I’m a big proponent of self-hosting, right to repair, and rolling your own whatever when you can. That probably started as teenage rebellion that got baked in - I was lucky enough to read both Walden and The Hobbit during a week-long cyclone lockdown several decades ago - but I suspect there’s a non-trivial overlap between that space and privacy-minded people in general.
My endgame is a self-sufficient intranet for myself and family: if the net goes down tomorrow, we’d barely notice.
I also use LLMs as a tool. True self-hosted equivalence to state-of-the-art models is still an expensive proposition, so like many, I use cloud-based tools like Claude or Codex for domain-specific heavy lifting - mostly coding. Not apologising for it; I think it’s a reasonable trade-off while local hardware catches up.
That context is just to establish where I’m coming from when I say this caught my attention today:
https://support.claude.com/en/articles/14328960-identity-verification-on-claude
To be accurate about what it actually says: this isn’t a blanket “show us your passport to use Claude.” Not yet.
The policy as written is narrower than it might first appear.
My concern isn’t what it says - it’s that the precedent now exists. OAI will do doubt follow suite.
Scope creep is a documented pattern with this kind of thing, and “we only use it for X” describes current intent, not a structural constraint.
Given the nature of this community, figured it was worth flagging.
The intranet becomes the internet :) Everything is local, accessible from multiple devices within my wLAN. The main box plugs into the router and serves everything over Wifi to trusted devices - my documents, media, books, games etc.
I wrote (flippantly) about the bones of the system here, 3 or 4 months ago. It’s more complex now, but the endgame has always been “what if cloud, but you are your own cloud?”
https://lemmy.world/post/41315607/21438607
It may not be fresh (if the net goes down), but it would be local. The only real question I have to grok for myself is if I want to mirror curated section of Wikipedia, books etc.
https://en.wikipedia.org/wiki/Kiwix
Probably I should. May as well go full data-horder. Good excuse to get a few more TB of storage. What I’ve done so far is all within 4TB, using clever tricks and black magic but there’s a limit to 4TB. Fortunately, hard drives are still not too $$$. +4TB is about $200 here locally. So the entire set up is still around $600-700 AUD (around $350-400 USD)
All the other stuff I have more or less tee-ed up (barring the UPS + solar kit I am building later in the year).
https://www.youtube.com/watch?v=1q4dUt1yK0g
Anyhow, 8TB local should just about cover what I have in mind (he said, fully aware of dragon horder sickness). Then I’ll grab something for offsite storage for critical docs - I have an old raspberry pi with a 256GB NVMe ssd I can use for that.
I’m semi tempted (because fuck it, why not have fun) to look into LoRA after that.
https://en.wikipedia.org/wiki/LoRa
What I really NEED to do is finish the LLM stack (I’m on it and nearly done) and then do a curated Youtube replacement with yl-dlp feeding into Nova-player or Jellyfin, once/if SmartTube etc gets shit-canned. The youtube thing I’m kinda excited about because I’ve figured out how to squeeze ~1500 videos in around 250ish GB of storage, with TTL (time to live) mechanics, download replacement schedule etc. The kids watch too much random shit on YT, so daddy will make YT at home (ha!).
I have some other wild ideas too…it’s a whole other thing…don’t get me started :)
Once I’m finished, I will open-source the entire thing, post about it here, and let others replicate / improve on it. And so it goes. Once you begin walking down the dark path, you are forever doomed. Be careful :)
OK I have you. You dont need the internet because you have the internet in your terabyte farm. Pretty cool.
Thanks for the detailed reply.
One final question, I’m sure its dark at the bottom of the deep rabbit hole you are in, what do you do for batteries for your head torch?! 😀
Exactly so. Mom - can we get the internet? Mom: we have the internet at home.
Batteries? I don’t need batteries. I have the never-ending warm glow of weaponized autism. And that’s not even a joke.
I tend to hyper-fixate on something until either it breaks or I do. It’s usually 70/30 in my favour :)
Thats weird, i don’t remember having an alt account called SuspiciousCarrot78 but surely you must be me, same project, same neurology… same fixation pattern.
