Maybe it'll be politics related. Or maybe it'll be a hit tweet they'll edit to make a bunch of users look like they are cheering for AI.
Without guardrails the possibilities are endless.
I'm waiting for the first "viral hit post -> edit into context the replies diametrically oppose/hate -> clip and shame participants" wave to hit.
It'll be interesting to see what it'll be.
It's not a browser standard def. And yeah I'm still waiting to see if Firefox picks up a TL standard.
But firefox's TL stuff that they already do is pretty much good enough IMHO since it's just "right click -> translate" and it works quite well.
It'd be nice to have an app level TL tho.
Well there's an argument that edits should be added as a record linked against the original. That way it's all user owned.
Then the links against the original content hold and if you change/delete the original then it breaks the link in the appview and/or shames the user.
Oh so it's just the visuals of it then and not the stuff under the hood. But certainly either way it needs more work before it's ready for big bluesky and not just other apps.
It depends on the browser tbh. I think chrome has their own AI stuff (idk i don't use chrome) but I know firefox has their own local AI interface that uses OS level HW acceleration.
And I think non-extension JS can access it as well but im not sure.
blog.mozilla.org/en/firefox/f...
It actually does. Firefox implements this and has for years and it's hardware accelerated just like on phones.
It started as an extension in 2022, then was upstreamed in 2024 and started shipping to mobile firefox last year.
And of course android + iOS have their own native TL stuff now too.
Currently you can do edits (like other apps do) but they aren't perfect and they break things. It breaks likes, reposts, QRTs, and replies.
When bluesky actually implements edits it will require changes to avoid breaking those things or edits will be broken forever.
The reason bluesky themselves don't add edits is because when bluesky adds a feature, it's the "reference implementation" that all other projects on atproto follow and need to be compatible with.
And native AI translation (what is used on android, iOS, and firefow) is running a local language translation model on your device so that what you are translating doesn't get leaked/sent to a server. i.e. it's a privacy feature.
To further explain this. Basically all modern language translation is AI/mahcine learning based. Structural translation rules based on grammar and dictionary used to be the main technique but now they are only a small part and the rest is all AI based (and has been for like a decade)
Turns out lawlessness is not a winning strategy. See you at Nuremberg 2.0
64/100 - Nameless
#strangefake
Congrats on 1 mil on Snake Eyes!
#BaelzBrush
test
I doubt the Iranian Kurds would even support a Kurdish run Iran. They mostly just want to break off and form their own independent territory given Iran won't respect their autonomy.
Scanning test ballots in New Hampshire with a VotingWorks tabulator, screen reads "Your ballot was counted!"
Testing our latest voting equipment in New Hampshire today. So proud of what we've built at @voting.works. A robust tabulator, built on a secure foundation, and fully open source.
iron
I put my thoughts together in what is /hopefully/ a more cogent way in this article.
trezy.com/blog/atproto...
#strangefake
Yeah I agree that there's better ways of approaching this.
My only gripe with passkeys is with the UX some orgs/sites are providing while implementing them.
I honestly prefer passkeys tbh. I just was commenting on the dangers of trying to avoid informing the user when doing stuff like this.
Yes. They can actually. That's the problem.
Okay but how is the user supposed to know this?
The amazon webpage isn't saying "create a passkey". You are just on the website trying to browse and it suddenly pops a system UI prompt without even telling you that it was the one to do it.
PIN+biometric and/or PIN+hardware key are my preference IMHO. quick and easy but with a fairly hard to cheat lockout mechanism
So the temp user calls the owner over and asks them to enter their biometrics/PIN because "the computer keeps trying to update and it's getting in my way". Neither user actually reads the OS-level biometrics prompt because it's not the web browser/app, just some separate popup
That's how it happens
As I said before, training the user to see the OS-level biometrics prompt the same way they see UAC, etc and not tying it to specific webpage/app prompts/actions makes them see the biometrics prompt as "oh the computer is trying to do updates" not "it wants to create a passkey for me".
(cont)
How is it not possible?
We're happy to announce a long-term partnership with Motorola. We're collaborating on future devices meeting our privacy and security standards with official GrapheneOS support.
motorolanews.com/motorola-thr...
If I the median user see these spurious "auth your PIN/biometric" prompts without context on my personal device on a semi regular to regular basis I learn to tune them out and so I go use a friend's device and fail to understand I'm saving my passkey to their device.
It's not that you are using your friend's device so regularly that you get tired with the prompt. It's that in general websites try to silently create passkeys without telling the user trains them to tune the prompts out which leads to security failure modes.
