As much as I like using Proton Mail and VPN, their current offerings have grown exponentially in size.
I would highly recommend anyone here to not put all their eggs in one basket. Proton can and has in the past disabled user accounts for no reason. This means that you will lose access to everything you use with them.
Only use Mail or VPN and use other services for other needs.
- Mail -> Tutanota
- Calendar -> Tutanota
- Drive -> Just make a NAS. I don't trust any provider with file storage.
- VPN -> Mullvad
- Pass -> Bitwarden or Keepass
- Wallet -> Don't buy crypto
- Docs -> ~~OpenOffice~~ LibreOffice
- Sheets -> ~~OpenOffice~~ LibreOffice
- Authenticator -> Aegis or Ente Auth
- Meet -> This is fine.
- AI -> Run something locally
Run deepseek/qwen locally or something. It's still useful. Never use AI as a service though.
By running AI locally you're avoiding the "easy" issues with AI: privacy and control. Sure.
But you're still falling victim to brain atrophy by a language prediction model pretending to be able to reason, you're frying your ability to decipher and organize complex information by yourself in exchange for an average-blob of collective internet answers.
This second part is significantly more damaging than the first, and we are just at the start. Don't engage with the slop, wether it runs on your computer locally, your phone, or wherever else.
it really depends on exactly how you engage with AI. using it doesn't magically make you stupid, it just makes it easier to be lazy in certain ways. I think LLMs are interesting for search (even though we absolutely still need normal boolean search because language models are lossy). They seem like they're good at synthesizing information, but that seems like an illusion to me. You gotta establish boundaries because the hype (and our cultural notions about AI) don't give us healthy or realistic ones by default.
As an example, I'm not a programmer. Not interested in being one. I set up Radarr on my home server this week and it wants all video files in their own named folders to import. I had like 400 movies I needed to manually make a folder for so Radarr can detect it. I asked AI to make me a script that just read the name of the movie, made a folder with the same name, then moved the movie into it. Took all of 3 minutes for a task I will never do again.