Well, there’s a quick check. Take a shot of whisky and I’ll see if can type “gottle of geer”.
Yep, definitely talking to myself again
Jokes aside, i haven’t seen you mention anything for media streaming.
I highly passionately recommend Navidrome for music. It is my absolute favorite and most used self hosted service.
For acquiring media like film and music depending where you live ripping those from your local library is in some places arguably a protected fair use. (Comes from the time mp3 players became common and runners used to take rented cds in their walkman outside before). In my experience, 480p dvd is much higher quality then internet 480p streams and the total size is much smaller then what you find in downloads.
ARM can help you automatically rip these as long as you have a drive in your pc. I got ARM running in a proxmox lxc with drive passtrough but that honestly was a pain to setup so not sure you should go that exact route, either way the moment arm is functional its smooth sailing and your only concern becomes storage space.
Well, this is going to freak you out, because I am (literally, right now) explicitly scoping out offline YouTube integration into Jellyfin, as a sort of rolling library. Jellyfin has been good to me, but I’ve been using Nova Player for a while now, since my Pi borked itself (Nova player is plug hard drive into router, install app on TVs, done). The limit is that yt-dlp doesn’t integrate very well with it. I mean, I could build something, or fork the repo myself…or I could just use what already exists.
So it might be time to restore the entire *arr stack.
The TL;DR: I want one front end for ALL my media - YouTube, instructionals, movies, TV shows. That immediately speaks to Jellyfin, which I’m very familiar with. The issue is YouTube. There’s too much slop on there, I want a curated experience for the kids, SmartTube won’t work forever, and the eldest is starting to go black-hat and screw around with settings. That’s accelerating the timeline.
The stack I’m scoping:
The YT stack: rolling library logic:
Scoping the maths at 200GB, 30-min average per vid, using compressed modern codecs:
Planning numbers per video: assume average video is 30mins. At 360p, that’s ~100MB per video. 480p ~160MB, 540p ~220MB, 720p ~320MB.
If I have a selection of “core keepers” at 720p H.265 (~300 videos), taking up ~80GB, that leaves ~120GB for the rolling pool:
I don’t need 4K…hell, 1080p is wasted on me. So I’m thinking… 300 core vids at 720p + rolling library at 540p = 845 videos, give or take. More than enough to keep the fam off my back once SmartTube goes tits up (they can’t play whack-a-mole for ever).
I would prefer a clean migration to other, live sources (I have those scoped out as well) but not all the Minecraft / gaming / pretend play / blah blah stuff the family watches is on Peertube/Odysee/Curiosity Stream.
PS: I see your 480p and raise you 60, because 540p is the forbidden resolution :)
PPS: I was planning on using JF for music too…but maybe I should look at Navidrome like you said.
The crazy idea that I had was to use AI to create an infinite playlist of sorts. Seed it with your own music, get it to generate tracks in THAT style as filler, intermingle them (so there’s always something new).
Finish off with AI DJ’s that pulls in “local news” from your curated RSS feeds.
Think: Three Dog from Fallout 3.
Basically what I spoke about here -
https://lemmy.world/post/43936980/22784324
I have a pretty clear idea of how to get that done. It could be amusing.
https://huggingface.co/ACE-Step/acestep-5Hz-lm-0.6B
Currently my setup for YouTube In jellyfin is pretty basic (but it works well)
Its one complex yt-dlp script that runs ecery hour, reads a list of channels from a txt file. Rss to check the last 2 vids per channel, if there not downloaded already (log with vid ids per channel) it starts to download on high quality preset, filter out shorts and “live”, if the result is getting above a certain size it aborts and tries again with lower quality preset.
I don’t have a system to auto dele yet cause the script is only for my favorites. For everything else or that uploads to regularly i host a local invidious.
But google really hates invidious and it breaks frequently, even got my original residential ip banned. I may take a look at tube archive cause i would prefer download and delete as a more stable flow.
Something else i am experimenting with is using the same data in multiple systems. My music collection is a proxmox dataset but as long as the read rights are good i see no reason why jellyfin can’t also read those files.
The output folder of ARM used to be its own library in jellyfin on my first server.