Yassin Mohamed

until Go. All developers did was stay where they were.

I love layout algorithms. With these amazing colors the layout of the nodes looks super nice. What are you using in this case ? is it generated by dot (like everyone does) or something else ?

every new year i get the latest nixos .11 release . that takes about 10 min as well with all my tools and configs reinstalled. This stopped me from distro hoping!

a donkey from the movie shrek is smiling Alt: a donkey from the movie shrek is smiling

nixos be like:

I am tempted to let claude loose on this and generate the equivalent go code !

In your estimation how much should one plan in terms of storage for just storing posts/text with no media from firehose ? 20g /month or 200g /month ?

i am planning on doing this before end of year

I imagine there is a way to make tailscale use a different port. Worst case it only takes one machine/container to drop you anywhere in your network. Then from there bob is your oncle :-)

for me tailscale ssh client from the web console was an amazing discovery. works from my pixel9 and tablets.

whats the big difference between this and wazero ?

part of me is like how many papers could be written from this if you were in academia instead of ... just doing it :-)

Big plus: i have identical setups across 3 laptops + 3 servers (Hostinger/DigitalOcean). Setting up or wiping a machine takes 5 min. Big minus: NixOS hates tools like npm that download random binaries and try to run it locally 😤

the numbers of times i stopped myself from throwing it and building a cli. the worst part is when you have to go look for models to download ... oh boy !

GitHub - radiatingreverberations/comfyui-docker: ComfyUI Docker Images ComfyUI Docker Images. Contribute to radiatingreverberations/comfyui-docker development by creating an account on GitHub.

i meant how many tasks can run concurrently. bad wording!

Can you run docker on these ? i used [1] to run comfyui on docker. Compiling took forever however. I hope you can cross compile

[1] github.com/radiatingrev...

how is the latency on these and how many can you run concurrently

1) summarise a large body of text with small/medium/large model
2) producing image/video using comfyui
3) pick a model that is good at coding

@hailey.at is there a limit to how many records you can fetch from jetstream by a single client ? i am about to switch to firehose.

pastila.nl?003e84a1/52a...

So i was finally able to put something that reads from jetstream using go. it stops reading at about 2000 eps (events per second). i tried on my desktop and on digital ocean. in both cases it just dies after about 20s consistently. i am not sure if it is getting disconnected or no

i think i would like a self labeled bot posts/accounts as a proof of good netizens.

but yeah spam was already terrible. even worst now marketing team calls then "cold email" with no shame :-( at least here they are sending to everyone not specifically to you

is it possible to know if a post was already moderated/labeled when its coming from jetstream ? in case you cant do your own moderation what are the alternatives ?

I was planning to play with jetstream locally. this is captivating to say the least !

one could make a custom ranking algorithm for stuff they are not following and/or particular topics.

its measured by how you treat those who are not kind to you.

its human to be kind to those who are already kind to you .. thats basically ground zero !

oh i thought that was embedding for the user content like text/links for search indexing. i see you are using something like open/elastic search. is the intent to move away in the future and move to a vector db based search or similar ?

big props for the stuff you folks are sharing! this is super refreshing !

Can you share what the plan is for the embedding part i see in the diagram ? in terms of what backend you are considering (qadrant, pinecore etc, .. ) or all in-house vector db/indexing ? i cant wait to see if that is .

super cool! can you share the latency requirements internally ? and maybe the hw you are running this ?

i had to make my own go/cue based config that spits out the nixos configs for my deskotps.laptops and digital-ocean servers.

nix is amazing for this sort of stuff but a go cli on top of it makes it super simple for me.

Do you have the source for these diagrams published ? I am being (very) lazy :-)

Open Social — overreacted The protocol is the API.

i wrote about atproto and why it matters

this seems like making or a uuid with Collision-free and coordination-free properties. last i looked i stumbled on [1]. do you need other properties like a bigger random payload and/or better PRNG maybe ?

[1]https://github.com/segmentio/ksuid

is there anything that can draw something like this from a goroutine trace/log ? that would be amazing !

it used to take me a day to format my machine once per year or install a new laptop/desktop when i buy new one. because linux ... with nix its 5min now !
i get all my tools with i3wm preconfigured. once i move to hyprland it would even be smoother on all desktops/servers i own.