set, I remember at one point there's the rollerblade song in the movie hackers that start up 🔥 :moar:

purchasing something called "knotting solution" from the hardware store and once again regretting the day i ever went online

The current level of academic discourse on LLM's 

Anyway I brought this up at the LLM discussion group at an academic conference this week:

time.com/6247678/openai-chatgp

(tl;dr—Kenyan workers are being paid exploitative-level wages to manually filter out the worst stuff from the internet and this is necessary for how LLM's work)

And in that hour, only one person even addressed the issue, and their take was ~ "colonial exploitation is good, actually" and everyone just sorta nodded their heads

I removed the custom CSS for my profile for now to show off the work @f0x did on the GoToSocial web view for profiles. Go open this in your browser, and go 'ooh' and 'ahhh' and 'woooo', and also 'hunnnhhhhhhhh'!

https://goblin.technology/@tobi

having to share Folding Ideas in a company meeting to try convince them that going with NFTs, surprisingly, isn't the ethical solution lol

i would just like to spend like 8 hours with nothing but a room of cats

people are liking gotosocial a bit too much, we should introduce something terrible next release

It has been proven repeatedly that platforms prioritizing growth over safety are toxic to everyone else who doesn't present as white, straight, and a man.

No, I'm not going to debate it anymore. No, I don't care if people think I'm wrong to center safety in my dev efforts because I can live with it going sideways if that happens.

But in my experience as a dev and a citizen of the web for more than two decades, the missing link to rehumanizing social media spaces is safety for the most marginalized people that have historically been ignored.

And I'm going to create around that ethos.

Show thread

I'm going to repeat this so it's clear.

If the fedi fractures around the ideological lines of safety vs. unsustainable growth, I'm okay with that.

I am very comfortable with being part of the fedi that actually gives a shit about people rather than treating them as product.

I have no interest in repeating the same mistakes and creating decentralized rage engines.

I believe this space can be better.

:boosts_ok_gay:​Help with Linux Screenreaders 

For better accessibility testing of the software I build, I've started testing more with screen readers.

On Android, enabling and using TalkBack is relatively straightforward, but for testing on my desktop, Orca is completely unusable.

I have a brand new Debian VM with GNOME, and Orca enabled, but actually using it in firefox just barely works. TalkBack has a nice workflow with highlighted elements that you move between as you read the page, but Orca seems delegated to using caret browsing and reading whatever your cursor is on? Am I missing shortcuts (which are also a pain, because the modifier can only be caps-lock or insert, not really conveniently placed)

Am I missing something, or is this the state of on ?
Also the tts quality is very poor, which would make this even worse for any prolonged testing, but maybe there are alternative engines?

Show older
Pixietown

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