Oh I'm not sure you're familiar with OpenObserve, but that's the platform I chose for my observability lately. I know you're a Grafana guy, but just wanted to throw it out there if you were interested. I assume your intentions are to link it up with grafana?

Data is cool man, that does sound neat.

GM. What does this mean?

I believe I followed the spec correctly. It's just a json document, no processing. Names and relays are returned.

https://gitcitadel.com/.well-known/nostr.json

Freetube is fine when it works. After about 2 months of it not working I uninstalled it and switched. I had too many issues with youtube blocking me then invidious going down etc. Newpipe still works on mobile for now when I need to actually go find a video. Newpipe is the only way I can browse anymore.

No Google is. Youtube requires cookies with ytld if you make more than a handful of connections/hour, even on a high trust score IP. So to use pinchflat/ytdlp you need to create an account and extract it's cookies, then use those cookies to download/fetch content.

All of your downloads link an IP to an account, so they can see with regularity what content you download, but they can't know when you watch it. It decouples the viewing behavior from the content. They just assume you watch it on the interval your downloads are set to. Ideally they can't see that your on the toilet, in the car, or at work anymore.

To get the cookie, you need an account, which needs a working cellular phone number (voip nor temp SMS solutions will not work) to create an account. So it can be personally linked to you.

Otherwise youd need a way to rotate your IP to trustworthy IPs (tor doesn't really work, and most popular vpns are well known and blocked). I was only able to get this working with a "fake" account, cookie, over proton US-based vpn only.

That's completely accurate XD

Yep, I remember last I used it, it was almost impossible to find content by people I was subscribed to. They'd bury it because id watch it, be content, then get off. Instead they want you stuck to the screen, and it works, as long as you let it. It's a natural human behavior that every normal human is vulnerable to, which is then exploited for their benefit, without regard to my quality or as you said, wellbeing. Which is when I transitioned to use the word evil.

If you can figure out how ~30b local models to actually make tool calls with cli tools, I have some sats to send your way.

Learning things, finding things, planning things. IDK about you, but search engines blow fucking hard lately, good luck finding a blog or a forum site about a topic anymore. Most very modern local models are quite good at presenting information correctly if you already know what you don't know. I like using it to teach and practice things I wouldn't have access to otherwise. School is hard for me, I learn things unconventionally. I can promt a model to exercise and train me on topics they way I understand them. It's basically impossible otherwise. Having time to 1:1 meet with a tenured research professor at a university teaching phd classes and asking them questions on your own time doesn't exist. Not for any reasonable amount of money, or without enrolling into a program you don't want.

Like: remind me how a forcing function works when used to evaluate the frequency response of a linear system.
Or: How can I practically switch to double entry accounting in my business?

with local models you can be a bit less privacy conscious and allow it to give you more relevant results based on _your_ work.

The other thing with ytdl/pinchflat that adds friction - there is no discovery outside of word-of-mouth.
Pro - can't get hooked on new channels easily
Con- can't algo discover new content that might interest you

Reading/writing helped me get past that. Now, if I want to watch something, it requires intent. I had to think on my own about what content would interest me, then decide if I have the server capacity to keep the channel around. Real world constraints force decisions XD

When using something like pinchflat, I give myself a week for most channels to auto-delete. Other educational accounts/playlist I store indefinitely, as things like MIT and many other university courses are open licensed making them legal to download and store.

It adds quite a bit of friction to my watching experience I slowed down a lot. To the point where I maybe watch ~3 hours of content/week. And I would have no shame in increasing that as I don't consume any other media, no podcasts, no tv, movies etc.

You can use uBlock origin to block google/youtube links so you can't accidentally watch things. You can add this to your My Filters list
||youtube.com^$all
||ytimg.com^$all
||youtube.com^$all
||youtube-nocookie.com^$all
||youtubei.googleapis.com^$all
||youtube.googleapis.com^$all