I do trust the devices on my network but I guess I’ll probably look into how to setup HTTPS.

How did you get this number?

Going to a PeerTube home page feels like a weird mix of Interdimensional Cable and very early YouTube.

Discoverability is almost 0 and I can’t easily find any videos that I’d like to watch, as you said. I assume most people would leave once they see what’s on the home page.

Found this for PeerTube.

How does it compare to Owncast?

I would argue that Google is much worse—their entire business model is built around collecting personal information and selling advertisements.

If we talk about smartphones, Android is open source, but is very limited without Google Services (GrapheneOS is great for privacy and security, but then you cannot pay with your phone, for example, because it needs to be approved by Google).

Linux on smartphones is more of a proof-of-concept, and not really usable as a daily driver.

So, Apple is the ‘least bad’ option now. (and there are ways to sideload FOSS apps)

Apple indeed has some great hardware, and their ARM-based computers have zero real competition. Among big companies, it’s also quite good re: privacy, for the most part.

They don’t even have to be on Lemmy. Other platforms in the Fediverse (like Kbin) are compatible with us.

Here is main takeaway from the abstract for those who don’t want to read the whole thing:

Through our experiments, we identify a key shortcoming of LLMs in terms of their causal inference skills, and show that these models achieve almost close to random performance on the task. This shortcoming is somewhat mitigated when we try to re-purpose LLMs for this skill via finetuning, but we find that these models still fail to generalize – they can only perform causal inference in in-distribution settings when variable names and textual expressions used in the queries are similar to those in the training set, but fail in out-of-distribution settings generated by perturbing these queries.

Yes, but if it’s too overwhelming, they’re not switching from said monopoly!

I have to go to sleep, it’s almost 3am LMAO

but it looks great, I’ve just set it up

None—it’s all very chaotic. That’s the problem. I don’t know how to solve this.

I’ve just found this but I haven’t tried it

Another tip for more advanced use cases: that command will also generate a json file for each submission with a LOT of metadata.

JSON Hero has a beautiful UI for exploring those files! (it can also be run locally)

maybe check this too: https://superuser.com/a/1767761

(not sure if the command works on Windows though)

I don’t use Windows, but yes, you do need Python.

Maybe check here: https://www.python.org/downloads/

Yes, I tried to keep it simple because it’s a limit that can’t be bypassed. Work around is running it, and deleting the most recent 1000. Then running it again.

Do you know of a good way to automate the removal?

I know what to do here now /s

Bluesky is invite-only and nothing like Reddit.

And while Lemmy and Kbin are mentioned quite a bit in some places, they don’t seem to be well discussed yet, to say the least.

While absolutely not important in the grand scheme of things, that day I opened the Twitter app and saw that fucking dogecoin icon as the splash screen, taking up all the screen real estate for a second or so, I was like: what am I even doing in this clownery of a platform??