Wow, as someone who has been hosting XMPP for family for 10+ years and only liking it, I’m amazed by the overwhelmingly supportive comments for it, it’s like people are recovering from a Matrix hangover and rediscovering the things they already had for their merit. So yeah that’s my +1: XMPP is really something dependable, it just works, and the sysadmin story is great too (my preference goes to ejabberd for the best “turnkey” experience, but you can’t go wrong with prosody either if you are willing to read some guide).

You don’t have to be on the NAS’ physical network to run some tests, if you can SSH into it, that’s enough to run some speed tests and latency tests to reference servers (e.g. Google, netflix, …), while doing the same from your PC simultaneously. Whenever you encounter “instabilities” you should be able to see which side of the connection is struggling more than usual. You should also be able to monitor the NAS itself for excessive IO or CPU using something like htop (even though that would be rather extraordinary if all that’s going on with it at that time is a Borg backup).

On the topic of Kopia, you can go the easy route of setting up a SFTP remote pointing to your NAS. All you need is SSH access to the NAS (so your current tailscale/borg set-up qualifies as is) and a folder where the SSH user can write. You can then compare backup times vs. Borg (in my use case, going from 18-25min with Borg down to 3-5min with Kopia). If you then want to try the HTTP repository server, there is a container for that, but the set-up is a bit annoying (the password to decrypt the repo needs to be stored on the NAS, you need to generate a SSL certificate pair and have the client whitelist the sha256sum of it, etc). Personally I haven’t found a substantial advantage to having the http repository server but YMMV.

Tailscale being merely a wrapper around Wireguard, which is pretty close to optimally efficient, I don’t think there’s much of an improvement to be expected by rolling your own or hopping over your VPS. You should probably start with investigating what’s causing the tailscale connection instability (is it your ISP? Is it the network on the side of the NAS?). If you are not bound to Borg, perhaps you can give a shot at Kopia, I found it to be much more performant for my use-cases, and it has a http “repository server” which you might be able to use with less protocols overhead than through your current setup.

Or plain old CalDAV

It’s there on the readme: https://hacdias.com/2026/07/28/filebrowser/

The project needs a rewrite from scratch, but the author had enough.

How does that compare to wallabag? (Does it?)

If you want notes “sometimes on the web, sometimes local first”, and possibly more than just notes (with the tools to extend into an actual knowledge base), you might be better served by https://triliumnotes.org/

I’m holding off on Fluxer until they decide how they’re going to implement federation

Seems wise. They seem competent in the front-end/client space and complete amateurs in the (difficult) protocol space. There is no example of successful tech (that I know of) that successfully added federation/P2P after the fact. It’s not an afterthought, it probably won’t ever happen.

Don’t do Element, Matrix is a nightmare (and a significant commitment) to self host. Other servers (recently, continuwuity) are a bit better on that front, but then you run into compatibility issues and edge cases as a forever second-tier citizen.

My advice is to just go with XMPP and ejabberd, and you will find clients for all kinds of usages and people (a free-er WhatsApp takes you to Conversations/Quicksy/Cheogram/Monocles/Monal, a better banquet/IRC-style rooms takes you to gajim/fluux, social networking and group calling takes you to Movim, etc).

Personally my needs are covered by Monocles on Android, Gajim on the PC, and Movim on occasions. Using multiple clients around the same protocol and account is a strength, not a weakness.

To add to the collection of chat apps enabling that, there is XMPP, too: https://f-droid.org/en/packages/de.monocles.chat/

That’ll work until they create for themselves new competitors that are good enough at previous (normal) market prices. And you don’t need incredible high foundry tech for that, so it’s closing up on them eventually.

It’s not as dramatic as GP makes it, but Matrix had been at the prototyping stage and at figuring out its core protocol and state resolution algorithm for more than a decade now, it’s pretty ridiculous in itself.

Isn’t that the premise of the “$4 ubuntu over droplet deployment” option?

Instead of having Secluso-Deploy ssh into some cloud box and prep it up with the server-side software, why not have a container deployment option as well? I get that you want to ensure that the server is publicly reachable for the mobile clients to work on the go, but ultimately (and in all honesty) at this point, that should be a user concern/choice (those more advanced users may be peculiar about running behind tailscale, a home-VPN, a port-routing config, …).

Needless to say, most people here might find it easier to work with containers and build trust in the project by having it run in an isolated environment with limited permissions than blindly trusting that the code is what it says it is and not quietly running a botnet at digitalocean with their PII attached.

No it’s not, and part of that is the current legislative laissez-faire in the US that put its regulatory bodies on a hiatus. Under normal circumstances, this stuff should have been under much more scrutiny and regulations. I’m not saying that the state should control what LLMs do or who’s access to them, but they could very much tackle the deceptive marketing, environmental and societal impact, unsound financing, abnormal market consolidation, and mitigate the overall financial risk.

Hoping this goes somewhere https://github.com/containerd/containerd/issues/13307

I still have mine, it’s a decent audiobook player for the kids, it even has a jack, a replaceable battery, induction charging and is so nimble. And did I mention the full keyboard? I hate that we followed Apple design ideals, for the sake of having today overpowered dumb, heavy , fragile and inconvenient consumption-only bricks.

+1 on babybuddy, I’ve only good things to say about it and its community. Also, it’s a Django app, so the bar to tweak it exactly the way you want is very very low.

Is that another overpowered smartphone SoC running a wasteful android with a thin UI and dumbed-down music player? If so, let me just repurpose an old smartphone and save the planet from some e-waste. If not, kudos.

Maybe I’m the weird one here, but when I look at fiio, and they are essentially smartphones (beefy Exynos or Qualcomm SoCs, running a bloated Android), minus some connectivity apps and a mobile antenna, I’d rather just use a phone, thanks. Most portable music players have this minimalistic appeal of devices that were engineered for just one single purpose and to do it well, and fiio certainly isn’t having that vibe.

Yes, the point is LLMs are AI.

So, we’ve gone full circle. LLMs is a sub-category within the collection of ML techniques within the collection of AI techniques. Point is, AI is just a term here. A label, that indicates nothing about the abilities of LLMs to exert any form of intelligence or reasoning. In other words, the whole field of AI could (should?) have been named “computational statistics” or “mathematics applied to numerical datasets” (or whatever else you want, really…), and LLMs would absolutely belong to “CoStats”/“MAND” fields, for the same reason we say they relate to “AI” today, it’s just that nobody would be silly enough to call them “artificially intelligent”.

ML is AI too. But sales didn’t call it that because AI had the reputation of “just brute force with some heuristics”.

What sales are you even thinking about? What do even presume the market for ML algorithms to be? Nobody was shopping for support vector machines as a service, or spending tokens on linear regression, or using convolutional neural networks via API before the current LLM craze. What the field is experiencing right now is unheard of. Never before did the private sector jump gun on a niche technique and spent trillions to package, anthropomorphise and market it as if “AI has finally been figured out, and we happen not only to own it, but also to sell it to you”. This deceptive rhetoric would be a much tougher sell if some early computer scientists hadn’t happened to name their field “AI”.