redcalcium

You can already make deepfake videos from a single target face image, and now you can make AI generated voice from two seconds of sample? The future is going to be interesting to say the least.

That's exactly what happen in my personal instance.

I see, looks like it's a correct decision to me. Let the Lemmy developers worry about which version of pictrs to use themselves.

I haven't been to Reddit for a few days and they did these stuff already? Let's keep this up.

Reddit used to be open source and the password was hashed using bcrypt.

I'll have more respect if the leak were done by disgruntled employees, but this attempt to leak is done by a ransomware operator who failed to extort them in the first place.

Ransomware operators are scum and should not be trusted, let alone paid.

Nice! Looks like it even has update checker as well. Is there any reason why pictrs is not included in the update checker and hardcoded to version 0.3.1?

Decentralized Identity could be implemented relatively easily just by allowing users to enter a their public key, like in git or PGP. How to sync the data is a different matter though. Maybe you can enter a username (e.g. @user@instance) in your instance's search field and have it federated to your account there if the cryptographic signature matches?

Look for secondhand thin clients such as HP T620. They're usually can be had for $30 or less (or more) depending on the configuration. They also have low power usage, not as low as a pi, but still low enough (<10 W).

I host it on a low power PC in my closet. I also limit the resource usage to 2 cores and 2GB of RAM in Kubernetes (RKE2) config. My ISP won't give me any static IP, so I routed the traffic via a $3 VPS over tailscale network.

And give up their power as mods of a large subreddit and starting again from scratch? Most of them probably aren't willing to do that.