• 0 Posts
  • 14 Comments
Joined 1 year ago
cake
Cake day: June 29th, 2023

help-circle
  • It just makes too much sense… The only way to get past electron is a better electron. Or just fix electron

    We’ve been going after this concept for decades now. That’s what java swing was supposed to be, what python gtlk was supposed to be, and I’m sure there were others before that and there’s been a hell of a lot since then

    It’s all trade-offs between flexibility, ease of use, and performance. Also between maintenance cost, portability, and existing library support

    Electron is a good compromise. The execution could be better, but it’s come a long way. There is no one size fits all solution, but there are some decent options that handle that compromise differently


  • Just to put this in context:

    There’s only so many ways to turn a bunch of files into one - mainly, you stick them back to back. Easy.

    Then, there’s an infinite ways to compress that file… You could come up with you own method, but what good is that? It’s better and smarter to use a format already supported by your users

    So of course most bundles are the same archive type under the hood. Everything from backups to installers - you shouldn’t be inventing new formats without a damn good reason





  • Nah, I’m thinking much bigger. I’ve got an AI that can transcribe video, I’m working on one to summarize and put facts into a knowledge graph, I’ve got one that can hold a conversation, and I’ve got a script that scrapes sites and does natural language processing. I just need an agent to tie the pieces together and some control scripts to manage the containerized pieces

    The idea is, my assistant will go out, read up on programming topics and build knowledge graphs with references to the source, and I’ll fix my biggest issue - shittified searches crippling my work speed

    Then, I’ll send it off to find content. It’ll transcribe/summarize videos and rank them, research topics and come back with reports, and trawl my socials to find new things I might find interesting

    I plan to take all that, then let my assistant create video channels to watch and additional content to read if Lemmy is slow. And if my friends and family show interest, I’ll add in hosting and an internal social media and convince them to run additional nodes at home

    I’ve been working on it for a while because I saw this coming, I’ve got most of the key pieces already.

    And that’s the bubble of Internet I’m building - AI curation of my Internet life, it’ll happily work away the hours deshittifying a bubble of Internet




  • Ditto. I’m not going to put in the time to clean up a project working fine for me just to have it ignored or complained about. It’s all expectation, very little reward

    And even if I did get volunteers, I’m not coordinating people on my personal projects. Sounds like a great way to take all the joy out of it…

    Open source is just broken right now, it runs by draining the passion out of people


  • Well hey listen, I appreciate it. I would’ve spent who knows how long waffling between distos that I don’t feel drawn to, and even if I came across an atomic flavor, I probably would’ve just assumed it was marketing fluff

    Good ideas need advocates, and this is a good idea… It’s a promise of an OS I want, not just running from one I don’t

    I’m probably going to look at bazzite first. If I have containers that can run LLMs on my GPU, that checks off everything on my wish list except gaming. I’ll read up on it though, you’ve given me the context I need to care about learning more


  • Ok, when I googled it earlier I saw “containers and roll back to previous version” and I made a note to do more reading

    Your write up was good, much clearer than what’s on fedora and Wikipedia. And the fact you pitched immutable OS’s in general first caught my attention… The concept is a no brainer. Decouple the os and the rest of the software, and don’t bother digging into one of a kind conflicts when updating things - just make it rebuildable and create it fresh. You never know when the wrong bit will flip

    Nix’s “learn this one thing, configure it once, and you’re done” stuck in my head. And after a different distros, a couple lines installed Nvidia, Nvidia’s docker package and docker

    But then I had to configure WiFi and spend half an hour learning why I couldn’t mount an external drive and how to manage it… I still have no regrets, I’ve got a USB that should start converting my friends and family’s old PCs into a self organizing AI/self hosting cluster… Hopefully it works next month lol

    But not what I want in a daily driver. I want something that’ll quickly do what I tell it and gracefully handle the fact I have 6 versions of Java and no idea why I need a version from 2018 specifically. And that I’m going to add a repo to install something and instantly forget what I did if it seems like the best path forward at the time

    You’ve sold that pretty well - my takeaway was that atomic fedora is very modular and low side effect and also an interchangable foundation I can swap out and roll back easily… At this point, if it can run containers and the drivers I need, it sounds like a great option.

    I used to use VMs so every 6-12 months I could start clean with the latest and run setup scripts for my dependencies… It was just easier than debugging some conflict. This sounds even cleaner - I swap out the base at will, and the stuff I’ve built on it should stay intact. Plus it sounds much more testable

    So my main concern is will it run on an HP omen - it has zero Linux support and a bunch of concerning driver needs, but it does have a second m2 slot… What’s the worst that can happen? Except apparently some models forget they have fans in Linux and I just know the iGPU-GPU switch will cause some problem with sleeping… But Windows is only going to get worse

    Now that you’ve convinced me this might be the best course (I only see less problems than other distros would have), and I’ve talked myself into giving it a go, is there any recommended reading or key concepts I should look into? Any particular flavor(s) you’d point me to first?


  • Fedora Atomic a chance, it’s an extremely nice family of distros (e.g. Bluefin/ Aurora, Bazzite, etc.)!

    Can you elaborate on this? I landed on nix for my PC turned server and haven’t regretted it, but I’ve been hesitant to go all in on my main laptop (I’m wary of my laptop iGPU and GPU switching becoming a config issue, and I’m dreading having to configure my wsl dev environments again…)

    Windows is getting blatantly terrible enough I know I’m just putting it off, maybe a cool new technology might help make it sound more fun


  • I mean, I’ve got one of those “so simple it’s stupid” solutions. It’s not a pure LLM, but those are probably impossible… Can’t have an AI service without a server after all, let alone drivers

    Do a string comparison on the prompt, then tell the AI to stop.

    And then, do a partial string match with at least x matching characters on the prompt, buffer it x characters, then stop the AI.

    Then, put in more than an hour and match a certain amount of prompt chunks across multiple messages, and it’s now very difficult to get the intact prompt if you temp ban IPs. Even if they managed to get it, they wouldn’t get a convincing screenshot without stitching it together… You could just deny it and avoid embarrassment, because it’s annoyingly difficult to repeat

    Finally, when you stop the AI, you start printing out passages from the yellow book before quickly refreshing the screen to a blank conversation

    Or just flag key words and triggered stops, and have an LLM review the conversation to judge if they were trying to get the prompt, then temp ban them/change the prompt while a human reviews it