Depends on the watermark method used. Some people talk about watermarking by subtly adjusting the words used. Like if there’s 5 synonyms and you pick the 1st synonym, next word you pick the 3rd synonym. To check the watermark you have to access to the model and probabilities to see if it matches that. The tricky part about this is that the model can change and so can the probabilities and other things I don’t fully understand.
How do you expect the packets to actually route? If you run Tailscale and your VPN on your phone, they might fight with each other for control of the routing table.
If you’re trying to use Tailscale exit note to then route through Tailscale to one node running gluetun to Mullvad. That’s going to be complex because against they both want to mess with the routing table.
Tailscale natively supports Mullvad: https://tailscale.com/mullvad
Okay it was a little hard to read since your post was missing formatting. TS_SUBNETS is what controls what CIDRs are announced through Tailscale. Since you’re not using Docker networking for Jellyfin, it would be whatever subnet the host is on. Maybe it’s 192.168.x.y
Gluetun doesn’t make any sense here. You’re forcing all the traffic for from Jellyfin to go through Mullvad, but you need to be able to connect to Jellyfin because Jellyfin is a service you connect to.
Since your Tailscale is host network mounted, you’ll be able to expose your Docker network subnets over Tailscale then access Jellyfin. This is done via the TS_SUBNETS env variable. Docker will use a 172.16.0.0/12 subnet.
You probably intend to gluetun your downloading software, not Jellyfin.
Was that anything more than just rumors? Letting a currently monopolistic company keep the browser because another bad billionaire might buy it and do something bad with it just prevents anything from changing.
Your options are to run smaller models or wait. llama3.2:3b fits on my 1080 Ti VRAM and is sufficiently fast. Bigger models will get split between VRAM and RAM and run slower but it’ll work.
Not all models are Gen AI style LLMs. I run GPU based speech to text models on my GPU too for my smart home.
It has to be the employees not the state because companies withhold it and remit directly to the IRS. Not saying you should do this, but if you increase your withholdings exemptions then it won’t go to the IRS. Though you will owe it in April and may have to pay penalties for underwitholding.
Who organized this form? Is there something official to make it look like it’s not just signing me up for spam?
I use a variant of this: https://github.com/linuxserver/docker-wireguard
You don’t need two different containers for this. They’re going to either fight each other for control over the networking tables or run wireguard in wireguard
So I had a chance to try this out. It wasn’t on Google Play Store, only F-Droid. There isn’t really SSO support, you either login with User/Password or a token. Instead, I login with my browser, get the token and paste it in. That works fine, but an ideal world is just pop up an browser WebView and go through the flow, then grab the token. Maybe it was intentional, but PaperlessShare registered as an Open handler for PDFs and the share menu, whereas this is only share menu. This seems to mean that I need to grant file access, whereas the open handler didn’t need that I think.
Overall, it does the job and gets my docs uploaded.
My prediction is that manually reviewing user creation won’t scale to a high level and unless systems develop spam detection and reputation management similar to email then it’s not going to be limited to just one or two bad instances.
Its trivial to create my own instance with a new domain and there’s no limitations against sending ActivityPub messages to a server. Unfortunately the simplest fix is for big instances to restrict what instances can communicate to it, but that causes centralization.
Plus, we don’t need to be huge. There’s no benefit from it.
The benefit is breadth and depth of communities. Reddit is great because if you are interested in a topic, there’s a bunch of people talking about it.
How hard would it be to contribute these improvements back to the project in the form of either distro package updates or documentation changes? Did you consider that?
Can you expand on the OIDC/OpenID support? Does it support SSO based authentication to Paperless? That’s my biggest problem with PaperlessShare right now. I can’t actually sign in because I use an SSO provider.
I respect your strong ethics and sticking to them, but saying they people support genocide for using software hosted on GitHub is an extreme position.
I could easily change this: Do you drive a car or fly a plane? Then you must have no red lines against climate change.
The problem is the payment processor. There’s only so many of them that customers actually choose to use.
How about some Yubikeys or smart cards instead of something that requires me to scan my retina and share it with Sam Altman
Right. Zero trust means at the very least you need to add AuthN and AuthZ to every endpoint with no exceptions for internal IP addresses.
Encryption at rest just means the data itself is encrypted when stored on disk and the key is somewhere. It doesn’t dictate that the key is not visible to the server.
Encryption in transit refers to an encrypted channel from client to server.
E2E encryption usually refers to encryption from one entity to another where any intermediary servers do not have the ability to decrypt
Source: too many years doing application security at my job
I’ve been running my own mail for 10+ years. I recommend rspamd for spam filtering. It took the place of SpamAssasin, grey listing, SPF checking, etc. All in one single system.