As a user, I think this is the biggest issue I have with this policy. Disclosure is a big part of open source and hiding whatever tools are used is fundamentally against that philosophy.

if you want to lock your phone to one app when you hand it to them you can use app pinning in settings. otherwise look into creating a user profile and see what you can do to lock that down.

what type of phone/OS are your phone and computer?

there’s some hands-on controls too that may be useful but it starts with just giving them their own profile or login that’s ideally with as few permissions as possible. For android, there’s profiles which you may be able to lock down a bit more than your own profile. Apple had family mode settings that you may be able to turn on and off when they have the device. For a computer, same approach. Standard user account with whatever you installed using an admin account and lock down content via DNS like cloudflare for families, https://adguard-dns.io/public-dns.html, a self hosted one or something like https://nextdns.io/.

“Excel is the second best tool for most jobs”

• From the passwords.doc file on the desktop

Worth a read if you haven’t: https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing

but hasnep covered it well. also emoji use, overly wordy and that graph/tree is a common style of llms.

how is there still 17% usage?!

Mr moneybags over here with half a grand in RAM.

has anyone noticed more “Spotify” created songs in their queue? I had a sleep playlist and discover weekly that had maybe 20% ai generated songs created by “Discover weekly” or Spotify.

Personal privacy is a right. If someone commits a crime and the equivalent of a warrant is issued that data becomes evidence and it becomes your job to discover what’s on the devices.

Jobs like these exist in that balance between privacy and justice. A judge or however your justice system works has said that the potential crimes that person has committed justifies waving that person’s right to privacy in the same way as searching a home.

Your job and scope is narrow and you should take that seriously but would you rather someone who doesn’t respect privacy (e.g. a department inside the police force) does this job?

For the tools, those are the industry standards. they’re just tools and yes they’ve been abused but they or tools like them are necessary to do that job.

you could definitely run proxmox in a VPS but I was thinking more of debian in the cloud. Then exporting that computer from wherever you host it as a ova file then importing it into your local server: https://pve.proxmox.com/wiki/Migrate_to_Proxmox_VE#Automatic_Import_of_Full_VM

and on the ram issue, I’m running on over 10 year old workstation hardware ddr3. it doesn’t take much to run.

NAS with zfs storage does typically want a lot of ram (e.g. truenas) but still for a small home environment it works fine with a lot less than 64gb.

I run proxmox bare metal (debian-based hypervisor) and virtualize debian which has all my docker containers.

with virtualization you can take a snapshot and backup everything that way. super simple to roll back if you mess something up. the added bonus is you could probably take a snapshot of your VPS and migrate it to a local vm.

Ubuntu is something I played with about a decade ago and I never felt like I was “missing” something for debian, especially on servers.

kde desktop, which you can install or select as an install on nearly Linux distro as your gui. It has themes that you can use to basically mimic apple’s feel. For hardware, framework laptops have great Linux support. I have a Dell and some things like fingerprint sensors took some tinkering to get working.

digging into sources/references I’d take this as a pretty reliable source. https://news.bloomberglaw.com/tech-and-telecom-law/big-tech-ai-spree-revives-accounting-devices-that-toppled-enron

ente photos and locker is open source and offers a paid hosted option or self hosted. https://ente.com/

https://youtu.be/MXEInc6Zy7Q making 3 AIs count to 100.

I’m not familiar with the source and they probably have some critical findings but this reads like ai slop wrote it.

I’m not entirely sure about the technical differences but from my understanding VPN connections are preferred. From a security perspective, ssh has some more considerations since it’s easier to detect it’s open, and you should lock down root access and other privileged accounts. but SSH seems simpler to actually get working vs a VPN solution which would probably require a reverse proxy or something to get the TV working.

For example, compromising a ssh service gives you access to the shell immediately vs wireguard or similar that historically (from my knowledge) has had fewer critical vulnerabilities that could lead to remote code injection or access. This is also why many corporate and best practices recommend layering ssh through a private VPN like IPsec, OpenVPN, wireguard, etc.

in practice it’s most likely fine as long as

• you don’t use root or an account with sudo to do the ssh forwarding
• require a ssh key for all connections (at minimum any remote/internet connections)
• update the system regularly. you can automate security updates with unattended upgrades on debian-based systems.

Are you connecting from a public network or something? like a hotel wifi or other?

The easiest solution would be to setup the pi as your router and use a VPN like wireguard (wg-easy) or tailscale.

if it is a public network, you can double NAT. There’s dedicated boxes like the GL.inet travel routers that support wireguard/openVPN and beta for tailscale. they have some features that work well with captive portals.

If it’s a home network, you can probably use your PI as a entry/exit node or VPN client instead of using ssh.

wireguard is self hosted and you do have to “expose” one UDP port. From the outside it’s difficult to detect that this “opening” exists because wireguard just listens and ignores everything unless you send the encrypted credentials. Compared to hosting a webpage or jellyfin directly this is much more secure. As long as you keep wireguard relatively up to date you don’t really have to worry much about it.

I personally use wg-easy. It’s designed to be deployed into docker (using docker compose is by far the easiest).

Then you can either use your IP address, or ideally a dynamic DNS provider so you’d connect to myexample.com:51820. Duckdns is free, otherwise options are available like cloudflare. If you can get jellyfin working, this should be relatively straightforward.

I recently had to increase my proxmox storage as well from an old 256 to 1TB. What I did was make a copy of /etc via PVE Host Backup and saved that on my NAS/external storage. Almost everything is in /etc/pve. Then I created backups of all the VMs and stored those on the same external storage. I then installed proxmox as normal and compared configs between backup and new configs then restored VMs from backup. The reason I did it this way is because 1) I had installed proxmox a while ago and new config > old config for stability after adding some necessary PVE scripts (e.g. intel chip, and 2) I’ve had weird issues before cloning drives and a fresh install was easier than risking some weird edge case troubleshooting. It also let me keep the old SSD as a backup in case something went wrong.

Edit: Also recommend going with zfs mirrored on the new install during the setup: target disks options and zfs mirrored. ZFS offers some benefits vs the default lvm.